Artificial intelligence-based sepsis prediction system and method

Abstract

There is an urgent need to identify measurable, disease-specific biomarkers that serve as indicators of the risk of progression to sepsis or its early onset, and to develop methods for using such markers in screening. [Solution] This disclosure encompasses a system and method for determining the risk of developing sepsis in an individual who has sepsis or is at risk of sepsis, including within about 24 hours of obtaining measurements and / or information of one or more factors related to the individual. In specific embodiments, one or more input features indicating whether an individual is at low, moderate, high, or very high risk of developing sepsis, such as one or more demographic attributes, patient assessments, vital signs, blood test measurements, chemo-panel measurements and / or sepsis-related biomarkers, may be used to generate an output such as a quantitative value.

Description

Technical Field

[0001] Cross - reference to Related Applications

[0001] This application claims the priority and benefit of U.S. Provisional Patent Application No. 63 / 459,568, filed on April 14, 2023, the content of which is incorporated herein by reference in its entirety.

[0002] Field

[0002] This disclosure generally relates to methods, compositions, and systems related to at least medicine, physiology, and biology. More particularly, this disclosure relates to the risk assessment of the onset of sepsis.

Background Art

[0003] Background

[0003] In many diseases and conditions, favorable treatment and / or prevention outcomes are strongly correlated with early and / or accurate diagnosis of the disease or condition. However, effective early diagnosis and treatment methods are often lacking, and as a result, the administration may be too late, inappropriate administration may be carried out, or it is often administered to individuals who will not receive its benefits.

Summary of the Invention

Problems to be Solved by the Invention

[0004]

[0004] Sepsis is a serious condition and often life-threatening disease, requiring urgent and comprehensive care. Treatment usually begins with antibiotics or similar drug therapy, depending on the type of infection. Sepsis can result from infections caused by a wide variety of organisms, making early prediction and diagnosis particularly difficult for effective intervention. Sepsis is an excessive and uncontrolled inflammatory response in the body caused by an inappropriate reaction of the body's immune system to pathogenic organisms. Furthermore, organisms are not present in large numbers in the body fluids or on accessible sites of the affected person, thus increasing the difficulty of diagnosis. Therefore, it is necessary to identify biomarkers that indicate the risk of sepsis or early onset, regardless of the causative pathogen, in order to enable early and effective intervention. Differentiating between patients at risk of developing sepsis and those who are not is also helpful in managing the disease state. Therefore, there is an urgent need for the identification of biomarkers that are measurable, specific to the disease state, and that indicate the risk of progression to sepsis or its early onset, as well as methods for using such markers in screening. [Means for solving the problem]

[0005] overview

[0005] Embodiments of the present disclosure include a method for generating an index of the risk of sepsis, comprising: receiving at least one input feature corresponding to a patient; analyzing the at least one input feature using a machine learning model to generate a predictive score indicating the probability that the patient will have or develop sepsis within a given period of time based on the at least one input feature, wherein the at least one input feature is related to sepsis.

[0006]

[0006] In some embodiments, a method for providing a diagnosis of sepsis includes receiving data corresponding to a patient, the data including the concentrations of one or more input features; analyzing the data using at least one machine learning model to generate a predictive score indicating the probability that the patient has or will develop sepsis within a given period, based on at least one input feature selected from a group of input features listed in any one of Tables 8-13, the groups of input features listed in Tables 8-13 being related to sepsis, and the groups of input features are listed in Tables 8-13 in terms of their relative importance to the predictive score; and generating a diagnostic output based on the predictive score. In specific embodiments, generating a diagnostic output includes determining that the predictive score falls into one of four categories, and generating a diagnostic output that, based on the category into which the predictive score falls, the patient has a low, moderate, high, or very high probability of having or developing sepsis within a given period. In some embodiments, the predictive score is in the range between 0 and 1. In some embodiments, the four categories include a low category, a medium category, a high category, and a very high category, where the low category is defined as having a predicted score below a first threshold, the medium category is defined as having a predicted score between a first threshold and a second threshold, the high category is defined as having a predicted score between a second threshold and a third threshold, and the very high category is defined as having a predicted score above a third threshold.

[0007]

[0007] In a detailed embodiment, there is a method for generating a diagnosis of sepsis, comprising receiving input features corresponding to a patient, analyzing the input features using a machine learning model to generate a predictive score indicating the probability that the patient has or will develop sepsis within a given period of time, based on at least one input feature selected from a group of input features listed in any one of Tables 8 to 13, wherein the group of input features listed in any one of Tables 8 to 13 is related to sepsis, and the group of input features is listed in Tables 8 to 13 in no particular order of importance. Furthermore, the group of input features listed in Tables 8 to 13 may be considered to be listed in terms of their relative importance to the predictive score. By any method, a diagnostic output may be generated based on the predictive score. In a particular embodiment, the input features include at least one clinical parameter and at least one biomarker listed in the group of input features listed in any one of Tables 8 to 13.

[0008] Brief explanation of the drawing

[0008] For a more detailed understanding of the principles and advantages disclosed herein, refer hereto to the following description in conjunction with the accompanying drawings. [Brief explanation of the drawing]

[0009] [Figure 1]

[0009] This is a description of the inputs, functions, artifacts, intermediate outputs, and final outputs used and generated in an example of the Sepsis ImmunoScore algorithm. Functions are represented as triggering logic, imputing input features, generating pre-calibration scores, generating SHAP values, calibrating predictions, and generating risk categories. Intermediate data are reflected as input features (unimputed) and pre-calibration Sepsis ImmunoScore. Artifacts are reflected as cube-shaped steps. The final output is represented as input features, feature SHAP values, Sepsis ImmunoScore, and Sepsis ImmunoScore risk categories. [Figure 2]

[0010] This is an exemplary method for generating indicators of sepsis risk according to various embodiments. [Figure 3A]

[0011] An example of SHAP values ​​for Model 1A, used individually in the first example, is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 3B]

[0011] A second example of SHAP values ​​for individually used Model 1A is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 3C]

[0011] A third example of SHAP values ​​for individually used Model 1A is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 4A]

[0012] An example of SHAP values ​​for Model 2, used individually in the first example, is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 4B]

[0012] An example of SHAP values ​​for Model 2, used individually, is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 4C]

[0012] A third example of SHAP values ​​for individually used Model 2 is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 5A]

[0013] An example of SHAP values ​​for Model 3, used individually in the first example, is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 5B]

[0013] An example of SHAP values ​​for a second individually used Model 3 is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 5C]

[0013] A third example of SHAP values ​​for individually used Model 3 is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 6A]

[0014] An example of SHAP values ​​for Model 4, used individually in the first example, is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 6B]

[0014] An example of SHAP values ​​for a second individually used Model 4 is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 6C]

[0014] A third example of SHAP values ​​for individually used Model 4 is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 7A]

[0015] An example of SHAP values ​​for Model 5, used individually in the first example, is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 7B]

[0015] An example of SHAP values ​​for a second individually used model 5 is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 7C]

[0015] A third example of SHAP values ​​for individually used Model 5 is provided. Values ​​that increase the risk of sepsis are represented by bars to the right, and values ​​that decrease the risk of sepsis are represented by bars to the left. [Figure 8]

[0016] These are block diagrams of computer systems according to various embodiments. [Figure 9A]

[0017] Examples of non-restrictive two-group or two-bin features applied to Model 1A according to various embodiments are provided. [Figure 9B]

[0018] Illustrate non-limiting examples of three-group or three-bin features applied to Model 1A according to various embodiments.

Best Mode for Carrying Out the Invention

[0010]

[0019] It should be understood that these figures are not necessarily drawn to scale, and the objects in the figures are not necessarily drawn to scale in relation to each other. These figures are a depiction intended to bring clarity and understanding to various embodiments of the devices, systems, and methods disclosed herein. The same reference numerals are used throughout the drawings, as far as possible, to refer to the same or similar parts. Further, it should be understood that the drawings are in no way intended to limit the scope of the present teachings.

[0011] Detailed Description I. Exemplary Explanation of Terms

[0020] As used herein, "a" or "an" can mean one or more. As used in the claims of this specification, "a" or "an" when used in conjunction with the phrase "comprising" can mean one or more than one. Some embodiments of the present disclosure can consist of or consist essentially of one or more elements, method steps, and / or methods of the present disclosure. Any method or composition described herein can be carried out with respect to any other method or composition described herein, and it is contemplated that different embodiments can be combined.

[0012]

[0021] The use of the term “or” in the claims is used to mean “and / or” unless it is expressly indicated that it refers only to the other option or that the other options are mutually exclusive; however, this disclosure supports the definitions of the other option only and “and / or.” For example, “x, y and / or z” may mean “x” alone, “y” alone, “z” alone, “x, y and z,” “(x and y) or z,” “x or (y and z),” or “x, or y, or z.” It is particularly intended that x, y, or z may be specifically excluded from embodiments. When used herein, “another” may mean at least two or more.

[0013]

[0022] The term "one" means two or more.

[0014]

[0023] As used herein, the term “plural” may mean two, three, four, five, six, seven, eight, nine, ten or more.

[0015]

[0024] As used herein, the term “set of” means one or more items. For example, a set of items includes one or more items.

[0016]

[0025] When used herein, the expression “at least one of ~” means, when used with a list of items, that one or more different combinations of the listed items may be used, and only one of those items in the list may be required. An item can be a specific object, thing, step, operation, process, or category. In other words, “at least one of ~” means that any combination of items or several items from the list may be used, but not all items in the list are required. For example, “at least one of item A, item B, or item C” means, without limitation, item A; item A and item B; item B; item A, item B, and item C; item B and item C, or item A and C. In some cases, “at least one of item A, item B, or item C” means, without limitation, two item A, one item B, and ten item C; four item B and seven item C, or any other preferred combination.

[0017]

[0026] As used herein, “substantially” means sufficient to serve its intended purpose. Thus, the term “substantially” allows for small, minor variations from absolute or perfect conditions, dimensions, measurements, results, etc., which are foreseeable to those skilled in the art but do not affect the overall performance to a degree that is acceptable. When used in relation to numerical values, parameters, or characteristics that can be expressed numerically, “substantially” means within 10 percent.

[0018]

[0027] Throughout this specification, unless otherwise specifically required by context, the phrases “comprise,” “comprises,” and “contains” imply that the described steps or elements or groups of steps or elements are included, but not that any other steps or elements or groups of steps or elements are excluded. “Consists of” means that anything following the expression “consists of” is included and limited to it. Thus, the expression “consists of” indicates that the listed elements are required and essential, and no other elements may be present. “Essentially consists of” means that any elements listed after this expression are included, and other elements are limited to those that do not interfere with or contribute to the activity or action specified in the disclosure for the listed elements. Thus, the expression “essentially consists of” indicates that the listed elements are required and essential, but other elements are not optional and may or may not be present, depending on whether they affect the activity or action of the listed elements.

[0019]

[0028] Throughout this specification, any reference to “one embodiment,” “embodiment,” “detailed embodiment,” “related embodiment,” “specific embodiment,” “additional embodiment,” or “further embodiment,” or any combination thereof, means that the specific features, structures, or characteristics described in relation to that embodiment are included in at least one embodiment of the present invention. Therefore, although the aforementioned expressions appear in various places throughout this specification, not all of them refer to the same embodiment. Furthermore, those specific features, structures, or characteristics may be combined in any preferred manner in various embodiments.

[0020]

[0029] The term "artifact," as used herein, refers to the digital output produced by an algorithm.

[0021]

[0030] As used herein, “artificial neural network” or “neural network” (NN) may refer to a mathematical algorithm or computational model that mimics a group of interconnected artificial nodes or neurons that process information computationally based on connectionist methods. A neural network may also be called a neural network and uses one or more nonlinear unit layers to predict an output for a given input. Some neural networks include one or more hidden layers in addition to an output layer. The output of each hidden layer is used as the input to the next layer in the network, i.e., the next hidden layer or output layer. Each layer of the network produces an output from a given input according to the current values ​​of the parameters of its respective set. In various embodiments, a reference to “neural network” may refer to one or more neural networks.

[0022]

[0031] Neural networks process information in two ways: when they are being trained, they are in training mode; and when they put what they have learned into practice, they are in inference (or prediction) mode. Because neural networks learn through a feedback process (e.g., backpropagation), the network can adjust the weight coefficients of individual nodes in the intermediate hidden layers (correcting their behavior) so that the output fits the output of the training data. In other words, a neural network learns from the training data (learning examples) it is supplied with, and eventually learns how to arrive at the correct output even when presented with a new range or set of inputs. Neural networks may include, for example, feedforward neural networks (FNNs), recurrent neural networks (RNNs), modular neural networks (MNNs), convolutional neural networks (CNNs), residual neural networks (ResNets), ordinary differential equation neural networks (neural-ODEs), or at least one of the other types of neural networks.

[0023]

[0032] The terms “declare,” “declare,” or “declaring,” as used herein, refer to the physician’s declaration process, typically involving one or more physicians reviewing a patient’s case through a thorough examination of medical records and making a retrospective determination of what has occurred. The term “biomarker,” as used herein, generally refers to any measurable substance taken as a sample from a subject whose presence indicates a certain phenomenon. Non-limiting examples of such phenomena include the subject’s age, disease state, pathological condition, or exposure to a compound or environmental condition. In various embodiments described herein, biomarkers included in analyses that also utilize age-related biomarkers may be used for diagnostic and / or therapeutic purposes (e.g., to diagnose health conditions, disease conditions). The term “biomarker” may be used synonymously with the term “marker.”

[0024]

[0033] The term “patient” as used herein generally refers to a mammalian subject. Mammals may be humans or, but are not limited to, animals including horses, pigs, dogs, cats, ungulates, and primates. In one embodiment, the individual is a human. The methods and uses described herein are useful for both medical and veterinary use. “Patient” refers to a human subject unless otherwise specified.

[0025]

[0034] When used herein, the term "training data" generally refers to data that can be input into any system or process capable of making predictions using a model, statistical model, algorithm and existing data.

[0026]

[0035] As used herein, “model” may include one or more algorithms, one or more mathematical techniques, one or more machine learning algorithms, or a combination thereof.

[0027]

[0036] As used herein, “machine learning” can be a practice technique that uses algorithms to parse data, learn from it, and then make decisions or predictions about something in the world. Machine learning uses algorithms that can learn from data rather than relying on rule-based programming. Machine learning algorithms may include parametric models, nonparametric models, deep learning models, neural networks, linear discriminant analysis models, quadratic discriminant analysis models, support vector machines, random forest algorithms, nearest neighbor algorithms, composite discriminant analysis models, k-means clustering algorithms, supervised models, unsupervised models, logistic regression models, multivariate regression models, penalized multivariate regression models, or other types of models.

[0028]

[0037] The term "LOINC," as used herein, refers to a Logical Observation Identifier, Name, and Code (LOINC), a standard that facilitates the exchange and pooling of results, such as clinical laboratory values ​​or vital signs, in clinical practice, outcome management, and research.

[0029]

[0038] When used herein, the term “Sepsis ImmunoScore” refers, in some embodiments, to the probability that a patient has or will develop sepsis as defined by Sepsis-3 within 24 hours of the time of requesting the Sepsis ImmunoScore.

[0030]

[0039] The term “Sepsis ImmunoScore algorithm,” as used herein, refers to the process of generating a Sepsis ImmunoScore using a set of values ​​measured in a patient’s live stream.

[0031]

[0040] When used herein, the term "SHAP value" refers to the Shapley value, which is a value related to the degree of contribution of each input feature to the final prediction.

[0032] II. Overview of Embodiments

[0041] Embodiments of this disclosure include generating any type of index or output, such as a quantitative output (e.g., a score) or a qualitative output, that reflects individuals who have sepsis or are at risk of developing sepsis (e.g., a higher risk than the general population), including at least some cases that develop sepsis within a specific period of time. Sepsis may be any type, including sepsis, severe sepsis, or septic shock. In specific embodiments, the output may or may not be live-streamed. In some embodiments, individuals have an infection as defined by the presence of one of the following criteria:

[0033]

[0042] 1. Four or more quality days (QADs) of antibiotic use that meet the criteria defined by the Centers for Disease Control and Prevention (CDC).

[0034]

[0043] 2. Positive culture after microbiological testing or excluding known contamination (known contamination in urine cultures may be defined as "Lactobacillus, Corynebacterium species, Gardnerella species, α-hemolytic streptococci, and aerobic bacteria").

[0035]

[0044] The QAD rules follow the instructions in the CDC Hospital Tool Kit for Adult Sepsis infection rules (https: / / www.cdc.gov / sepsis / pdfs / Sepsis-Surveillance-Toolkit-Mar-2018_508.pdf).

[0036]

[0045] In detailed embodiments, the methods and compositions of this disclosure relate to the probability or risk that a patient will have or develop sepsis as defined by Sepsis-3 (Singer et al., JAMA. 2016 Feb 23; 315 (8): 801-810 (which is incorporated herein by reference)) within 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 hours or not exceeding the time of patient evaluation. The output of the patient analysis may be referred herein to as the Sepsis ImmunoScore.

[0037]

[0046] The outputs of these methods may or may not be continuous values, such as a continuous value between 0 and 100. The outputs may, in at least some cases, be displayed to the user of a medical device program (SaMD). In some embodiments, the output may be interpreted as the probability that a patient has or develops sepsis as defined by Sepsis-3 within 24 hours from the time of requesting the Sepsis ImmunoScore. The Sepsis ImmunoScore may differ from an uncalibrated Sepsis ImmunoScore prediction in some embodiments by imposing a calibration that refines the interpretability of the uncalibrated Sepsis ImmunoScore prediction.

[0038]

[0047] In various embodiments, the individual requires a diagnosis of sepsis or determination of the risk of sepsis. The individual may be of any age, but in specific embodiments, the individual is at least about 65 years old or less than about 1 year old. The individual may be hospitalized or have recently been hospitalized. The individual may or may not be immunocompromised. In certain embodiments, the individual may have or be at risk of having an infection, including bacterial, viral, fungal, prion, or parasitic infections. The individual may or may not have chronic diseases, such as heart disease, lung disease, cancer, kidney disease, hepatitis, and / or diabetes, but these are merely examples. In specific embodiments, the individual may have infections such as SARS-CoV-2, influenza, meningitis, pneumonia, tuberculosis, Escherichia coli (E. coli), or C. difficile.

[0039]

[0048] The individual may have one or more symptoms of sepsis, including a high or weak heart rate, fever, chills or abnormal chills, confusion or disorientation, shortness of breath, severe pain or discomfort, and / or sticky or sweaty skin.

[0040]

[0049] Individuals may reside temporarily or permanently in facilities such as nursing homes, rehabilitation centers, advanced nursing facilities, and hospitals, together with other people.

[0041]

[0050] In a specific embodiment, the individual is an adult aged 18 or older, and is accompanied by a suspected serious infection, as defined by an initial blood culture request in an emergency room or hospital setting.

[0042]

[0051] In specific embodiments, the methods and systems of the Disclosure utilize one or more of the following: one or more demographic measurements, one or more patient assessments, one or more vital signs, one or more blood test values, one or more chemical test values, and / or one or more sepsis-related biomarker concentrations. In detailed embodiments, the methods and systems of the Disclosure utilize some or all of the following: (1) one or more demographic measurements, (2) one or more patient assessments, (3) one or more vital signs, (4) one or more blood test values, (5) one or more chemical test values, and / or (6) one or more sepsis-related biomarker concentrations. Examples of demographic measurements include at least age. Examples of patient assessments include at least the Glasgow Coma Scale (GCS). Examples of vital signs include systolic blood pressure, diastolic blood pressure, body temperature, respiratory rate, heart rate, blood oxygen saturation (SpO2), and / or inspired oxygen fraction (FiO2). Examples of blood test values ​​include white blood cell count, platelet count, lymphocyte count, and / or neutrophil count. Examples of chemical test values ​​include creatinine, blood urea nitrogen, potassium, chloride, total carbon dioxide (TCO2), sodium, albumin, and / or bilirubin. Examples of sepsis-related biomarkers include C-reactive protein, procalcitonin, and / or lactate. Other examples of input features are provided in Tables 8-13.

[0043]

[0052] In detailed embodiments, combinations of patient characteristics are measured, assayed, analyzed, obtained and / or considered to determine the presence of sepsis or sepsis at a particular stage, or the risk of having it. In some embodiments, combinations of patient characteristics are measured, assayed, analyzed, obtained and / or considered to determine the risk of developing sepsis within the following 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48 hours or longer. Depending on the circumstances, combinations of patient characteristics may be measured, assayed, analyzed, obtained and / or considered to determine the risk of developing sepsis within the following 1-48, 1-36, 1-24, 1-12, 1-6, 1-2, 2-48, 2-36, 2-24, 2-12, 2-6, 6-48, 6-36, 6-24, 6-12, 12-48, 12-36, 12-24, 24-48, 24-36, or 36-48 hours.

[0044]

[0053] Patient characteristics as input features may, in certain embodiments, be measured, assayed, analyzed, obtained, and / or considered in a detailed order, such as a specific predetermined order. The measurement, assay, analysis, acquisition, and / or consideration of one or more characteristics for one patient may, in specific cases, differ from the measurement, assay, analysis, acquisition, and / or consideration of one or more characteristics for another patient. In some embodiments, one or more characteristics are measured, assayed, analyzed, obtained, and / or considered before one or more other characteristics. In certain embodiments, the nature of one or more characteristics may trigger the need for consideration or discussion of one or more other characteristics, and such developments may occur in real time. In some embodiments, the determination of whether an individual has sepsis, has sepsis at a particular stage, or is at risk of sepsis (including, as described above, whether or not it is present within a particular time frame) may be made only after measuring, assaying, analyzing, obtaining and / or considering at least, exactly, or at most, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, or 58 or more patient characteristics.

[0045]

[0054] In specific embodiments, one or more of the following patient characteristics, namely age, albumin, total bilirubin, blood urea nitrogen, C-reactive protein, chloride, creatinine, diastolic blood pressure, heart rate, lactate, lymphocytes, neutrophils, platelets, potassium, procalcitonin, respiratory rate, sodium, SpO2, systolic blood pressure, body temperature, total carbon dioxide, and white blood cell count, are measured, assayed, analyzed, obtained, and / or considered to determine whether the patient has sepsis, is at risk of developing sepsis, or has a specific stage of sepsis. In specific embodiments, the following patient characteristics are considered: age, albumin, angiopoietin-1, angiopoietin-2, total bilirubin, blood urea nitrogen, C-reactive protein, CCL2 / MCP-1, CCL20 / MIP-3α, CCL4 / MIP-1β, chloride, creatinine, CX3CL1 / fractalkine, CXCL10 / IP-10, diastolic blood glucose, E-selectin, factor III / tissue factor, Flt-3 ligand, G-CSF, GM-CSF, granzyme B, heart rate, IFN-α, IFN-γ, IL-10, IL-15, IL-1β / IL-1F2, IL-1ra / IL-1F3, IL- 2. One or more of the following are measured, assayed, analyzed, obtained and / or considered to determine whether the patient has sepsis, is at risk of developing sepsis, or has a specific stage of sepsis. Various patient characteristics that may be available, such as those in any one of Tables 8-13, are provided elsewhere in this specification.

[0046]

[0055] In detailed embodiments, patient characteristics are input feature values ​​for a machine learning algorithm for sepsis risk diagnosis, sepsis staging, or sepsis risk prediction. In at least some embodiments, upon receiving a request (e.g., from a clinician's input device) to trigger a machine learning model to begin the process of generating a decision or prediction, one or more systems provided begin monitoring input feature values ​​from one or more measuring devices or user input devices. In other words, in such embodiments, the systems provided do not continuously monitor the input feature values, such as by constantly receiving them, but rather may monitor the input feature values ​​upon receiving a request until either the machine learning model generates a prediction or it is determined that the machine learning model cannot generate a prediction based on the available input feature values. In other embodiments, the systems provided may continuously monitor the input feature values. Input feature values ​​collected by measuring devices or user input devices are received (e.g., almost instantaneously) by one or more systems (e.g., computing devices) as soon as the input feature values ​​are collected (e.g., measured), so that one or more systems receive the input feature values ​​in real time. Once input feature values ​​are collected, the logic trigger determines whether and when all input features to the machine learning model have available values ​​(which may include value interpolation for the input features) that enable them to generate predictions.

[0047] III. Algorithm Description and Outline of Exemplary Workflow

[0056] In a detailed embodiment, the Sepsis ImmunoScore algorithm is a process that generates the Sepsis ImmunoScore and its auxiliary components using a set of live stream measurements. In a specific embodiment, the input to the algorithm is, for example, live stream measurements stored in a database, including the core input to the algorithm. The core output of the algorithm may be:

[0048]

[0057] 1. Sepsis ImmunoScore (Sepsis Risk Score)

[0049]

[0058] 2. Sepsis ImmunoScore Risk Stratification Categories

[0050]

[0059] 3. Input Features a. value b. Completed (True / False) c.SHAP value

[0051]

[0060] In detailed embodiments, the algorithm includes multiple functional components and artifacts that work together to produce an output, which can be considered Sepsis ImmunoScore and its auxiliary components. These functional components and artifacts reside within the SaMD backend and R engine module specified in U.S. Provisional Patent Application No. 63 / 459,521, filed April 14, 2023 (which is incorporated herein by reference in its entirety). For a visual representation of an example of the Sepsis ImmunoScore algorithm process, see Figure 1.

[0052]

[0061] For a brief description of some functional components or artifacts of at least one example algorithm as depicted in Figure 1, please refer to Table 1, which is explained in the table below.

[0053] [Table 1-1]

[0054] [Table 1-2]

[0055] [Table 1-3]

[0056] i. Streaming data and input features

[0063] In detailed embodiments, the Sepsis ImmunoScore algorithm receives and uses a variety of different streaming data features, including one or more demographic measurements, patient assessments, vital signs, blood test values, chemical test values, and sepsis-related biomarker concentrations. These input features may, in certain examples, originate from clinical findings and / or FDA-approved devices.

[0057]

[0064] In specific embodiments, Sepsis ImmunoScore uses LOINC code to identify streaming data features. In the software, a basic LOINC code may be provided for each feature. Because the LOINC database is dynamic in that there are differences in the implementation of input features, in certain embodiments, the relationships between input features and LOINC code may be updated during post-release installation and deployment.

[0058]

[0065] In specific embodiments, streaming data features can be transformed into input features using a function that triggers the logic of the algorithm. Each input feature may have an associated input feature selection methodology, which determines how to select the input features from the set of streamed measurements. In some embodiments, the input feature selection methodology is intended to be fixed after the release of Sepsis ImmunoScore's SaMD. A detailed description of this process is provided in U.S. Provisional Patent Application No. 63 / 459,521, filed April 14, 2023, which is incorporated herein by reference in its entirety. Examples of input features are listed in Tables 2 and 8-13.

[0059] [Table 2-1]

[0060] [Table 2-2]

[0061]

[0067] In detailed embodiments, the output is used to place individuals into one of four risk stratification categories: low, moderate, high, or very high. The risk stratification categories may be related to the occurrence of the following adverse clinical events based on the Sepsis ImmunoScore request time.

[0062]

[0068] 1. In-hospital mortality rate

[0063]

[0069] 2. Readmission within 30 days

[0064]

[0070] 3. Transfer to the ICU for treatment within 24 hours.

[0065]

[0071] 4. Mechanical ventilation within 24 hours

[0066]

[0072] 5. Use of vasopressors within 24 hours (e.g., norepinephrine, vasopressin, epinephrine, dopamine, angiotensin II, terlipressin, cerepressin)

[0067]

[0073] The stratified bands can be bounded by three thresholds identified during the development of the Sepsis ImmunoScore algorithm. The calculations for identifying risk stratification categories are described below.

[0068]

[0074] During the development of the Sepsis ImmunoScore algorithm, we generated receiver operating characteristic (ROC) curves using out-of-bag, uncalibrated predictions for the training labels. Using these ROC curves, we identified three thresholds based on the following criteria: 1. Lowest threshold - Identify the highest sensitivity threshold while maintaining a false positive rate of 50%. 2. Optimal threshold - The threshold is identified as the point at the "upper left corner" of the ROC curve, i.e., the point on the ROC curve where the Euclidean distance to (0,1) is minimized. 3. 95th percentile threshold - The threshold corresponding to the 95th percentile of the sepsis risk score within the training dataset.

[0069]

[0075] In a specific embodiment, possible output values ​​are provided in Table 3.

[0070] [Table 3]

[0071]

[0077] Table 4 provides examples of diagnostic interpretations and recommended responses for each of the potential risk stratification categories as outputs.

[0072] [Table 4]

[0073]

[0079] In various embodiments, an assessment of the presence or risk of sepsis in an individual is provided based on one or a variety of factors included in any one of Tables 8-13 (these may be referred to as input features, given the diversity of their properties (e.g., demographic attributes, patient assessment, vital signs, blood test measurements, chemical panel measurements, sepsis-related biomarkers, etc.)), one or a variety of factors included in any one of Tables 8-13 (these may be referred to as biomarkers, given their protein-like properties), or both. In some embodiments, one or more factors in any one of Tables 8-13 are used to generate an output that determines the presence or risk of sepsis.

[0074] Print.A. Example of input features

[0080] In a detailed embodiment, one or more of the following input features are used to determine whether an individual has sepsis or is at risk of developing sepsis, such as within 24 hours of measurement.

[0075] i. Demographic measurements

[0081] Age can be calculated as [current date - date of birth]. Since the incidence of sepsis increases with the patient's age, age has traditionally been considered a risk factor. This may be related, in specific embodiments, to an increase in chronic conditions that are significant in the progression or presence of sepsis.

[0076] ii. Patient evaluation

[0082] The Glasgow Coma Scale (GCS) is a cognitive impairment assessment administered by a qualified professional. The GCS is recorded on a scale of 3 to 15 points and is determined by a healthcare professional. The GCS is intended to assess the central nervous system component of the Sequential Organ Failure Assessment (SOFA) score (Jones et al., Crit Care Med. 2009 May; 37(5): 1649-1654 (which is incorporated herein by reference in its entirety)). By monitoring an individual's condition with the SOFA score, it is possible to determine the degree of organ function or failure rate of the individual. The score is based on six different scores, one for each of the respiratory, cardiovascular, hepatic, coagulation, renal, and nervous systems. In specific embodiments, sepsis is when an individual has organ dysfunction (acute change with a SOFA score ≥ 2), and septic shock is when an individual has organ dysfunction (acute change with a SOFA score ≥ 2) and refractory hypotension.

[0077] iii. Vital signs 1. Systolic blood pressure

[0083] Systolic blood pressure is the pressure exerted when the heart beats and blood is ejected into the arteries. This can be measured using a machine such as a blood pressure monitor. Systolic blood pressure is used to calculate mean arterial pressure, which is intended to assess the cardiovascular component of the SOFA score.

[0078] 2. Diastolic blood pressure

[0084] Diastolic blood pressure is the pressure in the arteries when the heart is at rest between beats. It can be measured using a machine such as a blood pressure monitor. Diastolic blood pressure is used to calculate mean arterial pressure, which is intended to assess the cardiovascular component of the SOFA score.

[0079] 3. Body temperature

[0085] Body temperature is the inherent degree of heat in the body. Body temperature can be measured in various ways. One commonly used method is with a thermometer. Body temperature can be used to determine whether a patient has a fever. Fever patterns can be used to identify possible causes of infection, as remittent fever (a fever that fluctuates by more than 1.1°C and never returns to normal) is often caused by infection.

[0080] 4.Respiration rate

[0086] Respiratory rate is defined as heart rate per minute. Respiratory rate can be measured manually or using a device such as a pulse oximeter. Abnormal respiratory rate may indicate a disturbance in the amount of available oxygen a patient has. Tachypnea (increased respiratory rate) is common in sepsis patients and is associated with increased 28-day mortality in sepsis patients. Respiratory rate is also associated with SIRS, which was part of the old Sepsis-2 definition.

[0081] 5. Heart rate

[0087] Heart rate is measured as the number of heartbeats within a given period. Heart rate can be measured manually or using a device such as a heart rate monitor. Irregular heartbeats may be associated with poor blood circulation and either low or high blood pressure. Heart rate is also relevant to SIRS, which is part of the old Sepsis-2 definition.

[0082] 6.Blood oxygen saturation (SpO2)

[0088] SpO2 is a measure of the amount of oxygen-carrying hemoglobin in the blood compared to oxygen-depriving hemoglobin. SpO2 can be calculated and measured in many ways, one of which is the use of a pulse oximeter. SpO2 is a non-invasive method for estimating arterial oxygen saturation (SaO2), which is used in the calculation of PaO2 / FiO2 in the respiratory system SOFA score calculation.

[0083] 7. Inhaled oxygen fraction (FiO2)

[0089] FiO2 is the concentration of oxygen in the inhaled gas mixture. The gas mixture of room air has a 21% inhaled oxygen fraction. An increase in FiO2 may indicate that the patient requires a higher percentage of oxygen. FiO2 is used in the PaO2 / FiO2 calculation for the respiratory system SOFA score.

[0084] iv. Blood test results 1.White blood cell count

[0090] White blood cell count is a measure of the number of white blood cells in a patient's blood. White blood cell count can be measured using a blood test. An abnormal white blood cell count may indicate that a person is dealing with an infection.

[0085] 2. Platelet count

[0091] Platelet count is a measure of the number of platelet cells in a patient's blood. Platelet count can be measured using blood tests. An abnormal platelet count may indicate that an individual has difficulty coagulating. Platelets are involved in both hemostasis and immune responses and play a crucial role in sepsis. Platelets are used to assess the coagulation component of the SOFA score.

[0086] 3. Lymphocyte count

[0092] Lymphocyte count is a measure of the number of lymphocyte cells in a patient's blood. Lymphocyte count can be measured using a blood test. Lymphocytes are a type of white blood cell that is part of the immune system. An abnormality in lymphocyte count may indicate the presence of an infection. Lymphocytes can be used in combination with neutrophils to calculate the neutrophil-to-lymphocyte ratio (NLR). NLR is a metric associated with mortality in sepsis patients, as non-survivors exhibit significantly higher NLRs.

[0087] 4. Neutrophil count

[0093] Neutrophil count is a measure of the number of neutrophil cells in a patient's blood. Neutrophil count can be measured using a blood test. Neutrophils are a type of granulocyte, a type of white blood cell that is part of the immune system. An abnormality in lymphocyte count can indicate the presence of infection. Lymphocytes can be used in combination with neutrophils to calculate the neutrophil-to-lymphocyte ratio (NLR). NLR is a metric associated with mortality in sepsis patients, as non-survivors exhibit significantly higher NLRs.

[0088] v. Chemical test results 1. Creatinine

[0094] Creatinine is defined as the concentration of creatinine in the blood. Blood creatinine levels can be measured using a chemical panel. High blood creatinine levels may indicate that the renal system is not effectively filtering creatinine into the urine for excretion. Creatinine is used to assess the renal system component of the SOFA score.

[0089] 2. Blood urea nitrogen

[0095] Blood urea nitrogen (BUN) is the amount of blood urea nitrogen remaining in the blood. BUN levels in the blood can be measured using a chemical panel. Blood urea nitrogen is a waste compound transferred from the blood to the urine by the kidneys. High BUN levels may indicate that the renal system is not functioning effectively. Elevated BUN has been associated with the development or presence of sepsis.

[0090] 3. Potassium

[0096] Potassium is measured as the concentration of potassium ions in the blood. Blood potassium levels can be measured by performing chemical tests. Potassium plays a role in maintaining intracellular fluid levels in the body. An imbalance in potassium can lead to fluid imbalances. Fluid imbalances are associated with many patients admitted to the ICU, and fluid resuscitation is effective in treating patients with severe sepsis or septic shock.

[0091] 4. Chloride

[0097] Chloride is measured as the concentration of chloride ions in the blood. Blood chloride levels can be measured by performing chemical tests. Chloride plays a role in maintaining intracellular fluid levels in the body. An imbalance in chloride can lead to fluid imbalances. Fluid imbalances are associated with many patients admitted to the ICU, and fluid resuscitation is effective in treating patients with severe sepsis or septic shock.

[0092] 5. Total Carbon Dioxide (TCO2)

[0098] Total carbon dioxide (TCO2) is a measure of carbon dioxide found in the blood. TCO2 can be measured in the blood using chemical tests. TCO2 can be used to identify metabolic acidosis, a function of the respiratory system. An imbalance in TCO2 may be associated with respiratory disorder. Elevated TCO2 levels have been associated with increased mortality in patients with sepsis.

[0093] 6. Sodium

[0099] Sodium is measured as the concentration of chloride ions in the blood. Blood sodium levels can be measured by performing chemical tests. Sodium plays a role in maintaining intracellular fluid levels in the body. An imbalance in potassium can lead to fluid imbalances. Fluid imbalances are associated with many patients admitted to the ICU, and fluid resuscitation is effective in treating patients with severe sepsis or septic shock.

[0094] 7. Albumin

[0100] Albumin is measured as the concentration of albumin present in the blood. Measuring albumin in the blood can be achieved by performing chemical tests. Albumin is a protein synthesized by the liver that minimizes capillary leakage into body tissues. Disruptions in albumin levels may indicate problems in the hepatic or renal system. Serum albumin has been identified as a predictor of mortality in patients with sepsis.

[0095] 8. Bilirubin

[0101] Bilirubin is measured as the concentration of bilirubin present in the blood. Blood bilirubin measurement can be achieved by performing chemical tests. Bilirubin is a protein produced during the breakdown of red blood cells. In healthy individuals, bilirubin is a waste product filtered by the liver. Bilirubin is used to assess the liver component of the SOFA score calculation.

[0096] vi. Sepsis-related biomarker concentrations 1. C-reactive protein

[0102] C-reactive protein (CRP) is the concentration of CRP in the blood. Measuring CRP in the blood can be achieved by performing chemical tests. CRP is a protein produced by the liver, and its levels increase with inflammation. Studies have shown that combining CRP with body temperature may increase the detection of infections.

[0097] 2. Procalcitonin

[0103] Procalcitonin (PCT) is measured as the concentration of PCT present in the blood. PCT can be measured by clinical tests. PCT is released into the bloodstream by tissues while inflammation is present. Bacterial infections have been shown to have a significant impact on PCT synthesis because bacterial endotoxins and cytokines facilitate PCT secretion by host cells.

[0098] 3. Lactate

[0104] Lactate is measured as the concentration of lactate present in the blood. Lactate can be measured by clinical laboratory tests. Lactate (or lactate) is produced by tissues during hypoxic conditions. Increased blood lactate levels may indicate that a patient is in organ failure or shock. Lactate has emerged as part of the septic shock label in the Sepsis-3 definition.

[0099]

[0105] In some embodiments, individuals at risk of or suspected of having sepsis undergo analysis, measurement, consideration, and / or assay of one or more samples. Examples of samples include blood, plasma, serum, urine, tissue, skin, saliva, sputum, and / or mucus. Levels of any one or more patient characteristics from one or more samples may be used as input features for models incorporated herein. Alternatively or in addition, individuals may be subjected to blood pressure testing, imaging (X-ray, ultrasound scan, or computed tomography (CT) scan), etc.

[0100]

[0106] In some embodiments, multiple input features are selected from one, two, three or more categories, and such features may or may not be used in any model contained herein. In specific embodiments, the multiple input features include at least one input feature from a first group, at least one input feature from a second group, and at least one input feature from a third group. In some embodiments, the first group includes input features in a group consisting of age, sex at birth, race, ethnicity, patient's medical history, patient's current complaints or symptoms, neurological assessment, clinical decision support alerts, clinician or other medical records, diagnostic codes, procedures performed on the patient, patient's current drug therapy, interventions, patient treatment environment, medical imaging data or assessments, electrocardiograms, endoscopy, systolic blood pressure, diastolic blood pressure, body temperature, respiratory rate, heart rate, blood oxygen saturation, inhaled oxygen fractions, and combinations thereof. In some embodiments, the second group includes input features in a group consisting of white blood cell count, lymphocyte count, neutrophil count, platelet count, creatinine, blood urea nitrogen, potassium, chloride, total carbon dioxide, sodium, albumin, bilirubin, lactate, glucose, calcium, hematocrit, hemoglobin concentration, laboratory-derived biomarkers, clinical severity scales and overall scores, biopsy data, microbiological test data, and combinations thereof. In some embodiments, the third group includes input features in a group consisting of procalcitonin, C-reactive protein, protein biomarkers, genomic biomarkers, gene expression or transcriptome biomarkers, composite biomarker scores, and combinations thereof.

[0101]

[0107] In specific embodiments, there are the following categories, and one or more input features from one or more of these categories can be used to determine whether an individual has sepsis or is at risk of developing sepsis within a given time.

[0102]

[0108] Category 1 • Vital signs (e.g., body temperature, heart rate, blood pressure, respiratory rate, oxygen saturation), • Demographic attributes (e.g., age, sex, race, ethnicity), ·Past history, • Current illness or symptoms, • Neurological assessment (Glasgow Coma Scale, Full Outline of Unresponsiveness (FOUR)) • Clinical decision support alerts, • Clinician or other medical records, • Diagnostic code (e.g., ICD-10) • Procedures (including CPT codes), • Current drug therapies (including NDC codes), ·intervention, ·Patient treatment environment (ICU, ED, hospital bed, outpatient treatment, etc.), • Imaging methods (e.g., X-ray, CT scan, MRI, fMRI, ultrasound) • Electrical recording diagrams (e.g., EEG, EKG), • Endoscopic examination

[0103]

[0109] Category 2 • Clinical test values, • Laboratory-derived biomarkers (e.g., monocyte distribution width, Cytovale IntelliSep), • Clinical severity scales and overall scores (e.g., Sequential Organ Failure Assessment (SOFA) score, SOFA component score, Charlson Comorbidity Index, Acute Physiological and Chronic Health Assessment (APACHE) I and II), ·biopsy, • Microbiological testing (culture, PCR testing, antigen testing)

[0104]

[0110] Category 3 • Protein biomarkers (angiopoietin-1, angiopoietin-2, C-reactive protein, cystatin C, D-dimer, E-selectin, fractalkine, FLT3-ligand, GCSF, GDF15, GMCSF, granzyme-B, IFN-α, IFN-γ, IL1-β, IL1-RA, IL-2, IL-4, IL-6, IL-7, IL-8, IL-10, IL-15, IP-1) (Contains 0, lactate dehydrogenase (LDH), lipopolysaccharide-binding protein (LBP), leptin, MCP1, MIP1-α, MIP1-β, MIP3-α, NGAL, pancreatic stone protein, PDL-1, pentraxin-3, procalcitonin, protein C, S100b, TGF-α, thrombomodulin, tissue factor, TNF-α, TRAIL, TREM-1, troponin, VCAM-1, and VEG-F), • Genomic biomarkers (e.g., one or more SNPs in CD14, TLR1, TLR2, TLR4, TLR6, TNF-α, IL-6, factor V, factor XII, FER, MAN2A1), • Gene expression or transcriptome biomarkers (CEACAM4, LAMP1, PLAC8, PLA2G7, CEACAM1, ZDHHC19, C9orf95, GNA15, BATF, C3AR1, KIAA1370, TGFBI, MTCH1, RPGRIP1, HLA-DPB1, nuclear factor kappa B (NF-κB), heme oxygenase-1 (HO-1), FCGR2C, CRP, IL-6, IL-10, IL-12, TNF, LY6E, LAMP3, ISG15, USP18, RSAD2, IFI44L, NCOA7, CMPK2, IFI 44, BATF2, GBP4, IFIH1, GMPR, CXCL10, IFI27, ISG15, CCL2, OAS3, P2RY14, GCH1, CD274, ID01, ARL4A, BPGM, TNFAIP3, TNFAIP6, IL1R1, PROK2, RGS1, CXC R6, RTP4, TLR3, FZD5, KCNJ2, IRAK2, RGS16, CXCL8, OLR1, CCRL2, SPP1, ST3GAL4, CST7, IL1R2, IL4R, GALM, CTLA4, SOCS1, BCL2L1, TGM2, SLC39A8, CA2, PN P, GATA1, SLC1A5, RGS16, CD44, PDGFC, CSF2RA, CXCL1, CSF3R, SOCS3, IL18R1, PIM1, LEPR, IL1B, CSF2RB, ITGB3, STAT1, FAS, EFNA1, IER3, EETS2, SLC2A 3, BCL2A1, PFKFB3, CEBPB, BIRC2, ATF3, TUBB2A, G0S2, MERTK, BIK, QSOX1, PYGL, TGFA, P4HA2, LDHA, IRS2, TSPO, HGF, GADD45A, TGFBR3, CD38, PRF1, FASL G, TIMP3, ANKH, LGALS3, PPP2R5B, GPX1, BNIP3L, BCL2L1, KLF7, FOSL2, ADM, MXI1, SELENBP1, PGF, FOXO3, NEDD4L, SLC2A1, SIAH2, MMP9, CXCR3, CD8A, IF NG, CD8B, JAK2, CCL4, ICAM1, WARS1, KRT1, GPR65, DYRK3, ACHE, ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4,TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, CD24, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2 , including ZCCHC4, DDX6, SENP5, RAPGEF1, DTX2, RELB, DYRK2, CCNB1IP1, TDRD9, ZAP70, ARL14EP, MDC1, ADGRE3, • Composite biomarker scores (e.g., SeptiCyte LAB and Septicyte RAPID, Inflammatix TruVerity)

[0105] Torb.B. Algorithm Output

[0111] In a detailed embodiment, the core output of the Sepsis ImmunoScore algorithm is as follows: 1. Sepsis ImmunoScore (Sepsis Risk Score) 2. Sepsis ImmunoScore risk stratification categories, 3. Input features, a. Value, b. Completed (True / False), c.SHAP value

[0106] i.Sepsis ImmunoScore

[0112] In detailed embodiments, the Sepsis ImmunoScore is a continuous value between 0 and 100 and is the output of a function that calibrates predictions, as specified in other parts of this specification. In specific embodiments, the Sepsis ImmunoScore is intended to be displayed to the user of SaMD. In certain embodiments, the Sepsis ImmunoScore may be interpreted as the probability that a patient will have or develop sepsis as defined by Sepsis-3 within 24 hours from the time of the Sepsis ImmunoScore request. In specific embodiments, the Sepsis ImmunoScore may differ from an uncalibrated sepsis risk score in that it imposes a calibration that refines the interpretability of the uncalibrated sepsis risk score.

[0107] ii. Risk Category

[0113] In a detailed embodiment, the Sepsis ImmunoScore is used to place patients into one of four risk stratification categories: low, moderate, high, or very high. The risk stratification categories may be related to the occurrence of the following adverse clinical events based on the Sepsis ImmunoScore request time. 1. In-hospital mortality rate, 2. Readmission within 30 days. 3. Transfer to the ICU for treatment within 24 hours. 4. Mechanical ventilation within 24 hours. 5. Use of vasopressors within 24 hours

[0108]

[0114] In certain embodiments, risk stratification categories are defined by three thresholds identified during the development of the Sepsis ImmunoScore algorithm. The calculations for determining risk stratification categories are described in other parts of this specification.

[0109] iii. Input features

[0115] In detailed embodiments, the Sepsis ImmunoScore holds the values ​​of the input features used in its calculation, a true or false value indicating whether the values ​​were imputed, and associated SHAP values. SHAP values ​​are numerical values ​​indicating the local feature importance of the Sepsis ImmunoScore for a particular set of input features. In certain embodiments, the use of SHAP values ​​is helpful in interpreting and trusting the Sepsis ImmunoScore. In detailed embodiments, the sum of SHAP values ​​for a single Sepsis ImmunoScore equals the uncalibrated Sepsis ImmunoScore prediction—the mean value of a predefined data set (the uncalibrated Sepsis ImmunoScore prediction). From this, we can observe the effect of summing each input feature on the uncalibrated Sepsis ImmunoScore prediction.

[0110]

[0116] The SHAP value is a characteristic of any newly generated score, rather than a global algorithmic characteristic, and in some embodiments, it localizes the commonly used concept of feature importance for individual scores.

[0111] III.C. Function Components iv. Trigger the logic input Streaming data features output Sepsis determines whether ImmunoScore can generate results. Input features with missing values

[0112] A logic trigger can determine whether it is ready to generate a Sepsis ImmunoScore for a valid Sepsis ImmunoScore request. If the logic trigger indicates that it is ready to generate a Sepsis ImmunoScore, it passes the input features to the interpolation function component, in certain embodiments. Further information regarding the specification and description of the logic trigger can be found in U.S. Provisional Patent Application No. 63 / 459,521, filed April 14, 2023, which is incorporated herein by reference in its entirety.

[0113] v. Interpolate input features input Input features with missing values output Input features without missing values artifact Completion template

[0114]

[0117] The function that imputates input features takes the input features identified from the module that triggers the logic and imputates any missing values. In a specific embodiment, the input feature imputation may be performed in the following steps in order.

[0115]

[0118] 1. Supplement missing FiO2 measurements with 21.

[0116]

[0119] 2. Supplement missing GCS measurements with 15.

[0117]

[0120] 3. Perform bag imputation using the completion template. Further information regarding the completion methodology can be found in other parts of this specification.

[0118] vi.Generating uncalibrated predictions input Input features with missing values ​​imputed. output Uncalibrated Sepsis ImmunoScore. Uncalibrated raw model prediction score. artifact Machine learning models

[0119]

[0121] In a detailed embodiment, the function that generates the pre-calibration prediction uses the output from a function that complements the input features, along with a trained machine learning model, to generate the raw pre-calibration Sepsis ImmunoScore as a continuous value between [0,1]. The machine learning model used may be a random forest model, implemented using the Ranger R package. Further information regarding the machine learning model can be found in other parts of this specification.

[0120] vii. Calibrate predictions input Uncalibrated sepsis risk score output Sepsis risk score artifact Calibration model

[0121]

[0122] The function that calibrates the predictions uses an uncalibrated Sepsis ImmunoScore along with a calibration model to generate a Sepsis ImmunoScore. Further information regarding the calibration model can be found in other parts of this specification.

[0122] viii. Generate SHAP values input Input data (completed) output SHAP value artifact Machine learning models Training data object

[0123]

[0123] In certain embodiments, the function that generates SHAP values ​​uses the complemented input features, the trained machine learning model, and the training object to generate the SHAP value contribution of the input features. In specific embodiments, the calculation of SHAP values ​​uses game-theoretic methods to identify the contribution that features make to the prediction of an observation (Scott, Lundberg M, and Lee Su-In. “A Unified Approach to Interpreting Model Predictions.” 31st Conference on Neural Information Processing Systems, 2017). In specific embodiments, the SHAP values ​​may use the training data objects to identify an estimate of the feature contribution to the training dataset. Due to the computational complexity in calculating SHAP values, the function that generates SHAP values ​​may, in at least some cases, use an implementation of SHAP estimation by the fastshap R package. In some embodiments, the fastshap R package uses 10 rounds of Monte Carlo simulation to estimate the feature contribution of the training data objects with a fixed seed. Applying this estimate to a new observation can, in some embodiments, generate an overall contribution to the prediction.

[0124] ix. Generate risk stratification categories input Sepsis risk score output Risk stratification categories artifact Threshold list

[0125]

[0124] In a particular embodiment, the function that generates risk stratification categories uses a predetermined threshold list to place the Sepsis ImmunoScore sepsis risk score value into one of four categories. The function that generates risk stratification categories places patients into risk categories using the following logic:

[0126] [Table 5]

[0127] 3.D. Artifact

[0125] The Sepsis ImmunoScore algorithm, in certain embodiments, exhibits five artifacts: 1. Completion template, 2. Pre-trained machine learning algorithms, 3. Calibration model, 4. Training data object, 5. Threshold List Includes.

[0128]

[0126] Artifacts can all be generated during the development of the Sepsis ImmunoScore algorithm. In detailed embodiments, the artifacts are intended to be fixed upon the release of the SaMD of Sepsis ImmunoScore.

[0129] i. Completion templates

[0127] A completion template is, in a specific embodiment, a pre-trained template used to complete the values ​​of missing input features. The completion template may be stored as an R data structure file. In one embodiment, the purpose of this artifact is to ensure that all input features have available measurements before use in generating an uncalibrated sepsis risk score.

[0130] ii. Pre-trained machine learning models

[0128] In certain embodiments, the pre-trained machine learning model is a supervised machine learning model constructed using a training dataset that includes input features and training labels along with hyperparameters. In detailed embodiments, this machine learning model is stored as an R data structure file. In detailed embodiments, the objective of the machine learning model is to learn the relationship between input features and the risk of having or developing sepsis within 24 hours of the Sepsis ImmunoScore request time. This relationship can, in at least some cases, be applied to unobserved patients in the future who have the corresponding input features.

[0131] iii. Calibration Model

[0129] In detailed embodiments, the calibration model is a logistic model that generates a Sepsis ImmunoScore sepsis risk score using an uncalibrated sepsis risk score as input. This calibration model can be saved as an R data structure file. In one embodiment, the objective of the calibration model is to establish a relationship between an uncalibrated sepsis risk score and a classification of having or developing sepsis within 24 hours of a Sepsis ImmunoScore request. This relationship can, in certain embodiments, be applied in the future to convert an uncalibrated sepsis risk score into a sepsis risk score.

[0132] iv. Training data object

[0130] In detailed embodiments, the training data object is a data frame containing the input features of the training dataset. The training data object may be stored as an R data structure file. In some embodiments, the purpose of the training data object is to provide data for learning the relationship between the SHAP values ​​of the input features and their corresponding uncalibrated sepsis risk score values. In some embodiments, this relationship may be applied to a Sepsis ImmunoScore request in the future to generate the SHAP values ​​of the corresponding sepsis risk scores.

[0133] v. Threshold List

[0131] In some embodiments, the threshold list is a vector containing three thresholds between 0 and 1 related to the boundaries of the Sepsis ImmunoScore risk category. The threshold list may be stored in a JSON file. In specific embodiments, the three values ​​are ranked in ascending order and referred to as the “lowest,” “optimal,” and “95th percentile” thresholds, respectively.

[0134] III.E. Development Datasets

[0132] An algorithm was developed using retrospective data from anonymized patient data and samples derived from the multicenter NOSIS dataset and Biobank.

[0135]

[0133] In a detailed embodiment, the NOSIS dataset and Biobank are integrated datasets consisting of prospectively collected clinical data (electronic medical record data), time-series biological samples, and biomarker measurements obtained from these samples. This dataset is constructed through prospective clinical trials sponsored by the applicant in a community of clinical facilities.

[0136]

[0134] III.F. Algorithmic Artifact Design

[0137]

[0135] The following sections define the methods used to design each of the Sepsis ImmunoScore algorithm functions and artifacts.

[0138] i. Completion templates

[0136] In some embodiments, the completion template is generated by using bag imputation on the Sepsis ImmunoScore algorithm training dataset as identified in other parts of this specification. Bag imputation is, in specific embodiments, a statistical method for constructing a random forest model for each input feature of the Sepsis ImmunoScore algorithm. Each random forest model may generate completion values ​​using the remaining observed input features. In some embodiments, the completion template may be performed on unseen data and determined by calculating the R² of the completion value versus the known value for each input feature.

[0139]

[0137] The bag imputation was implemented using the imputeMissings R package.

[0140]

[0138] In certain embodiments, the data used to develop the complementary template is the same as that used for the machine learning model.

[0141] ii. Pre-trained machine learning models

[0139] A pre-trained machine learning model can be developed by training a supervised machine learning model using a training dataset that includes input features and training labels along with hyperparameters relevant to the machine learning model. In certain embodiments, a pre-trained machine learning model is determined by looking at specific performance criteria related to diagnostic and prognostic performance.

[0142] iii. Machine Learning Models

[0140] In a specific embodiment, the machine learning model used is a random forest model. A random forest model is a commonly used ensemble model that includes many simple tree models to generate predictions (Breiman, Leo. "Random forests." Machine learning 45, 2001, pp. 5-32). The random forest performs bagging, which is a method of resampling the dataset. Individual simple models can be trained on this sampled dataset. In a specific embodiment, this resampling and subsequent training is performed n times to generate an ensemble or forest of simple models. The random forest used in the development of the Sepsis ImmunoScore algorithm, in a specific embodiment, generated a probabilistic forest using a regression tree model of 1000 as the base model (Malley, James D., Jochen Kruppa, Abhijit Dasgupta, Karen G. Malley, and Andreas Ziegler. “Probability machines.” Methods of information in medicine 51, no. 01, 2012, pp. 74-81). In a specific embodiment, the random forest model was trained by maximizing the area under the receiver operating characteristic curve (AUROC) between predictions and training labels.

[0143]

[0141] The advantage of the random forest model is the use of out-of-bag (OOB) errors. Because each decision tree is generated by bagging, there is a probability that a certain observation will not be included in the training of the decision tree. This set of excluded observations can be defined as OOB in certain embodiments. In specific embodiments, by generating predictions for OOB, it is possible to identify errors for all observations instead of relying on traditional training / test splits.

[0144]

[0142] The random forest was implemented using the ranger R package.

[0145] iv. Training dataset

[0143] The training dataset is a subset of data from the NOSIS dataset and Biobank, and was extracted according to the following criteria.

[0146]

[0144] This was a multicenter prospective observational study of patients suspected of having sepsis. Patients were identified by querying the hospital's electronic medical record system against study enrollment criteria. Patients who had a relevant request for a test indicating suspected bacteremia (blood culture) were eligible for enrollment. A list of eligible patients was compiled, and researchers or hospital staff evaluated the presence of residual samples from lithium heparin tubes with gel separators (pale green caps) in the clinical laboratory. Patients with available samples were enrolled as part of the NOSIS dataset.

[0147]

[0145] Adult patients (18 years or older) who presented to or were hospitalized in the emergency department with suspected sepsis were included. Suspicion of sepsis was inferred from requests for blood cultures. Patient data, including demographic data, comorbidities, vital signs, laboratory results, electronic record interventions, and ICD-10 codes, were extracted from the hospital's electronic health record (EHR). In addition, residual plasma samples remaining from routine clinical blood collections were collected, each containing four 175 μL aliquots, frozen at -80°C, and sent to the central laboratory. Each sample contained at least one prospective -80°C frozen plasma aliquot from the time the sample became available until three days prior to registration and during hospitalization.

[0148]

[0146] For each patient, one discarded plasma sample was assayed for approximately 40 protein biomarkers, including CRP and PCT, using patient subsamples from the NOSIS biobank. The work was performed sequentially for each facility based on the patient registration date, and the inventors selected the sample closest to baseline, defining this as the blood culture request time. Patients without biobanked samples collected within 3 hours of the initial blood culture request were excluded. Protein concentrations were measured using the multiplex Luminex MagPix assay developed by R&D Systems. Laboratory quality control procedures strictly adhered to the FDA's industry guidance on bioanalytical method validation to ensure data quality. 2,367 patients recruited from the three hospital facilities in this study were identified as triggerable by the logic trigger and served as data to train the algorithm.

[0149] v. Input Feature Selection

[0147] The selection measures used for the input features correspond to the logic specified in U.S. Provisional Patent Application No. 63 / 459,521, filed on April 14, 2023 (which is incorporated herein by reference in its entirety). Measurement completion was performed using the same criteria specified elsewhere in this Spec. During development, the request for Sepsis ImmunoScore was replaced with a request for an initial blood culture during the encounter.

[0150] vi.training labels

[0148] The labels used to train the Sepsis ImmunoScore machine learning model were derived Sepsis-3 labels within 24 hours of the initial blood culture request. In specific embodiments, the derived Sepsis-3 labels were calculated based on the presence of organ dysfunction and infection. The timing of the organ dysfunction event corresponds to the Sepsis-3 time, given that the patient met the infection criteria. Further details of the organ dysfunction component and infection criteria are listed below.

[0151] vii. Organ dysfunction

[0149] Organ dysfunction was calculated using available drug therapies, clinical laboratory results, vital signs, and patient assessments available in the NOSIS database during the encounter. Organ dysfunction is calculated using the following formula: SOFA-SOFA baseline ≥ 2

[0152]

[0150] The SOFA score is calculated as the sum of the CNS, cardiovascular, respiratory, renal, coagulation, and hepatic SOFA components. The method for calculating the SOFA score is specified in Table 6 below. Baseline SOFA was calculated as the SOFA component before the encounter, if available. If pre-encounter data was not available, the baseline was assumed to be 0.

[0153] [Table 6-1]

[0154] [Table 6-2]

[0155] viii. Infection

[0151] If any of the following criteria are met, the Sepsis ImmunoScore infection criterion is identified. 1. Within a two-day window before and after the initial blood culture request, there are four or more quality of day administrations (QADs) that meet the criteria. 2. Positive culture after microbiological testing or excluding known contamination.

[0156]

[0152] The QAD rules follow the instructions in the CDC's Hospital Toolkit Infection Rules for Adult Sepsis (https: / / www.cdc.gov / sepsis / pdfs / Sepsis-Surveillance-Toolkit-Mar-2018_508.pdf). Microbial positivity is an addition to the CDC's Hospital Toolkit for Adult Sepsis. See Table 7 for a list of known contaminations in blood cultures.

[0157] [Table 7]

[0158]

[0153] Known contaminations in urine cultures are defined as "Lactobacillus, Corynebacterium species, Gardnerella species, α-hemolytic streptococci, and aerobic bacteria" (Franz, Martina, and Walter H. Hoerl. "Common Errors in Diagnosis and Management of Urinary Tract Infection. I: Pathophysiology and Diagnostic Techniques." Nephrology Dialysis Transplantation, vol. 14, no. 11, 1999, pp. 2746-2753., https: / / doi.org / 10.1093 / ndt / 14.11.2746).

[0159] ix. Hyperparameters

[0154] In detailed embodiments, the hyperparameters related to the random forest model are mtry, minimum node size, and split rule. Various combinations of mtry, minimum node size, and split rule were considered. In a particular embodiment, the combination that repeatedly provided the best 5x cross-validated AUROC results three times was selected as the optimal hyperparameter. The R package "caret" was used to facilitate the selection of the optimal hyperparameter.

[0160] x.Performance judgment

[0155] In certain embodiments, the pre-trained machine learning model is evaluated by assessing its diagnostic performance and prognosis prediction performance. Both performances are evaluated using the sepsis risk score calculation and risk threshold identified during the development of the Sepsis ImmunoScore algorithm.

[0161] xi. Creating risk thresholds

[0156] During the development of the Sepsis ImmunoScore algorithm, receiver operating characteristic (ROC) curves were generated using uncalibrated predictions for training labels. Using these ROC curves, three thresholds were identified in a particular embodiment based on the following criteria.

[0162]

[0157] 1. Lowest threshold - Identify the highest sensitivity threshold while maintaining a false positive rate of 50%.

[0163]

[0158] 2. Optimal threshold - The threshold is identified as the point at the "upper left corner" of the ROC curve, i.e., the point on the ROC curve where the Euclidean distance to (0,1) is minimized.

[0164]

[0159] 3. 95th percentile threshold - The threshold corresponding to the 95th percentile of the sepsis risk score within the training dataset.

[0165] xii. Diagnostic performance

[0160] The global diagnostic performance of the machine learning model was measured by identifying the area under the global OOB precision-recall curve (AURPC) and the area under the OOB precision-recall curve (AURPC). Furthermore, diagnostic metrics (e.g., sensitivity, specificity, positive predictive value, negative predictive value, etc.) were evaluated at various thresholds.

[0166] xiii. Prognostic performance

[0161] Prognostic performance was determined by identifying the occurrence of adverse events as specified in other parts of this specification for each of the four risk categories.

[0167] xiv. Calibration Model

[0162] Classifiers generated by random forest models, despite achieving both accuracy and high AUC, tend to produce lower-quality class probabilities when measured by the squared error of the predicted probability (Bryar score). One approach to address this problem is through a calibration model, which involves learning a function that maps the original probability estimate or score to a more accurate probability estimate.

[0168]

[0163] In a detailed embodiment, a calibration model was generated by performing Pratt calibration. Pratt calibration can be produced by training a logistic regression model on the Sepsis ImmunoScore before calibration to predict the Sepsis-3 training label. The calibration model can be evaluated by identifying the proportion of patients with sepsis given the Sepsis ImmunoScore.

[0169] IV. Exemplary Methodology IV.A. Exemplary Methodology

[0164] Figure 2 illustrates an exemplary method 200 for generating an index of sepsis risk according to various embodiments. Step 210 of method 200 includes the step of receiving at least one input feature corresponding to a patient. Step 220 further includes the step of analyzing at least one input feature using a machine learning model to generate a predictive score indicating the probability that a patient will have or develop sepsis within a given period of time based on at least one input feature. In various embodiments, optionally (step 230), at least one input feature is related to sepsis.

[0170]

[0165] In various embodiments, at least one input feature can be divided into two groups. These two groups may be intensity values, where the intensity value of the first group has a lower relative importance than the intensity value of the second group.

[0171]

[0166] In various embodiments, at least one input feature may include procalcitonin and be divided into two groups. These two groups may be procalcitonin concentration values, where the procalcitonin concentration values ​​of the first group have a lower relative importance than the procalcitonin concentration values ​​of the second group.

[0172]

[0167] In various embodiments, at least one input feature can be divided into three groups. These three groups may be intensity values, where the intensity value of the first group has a lower relative importance than the intensity value of the second group, and the intensity value of the third group has a higher relative importance than both the intensity values ​​of the first and second groups.

[0173]

[0168] In various embodiments, at least one input feature may include procalcitonin and be divided into three groups. These three groups may be procalcitonin concentration values, where the procalcitonin concentration values ​​of the first group have a lower relative importance than the procalcitonin concentration values ​​of the second group, and the procalcitonin concentration values ​​of the third group have a higher relative importance than both the procalcitonin concentration values ​​of the first and second groups.

[0174]

[0169] Figure 9A illustrates an unrestricted example of 2-group or 2-bin features applied to Model 1A (discussed in detail below). Figure 9B illustrates an unrestricted example of 3-group or 3-bin features applied to Model 1A.

[0175]

[0170] This exemplary method can be further improved in various embodiments, for example, as provided in claims 2 to 171. The further features included therein can be implemented in any combination as needed. [Examples]

[0176] V. Examples

[0171] The following examples are included to demonstrate preferred embodiments of the present invention. Those skilled in the art will understand that the techniques disclosed in these following examples correspond to techniques discovered by the inventors to work well in the implementation of the present invention and may therefore be considered to constitute preferred embodiments for implementation. However, those skilled in the art will understand that many modifications can be made to the specific embodiments disclosed in light of this disclosure and still obtain similar or similar results without departing from the spirit and scope of the present invention.

[0177] Example 1 Examples of algorithmic results and feature importance rankings for exemplary models.

[0172] This embodiment provides examples of algorithm results and feature importance rankings for multiple versions of the sepsis algorithm. These versions differ in the set of input features used to predict sepsis. In a detailed embodiment, six different models utilize one or more various demographic measurements, one or more patient assessments, one or more vital signs, one or more blood test values, one or more chemical test values ​​and / or one or more sepsis-related biomarker concentrations.

[0178] VA Model 1A

[0173] Exemplary input features for Model 1A are provided below in no particular order.

[0179]

[0174] Table 8: Example of input features for Model 1A [Table 8-1]

[0180]

[0175] In one embodiment of Model 1A, examples of SHAP values ​​for three different patients at risk of sepsis are provided in Figure 3A (no sepsis within 24 hours), Figure 3B (sepsis within 24 hours), and Figure 3C (no sepsis within 24 hours). The ranking of the importance of the input features differs among the three patients.

[0181]

[0176] Figure 3A illustrates a snapshot of the SHAP values ​​for patient 1 for notable input features. For this patient, most feature values, particularly his young age and normal BUN, platelet, and creatinine levels, have negative SHAP values, i.e., they reduce his Sepsis ImmunoScore and indicate no signs of dysfunction in these organs. However, his elevated bilirubin is a sign of abnormal liver function, and his elevated respiratory rate and, to a lesser extent, his body temperature exhibit substantially positive SHAP values.

[0182]

[0177] Figure 3B illustrates a snapshot of patient 2's SHAP values ​​for notable input features. This elderly patient had positive SHAP values ​​for most features, with the highest contributions coming from platelet count, low pulse oximetry and BUN levels, which is an indication of organ dysfunction consistent with Sepsis-3's definition of organ dysfunction (an increase of at least 2 points from baseline in the SOFA score).

[0183]

[0178] Figure 3C illustrates a snapshot of patient 2's SHAP values ​​for notable input features. This patient has negative SHAP values ​​for almost all features, with the exception of age and a few miscellaneous laboratory values ​​that are barely positive. Both organ function and infection-related features are represented as SHAP values ​​in the most important tier, all in the negative direction. This patient is unlikely to develop sepsis within the next 24 hours.

[0184]

[0179] The following are exemplary results regarding the diagnostic performance and prognostic prediction performance of Model 1A.

[0185] [Table 8-2]

[0186] VB Model 1B

[0180] Exemplary input features for Model 1B are provided below in no particular order. In detailed embodiments, the input features for Model 1B include IL-6, which may be missing from the input features for Model 1A.

[0187]

[0181] Table 9: Example of input features for Model 1B

[0188]

[0182] [Table 9-1]

[0189]

[0183] The following are exemplary results regarding the diagnostic performance and prognostic prediction performance of Model 1B.

[0190] [Table 9-2]

[0191] [Table 9-3]

[0192] VC Model 2

[0184] The following are exemplary input features for Model 2, in no particular order.

[0193] [Table 10-1]

[0194]

[0185] In one embodiment of Model 2, examples of SHAP values ​​for three different patients at risk of sepsis are provided in Figure 4A (sepsis within 24 hours), Figure 4B (sepsis within 24 hours), and Figure 4C (no sepsis within 24 hours). The importance ranking of the input features differs depending on the patient.

[0195]

[0186] Figure 4A illustrates a snapshot of patient 1's SHAP values ​​for notable input features. For this patient, the SHAP values ​​for most features are positive, primarily due to his bilirubin and platelet counts, both of which are signs of organ dysfunction. The feature values ​​for several infection-related markers suggest that his organ dysfunction may be due to infection, as defined in Sepsis-3.

[0196]

[0187] Figure 4B illustrates a snapshot of patient 2's SHAP values ​​for notable input features. This patient has negative SHAP values ​​that correspond almost uniformly to extremely low Sepsis ImmunoScores. Of these, many of the most important layers (i.e., the largest of the top layers) are protein biomarkers, including IL-8, procalcitonin, IL-6, and pentraxin-3.

[0197]

[0188] Figure 4C illustrates a snapshot of patient 3's SHAP values ​​for notable input features. This patient has both large positive and negative SHAP values, including elevated respiratory rate and certain sepsis-related biomarkers such as IL-6, IL-8, and CRP. However, its normal blood pressure and platelet count are the strongest overall contributors.

[0198]

[0189] The following are exemplary results regarding the diagnostic and prognostic performance of Model 2.

[0199] [Table 10-2]

[0200] [Table 10-3]

[0201] VD Model 3

[0190] The following are exemplary input features for Model 3, in no particular order.

[0202]

Table 11-1

[0203]

[0191] In one embodiment of Model 3, examples of SHAP values for three different patients at risk of sepsis are provided in FIGS. 5A (no sepsis within 24 hours), FIG. 5B (no sepsis within 24 hours), and FIG. 5C (sepsis within 24 hours). The order of importance of the input features varies among the three patients.

[0204]

[0192] FIG. 5A illustrates a snapshot of the SHAP values for Patient 1 for the input features of interest. This patient has both high positive and high negative SHAP values. Abnormal blood pressure readings and increases in creatinine and several protein biomarkers (e.g., IL-8, TREM-1) all increase his Sepsis ImmunoScore. However, pentraxin-3, IL-6, CRP, and other biomarkers including his respiratory rate and normal blood oxygen decrease his risk score.

[0205]

[0193] FIG. 5B illustrates a snapshot of the SHAP values for Patient 2 for the input features of interest. Elevated IL-8, IP-10, and a substantially negative contribution due to advanced age provide a positive contribution to Patient's Sepsis ImmunoScore. FIG. 5C illustrates a snapshot of the SHAP values for Patient 3 for the input features of interest. Pentraxin-3 has the highest SHAP value for this patient, along with abnormal values for other infection-related markers and measures of organ dysfunction (platelets, pulse oximetry).

[0206]

[0194] Exemplary results of the diagnostic performance and prognostic prediction performance of Model 3 are as follows.

[0207]

Table 11-2

[0208] [Table 11-3]

[0209]

[0195] VE Model 4

[0210]

[0196] The following are exemplary input features for Model 4, in no particular order.

[0211] [Table 12-1] [Table 12-2]

[0212]

[0197] In one embodiment of Model 4, examples of SHAP values ​​for three different patients at risk of sepsis are provided in Figure 6A (no sepsis within 24 hours), Figure 6B (sepsis within 24 hours), and Figure 6C (no sepsis within 24 hours). Each patient has a different ranking of the input features.

[0213]

[0198] Figure 6A illustrates a snapshot of patient 1's SHAP values ​​for a notable input feature. Due to abnormal blood pressure, a sign of cardiovascular dysfunction, this patient's blood pressure reading is its most important SHAP value.

[0214]

[0199] Figure 6B illustrates a snapshot of patient 2's SHAP values ​​for notable input features. Low platelet count is the most important feature for this patient, which is appropriate for classifying the patient as organ dysfunction due to coagulation disorders.

[0215]

[0200] Figure 6C illustrates a snapshot of the SHAP values for Patient 3 for the input features of interest. The increases in creatinine, IL-8, and granzyme-B have positive SHAP values, while normal platelet, blood pressure, and other feature values all contribute negatively to the sepsis risk for this patient.

[0216]

[0201] Exemplary results of the diagnostic performance and prognostic prediction performance of Model 4 are as follows.

[0217]

Table 12-3

[0218]

Table 12-4

[0219] V.F. Model 5

[0202] Exemplary input features for Model 5 are provided below in random order.

[0220]

[0203] Table 13: Examples of input features of Model 5

Table 13-1

[0221]

[0204] In one embodiment of Model 5, examples of SHAP values for three different patients at risk of sepsis are provided in FIGS. 7A (sepsis within 24 hours), FIG. 7B (no sepsis within 24 hours), and FIG. 7C (no sepsis within 24 hours). The input features have different rankings of importance among the three patients.

[0222]

[0205] Figure 7A illustrates a snapshot of the SHAP values for Patient 1 for the input features of interest. The abnormal blood pressure indicating cardiovascular organ dysfunction was the most powerful feature as a contribution to the sepsis risk score for this patient.

[0223]

[0206] Figure 7B illustrates a snapshot of patient 2's SHAP values ​​for the input features of interest. This patient has negative SHAP values ​​for most features, but the most important individual contribution was its IL-8 value.

[0224]

[0207] Figure 7C illustrates a snapshot of patient 3's SHAP values ​​for notable input features. Abnormal pulse oximetry and IL-8 had high-importance positive SHAP values, while most other laboratory and biomarker values ​​had a negative contribution to its sepsis risk score.

[0225]

[0208] The following are exemplary results regarding the diagnostic and prognostic performance of Model 5.

[0226] [Table 13-2]

[0227] [Table 13-3]

[0228] VI. Embodiments of Input Features

[0209] Detailed embodiments of this disclosure provide risk stratification related to long-term and short-term prognostic indicators of clinical outcomes based on the analysis of one or more input features in any one or more of Tables 8 to 13.

[0229]

[0210] In various embodiments, an indicator of an individual's risk of sepsis is provided by one or more input features in one or more of the tables 8 to 13. In a detailed embodiment, an indicator of an individual's risk of sepsis is provided by input features 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 of Model 1A in Table 8. In a detailed embodiment, an indicator of an individual's risk of sepsis is provided by input features 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, or 23 of Model 1B in Table 9. In a detailed embodiment, an index of an individual's sepsis risk is provided by 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, or 27 input features of Model 2 in Table 10. In a detailed embodiment, an index of an individual's sepsis risk is provided by 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, or 29 input features of Model 3 in Table 11. In a detailed embodiment, an index of an individual's risk of sepsis is provided by 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, or 58 input features of Model 4 in Table 12. In a detailed embodiment, an indicator of an individual's risk of sepsis is provided by input features 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 of Model 5 in Table 13.

[0230]

[0211] In a specific embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 input features of Model 1A are used as indicators of the individual's risk of sepsis. In a specific embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, or 23 input features of Model 1B are used as indicators of the individual's risk of sepsis. In a specific embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, or 27 input features of Model 2 are used as indicators of an individual's risk of sepsis. In a specific embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, or 29 input features of Model 3 are used as indicators of an individual's risk of sepsis. In a specific embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, or 58 input features of Model 4 are used as indicators of an individual's risk of sepsis. In a specific embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 input features of Model 5 are used as indicators of an individual's risk of sepsis.

[0231]

[0212] In a specific embodiment, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 or fewer input features of Model 1A are used as indicators of the individual's risk of sepsis. In a specific embodiment, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, or 23 or fewer input features of Model 1B are used as indicators of the individual's risk of sepsis. In a specific embodiment, up to 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, or 27 input features of Model 2 are used as indicators of an individual's risk of sepsis. In a specific embodiment, up to 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, or 29 input features of Model 3 are used as indicators of an individual's risk of sepsis. In a specific embodiment, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, or 58 or fewer input features of Model 4 are used as indicators of the individual's risk of sepsis. In a specific embodiment, up to 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 input features of Model 5 are used as indicators of an individual's risk of sepsis.

[0232]

[0213] In various embodiments, the majority of the input features of Model 1A, 1B, 2, 3, 4, or 5 are used as an indicator of the individual's risk of sepsis. In some embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, or 95% of the input features of Model 1A, 1B, 2, 3, 4, or 5 are used as an indicator of the individual's risk of sepsis.

[0233]

[0214] In a detailed embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 input features of Model 1A are used to determine the risk of an individual developing sepsis or to determine that an individual has sepsis. In particular, the ranking of importance of the input features may be patient-specific.

[0234]

[0215] In a detailed embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, or 23 input features of Model 1B are used to determine the risk of an individual developing sepsis or to determine that an individual has sepsis. In particular, the ranking of importance of the input features may be patient-specific.

[0235]

[0216] In a detailed embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, or 27 input features of Model 2 are used to determine the risk of an individual developing sepsis or to determine that an individual has sepsis. In particular, the ranking of importance of the input features may be patient-specific.

[0236]

[0217] In a detailed embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, or 29 input features of Model 3 are used to determine the risk of an individual developing sepsis or to determine that an individual has sepsis. In particular, the ranking of the importance of the input features may be patient-specific.

[0237]

[0218] In a detailed embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, or 58 input features of Model 4 are used to determine the risk of an individual developing sepsis or to determine that an individual has sepsis. In particular, the ranking of the importance of the input features may be patient-specific.

[0238]

[0219] In a detailed embodiment, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 input features of Model 5 are used to determine the risk of an individual developing sepsis or to determine that an individual has sepsis. In particular, the ranking of importance of the input features may be patient-specific.

[0239]

[0220] In any model included herein, patient age may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any model included herein, patient albumin levels may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any model included herein, patient angiopoietin-1 levels may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any model included herein, patient angiopoietin-2 levels may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any model included herein, patient bilirubin levels may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's biological sex, such as being female, may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's blood urea nitrogen (BUN) level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's chloride level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's CO2 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's C-reactive protein level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's creatinine level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis.In any of the models contained herein, the patient's diastolic blood pressure level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's E-selectin level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's FLT-3 ligand level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's fractalkine level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's GCSF level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's GMCSF level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's granzyme B level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's glucose level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's heart rate may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's hematocrit level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's hemoglobin level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis.In any of the models contained herein, the patient's IFN-α level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IFN-γ level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-10 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-15 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-1-β level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-1-RA level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-2 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-4 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-6 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IL-7 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's IL-8 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis.In any of the models contained herein, the patient's IL-10 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's IP-10 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's lactate level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's leptin level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's lymphocyte level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's MCP1 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's MIP1-α level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's MIP1-β level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's MIP3-α level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's neutrophil level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's NGAL level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis.In any of the models contained herein, a patient's platelet level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's programmed death-ligand 1 (PD-L1) level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's pentraxin-3 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's potassium level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's platelet level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's procalcitonin level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's pulse oximetry may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's respiratory rate may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's sodium level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's systolic blood pressure may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's body temperature may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis.In any of the models contained herein, a patient's TGF-α level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's thrombomodulin level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, a patient's tissue factor level is the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's TNF-α level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's TRAIL level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the triggering receptor 1 (TREM-1) level expressed on the patient's myeloid cells may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's VCAM-1 level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's VEG-F level may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis. In any of the models contained herein, the patient's white blood cell count may or may not be the most important input feature for determining the risk of developing sepsis or the presence of sepsis.

[0240]

[0221] In various embodiments, the method of the present disclosure generates an index of sepsis risk, such as a predictive score indicating the probability that a patient has or develops sepsis within a given period, based on at least one input feature listed for any one or more models in Tables 8 to 13. In some embodiments, a particular input feature has a higher importance in the predictive score than one or more other input features.

[0241]

[0222] In a specific embodiment of Model 1A, age has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, albumin level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, bilirubin level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, the BUN level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, the chloride level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, CO2 has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, C-reactive protein level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, creatinine level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features.In a specific embodiment of Model 1A, diastolic blood pressure has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, heart rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, lactate level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, lymphocyte levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, neutrophil levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, platelet levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, potassium levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, procalcitonin levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, pulse oximetry has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features.In a specific embodiment of Model 1A, respiratory rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, sodium level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, systolic blood pressure has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, body temperature has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 1A, white blood cell count has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features.

[0242]

[0223] In a specific embodiment of Model 1B, age has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, albumin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, the bilirubin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, the BUN level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, chloride levels have a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, CO2 has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, C-reactive protein levels have a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, creatinine level has a higher importance in the prediction score than other input features 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22.In a specific embodiment of Model 1B, diastolic blood pressure has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, heart rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, IL-6 level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, lactate levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, lymphocyte levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, neutrophil levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, platelet levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, potassium levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features.In a specific embodiment of Model 1B, procalcitonin levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, pulse oximetry has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, respiratory rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, sodium level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, systolic blood pressure has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, body temperature has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features. In a specific embodiment of Model 1B, white blood cell count has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, or 22 other input features.

[0243]

[0224] In a specific embodiment of Model 2, age has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, albumin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the angiopoietin-2 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the bilirubin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the BUN level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the chloride level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the CO2 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features.In a specific embodiment of Model 2, C-reactive protein level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, creatinine level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, diastolic blood pressure has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, heart rate has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the IL-6 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the IL-8 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the IP-10 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the lactate level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features.In a specific embodiment of Model 2, lymphocyte levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, neutrophil levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the pentraxin-3 level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the platelet level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, potassium levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, procalcitonin levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, pulse oximetry has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features.In a specific embodiment of Model 2, respiratory rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, sodium level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, systolic blood pressure has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, body temperature has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features. In a specific embodiment of Model 2, the white blood cell count has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or 26 other input features.

[0244]

[0225] In a specific embodiment of Model 3, age has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, albumin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the angiopoietin-2 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the bilirubin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the BUN level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the chloride level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, CO2 has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features.In a specific embodiment of Model 3, C-reactive protein level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, creatinine level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, diastolic blood pressure has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, heart rate has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the IL-6 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the IL-8 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the IP-10 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features.In a specific embodiment of Model 3, lactate levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, lymphocyte levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, neutrophil levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, PD-L1 levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, pentraxin-3 level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, platelet level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, potassium level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features.In a specific embodiment of Model 3, procalcitonin levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, pulse oximetry has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, respiratory rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, sodium level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, systolic blood pressure has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, body temperature has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features. In a specific embodiment of Model 3, the TREM-1 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features.In a specific embodiment of Model 3, the white blood cell count has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, or 28 other input features.

[0245]

[0226] In a specific embodiment of Model 4, age has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, albumin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the angiopoietin-1 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the angiopoietin-2 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the bilirubin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the BUN level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the chloride level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the CO2 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the C-reactive protein level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, creatinine level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, diastolic blood pressure has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the E-selectin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the FLT3-ligand level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the fractal kine level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the GCSF level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the GMCSF level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the Granzyme-B level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, heart rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IFN-α level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IFN-γ level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the IL-10 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IL-15 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IL1-β level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IL1-RA level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, or 1. It has a higher importance than 4, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the O:-2 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IL-4 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IL-6 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the IL-7 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IL-8 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the IP-10 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the lactate level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, leptin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, lymphocyte levels have a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the MCP1 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the MIP1-α level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the MIP1-β level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the MIP3-α level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, neutrophil levels have a higher importance in the predictive score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the NGAL level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the PD-L1 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the pentraxin-3 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, platelet level has a higher importance in the predictive score than other input features of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57. In a specific embodiment of Model 4, potassium levels have a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, procalcitonin levels have a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, pulse oximetry has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the respiratory rate has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the sodium levels are predicted to be 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, and 45. , has a higher importance than 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56 or 57 other input features. In a specific embodiment relating to Model 4, systolic blood pressure has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56 or 57 other input features. In a specific embodiment of Model 4, body temperature has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the TGF-α level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the thrombomodulin level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the organizational factor level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the TNF-α level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the TRAIL level has a higher importance in the prediction score than other input features of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57. In a specific embodiment of Model 4, the TREM-1 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.In a specific embodiment of Model 4, the VCAM-1 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the VEGF level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features. In a specific embodiment of Model 4, the white blood cell count has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, or 57 other input features.

[0246]

[0227] In a specific embodiment of Model 5, age has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, angiopoietin-2 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, BUN level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, calcium level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, chloride level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, CO2 has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, creatinine level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, diastolic blood pressure has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, biological sex has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features.In a specific embodiment of Model 5, glucose value has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, heart rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, hematocrit has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, hemoglobin has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, the IL-8 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, the IP-10 level has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, pentraxin-3 level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, potassium level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, pulse oximetry has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features.In a specific embodiment of Model 5, respiratory rate has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, sodium level has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, systolic blood pressure has a higher importance in the predicted score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features. In a specific embodiment of Model 5, body temperature has a higher importance in the prediction score than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or 21 other input features.

[0247] VII. Embodiments of Treatment Methods

[0228] In some embodiments, individuals determined to have sepsis or be at risk of sepsis (including being at risk within or beyond 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 341, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48 hours) may be provided with one or more therapeutic doses. In specific embodiments, this therapy may include, for example, one or more antibiotics that can be administered orally (by pills, liquids or capsules), by injection, intravenous infusion and / or topically (by cream or ointment) and / or intravenously. Treatment may be on an order of 1–24, 1–12, 1–6, 6–24, 6–18, 6–12, 12–24, 12–18 or 18–24 hours, 1–7, 1–6, 1–5, 1–4, 1–3, 1–2, 2–7, 2–6, 2–5, 2–4, 2–3, 3–7, 3–6, 3–5, 3–4, 4–7, 4–6, 4–5, 5–7, 5–6 or 6–7 days, or 1–4, 1–3, 1–2, 2–4, 2–3 or 3–4 weeks or longer, depending on the type of infection and how it responds to antibiotics. In some embodiments, the initial treatment may be ineffective or less effective than desired, requiring a second treatment, such as with a different type of antibiotic.

[0248]

[0229] In detailed embodiments, when an individual requires antibiotics, one or more of the following may be administered: (1) ceftriaxone (e.g., 1 to 2 g IV or IM every 12 to 24 hours for adults, and 50 to 75 mg / kg / day for infants, children and adolescents) (1) IV or IM in divided doses every 12 to 24 hours, and 50 mg / kg / dose IV or IM every 24 hours for neonates and premature infants), (2) Piperacillin-tazobactam (e.g., 4.5 g (4 g piperacillin and 0.5 g tazobactam) IV every 6 hours for adults, 80 mg / kg / dose piperacillin component (90 mg / kg / dose piperacillin and tazobactam for infants, children and adolescents) IV every 6 hours, and 80 or 100 mg / kg / dose piperacillin component (90 mg / kg / dose piperacillin and tazobactam) IV every 6 hours for neonates), (3) Cefepime (e.g., 2 g for adults) (4) IV every 8 hours, 50 mg / kg / dose (maximum: 2 g / dose) for infants, children and adolescents, IV every 8 hours and 30 or 50 mg / kg / dose every 12 hours for neonates), (4) ceftazidime (e.g., 2 g for adults and adolescents) (5) Vancomycin (for example, a loading dose of 20 to 35 mg / kg / dose (maximum: 3,000 mg / dose) IV every 8 hours for adults, followed by 15 to 20 mg / kg / dose IV every 8 to 12 hours for children and adolescents, 60 to 70 mg / kg / day IV every 6 to 8 hours for infants and adolescents, 60 to 80 mg / kg / day IV every 6 hours for infants), and / or (6) Ciprofloxacin (for example, 600 mg IV every 12 hours for adults).

[0249]

[0230] Embodiments of the present disclosure include methods for treating sepsis in individuals who have, are suspected of having, or are at risk of having sepsis within a predetermined period (e.g., within 1 to 48 hours), including individuals whose input features are generated using a machine learning model to produce a predictive score indicating that the individual has or will develop sepsis within a predetermined period (e.g., within 1 to 48 hours).

[0250]

[0231] Embodiments of the present disclosure include a method for classifying a patient as having sepsis or at risk of developing sepsis within a given period of time by generating a diagnostic output based on a predictive score from any method incorporated herein, further comprising administering to the individual a therapeutically effective dose of treatment for sepsis. In specific embodiments, the treatment includes at least one of an antibiotic or an intravenous fluid based on the generated diagnostic output.

[0251]

[0232] Embodiments of the present disclosure include a method for treating sepsis in an individual having a score based on a plurality of input features indicating the probability that a patient has or will develop sepsis, based on any method incorporated herein, wherein the treatment comprises administering to the individual an effective amount of one or more sepsis therapies.

[0252]

[0233] Embodiments of the present disclosure include a method comprising measuring a plurality of input features in Tables 8, 9, 10, 11, 12 and / or 13 from an individual and / or a sample from an individual, wherein optionally, the measurements of at least a first input feature and at least a second input feature are sequential, and the measurement of the second input feature is based on the result of the measurement of the first input feature.

[0253]

[0234] Embodiments of the present disclosure include a method comprising measuring one, two, three, four, five, six, seven, eight, nine, ten, one, two

[0254]

[0235] Embodiments of the present disclosure include a method for treating sepsis in an individual, wherein the individual has one or more specific values ​​for any input feature in one or more of Tables 8, 9, 10, 11, 12, or 13, and the input feature indicates sepsis.

[0255]

[0236] Embodiments of the present disclosure include a method for treating an individual having one or more specific values ​​for one or more input features in any one or more of Tables 8, 9, 10, 11, 12, or 13, wherein the treatment comprises administering one or more antibiotics in a therapeutically effective amount.

[0256] VIII. Computer-Implemented Systems

[0237] Figure 8 is a block diagram of a computer system according to various embodiments. The computer system 800 can be one implementation of various methods, such as the method 200 described in Figure 2 above.

[0257]

[0238] In one or more examples, the computer system 800 may include a bus 802 or other communication mechanism for communicating information, and a processor 804 coupled to the bus 802 for processing information. In various embodiments, the computer system 800 may also include memory, which may be random access memory (RAM) 806 or other dynamic storage device, coupled to the bus 802 for determining instructions to be executed by the processor 804. The memory may also be used to store temporary variables or other intermediate information during the execution of instructions to be executed by the processor 804. In various embodiments, the computer system 800 may further include read-only memory (ROM) 808 or other static storage device, coupled to the bus 802 for storing static information and instructions for the processor 804. A storage device 810 for storing information and instructions, such as a magnetic disk or optical disk, may be provided and coupled to the bus 802.

[0258]

[0239] In various embodiments, the computer system 800 may be coupled via bus 802 to a display 812 for displaying information to the computer user, such as a cathode ray tube (CRT), liquid crystal display (LCD), or light-emitting diode (LED). An input device 814, including alphanumeric keys and other keys, may be coupled to bus 802 for communicating information and instruction selections to the processor 804. Another type of user input device is a cursor control 816, such as a mouse, joystick, trackball, gesture input device, eye-tracking input device, or cursor direction keys, for communicating directional information and instruction selections to the processor 804 and for controlling cursor movement on the display 812. This input device 814 typically has two degrees of freedom, a first axis (e.g., x) and a second axis (e.g., y), so that the position of the device can be specified on a plane. However, it should be understood that an input device 814 that enables three-dimensional (e.g., x, y, and z) cursor movement is also contemplated herein.

[0259]

[0240] In accordance with certain implementations of this teaching, the computer system 800 may provide results in response to a processor 804 executing one or more sequences of one or more instructions contained in RAM 806. Such instructions may be read into RAM 806 from another computer-readable medium or computer-readable storage medium, such as a storage device 810. The execution of the sequence of instructions contained in RAM 806 can cause the processor 804 to perform the process described herein. Alternatively, hardwired circuits may be used instead of or in combination with software instructions to implement this teaching. Thus, the implementations of this teaching are not limited to any particular combination of hardware circuits and software.

[0260]

[0241] The term “computer-readable medium” (e.g., data store, data storage, memory device, data storage device, etc.) or “computer-readable storage medium” as used herein refers to any medium involved in providing instructions for execution to the processor 804. Such medium can take many forms, including, but are not limited to, non-volatile media, volatile media, and transmission media. Examples of non-volatile media include, but are not limited to, optical disks, solid disks, and magnetic disks, such as in the memory device 810. Examples of volatile media include, but are not limited to, dynamic memory, such as in RAM 806. Examples of transmission media include, but are not limited to, coaxial cables, copper wires, and optical fibers, including wires with bus 802.

[0261]

[0242] Common forms of computer-readable media include, for example, floppy disks, flexible disks, hard disks, magnetic tapes or any other magnetic media, CD-ROMs, any other optical media, punch cards, paper tapes, any other physical media having a pattern of holes, RAM, PROMs and EPROMs, flash EPROMs, any other memory chips or cartridges or any other tangible media that can be read by a computer.

[0262]

[0243] In addition to computer-readable media, instructions or data may be provided as signals on a transmission medium included in communication equipment or a system, which provides a sequence of one or more instructions for execution by a processor 804 of a computer system 800. For example, communication equipment may include transceivers having signals which are instructions for instructions and data. Instructions and data are configured to cause one or more processors to implement functions outlined in the disclosure herein. Typical examples of data communication transmission connections include, but are not limited to, telephone modem connections, wide area networks (WANs), local area networks (LANs), infrared data connections, NFC connections, optical communication connections, and the like.

[0263]

[0244] It should be understood that the methodologies, flowcharts, diagrams and accompanying disclosures described herein may be implemented using the computer system 800 as a standalone device or on a distributed network of shared computing resources, such as a cloud computing network.

[0264]

[0245] The methodologies described herein may be implemented by various means depending on the application. For example, these methodologies may be implemented in hardware, firmware, software, or any combination thereof. In terms of hardware implementation, the processing unit may be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing units (DSPDs), programmable logic units (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, electronic devices, other electronic units designed to perform the functions described herein, or a combination thereof.

[0265]

[0246] In various embodiments, the methods described herein may be implemented as firmware and / or software programs and applications written in conventional programming languages ​​such as C, C++, and Python. When implemented as firmware and / or software, the embodiments described herein may be implemented on a non-temporary computer-readable medium where a program for causing a computer to perform the methods described above is stored. The various engines described herein may be provided to a computer system, such as computer system 800, and it should be understood that the processor 804 may perform the analysis and decision-making provided by these engines, subject to instructions provided by one or a combination thereof of memory elements RAM 806, ROM 808, or storage device 810, and user input provided by input device 814.

[0266] IX. Additional considerations

[0247] The headings and / or subheadings between sections and subsections in this specification are included solely for the purpose of improving readability and do not imply that features should not be combined across sections and subsections. Accordingly, sections and subsections do not describe separate embodiments.

[0267]

[0248] Although this teaching is described in conjunction with various embodiments, it is not intended to limit this teaching to such embodiments. Rather, this teaching encompasses a variety of alternative forms, variations and equivalents as will be understood by those skilled in the art. The description herein provides preferred exemplary embodiments and is not intended to limit the scope, applicability or configuration of the disclosure. Rather, the description herein of preferred exemplary embodiments provides to those skilled in the art an effective explanation for realizing various embodiments.

[0268]

[0249] It is understood that various modifications to the function and configuration of the elements may be made without departing from the spirit and scope set forth in the attached claims. Thus, such improved and modified forms shall be considered to be within the scope set forth in the attached claims. Furthermore, the terms and expressions used are for illustrative purposes only, not limitation, and the use of such terms and expressions is not intended to exclude any equivalents of the illustrated and described features or parts thereof, but it is recognized that various modifications are possible within the scope of the claimed invention.

[0269]

[0250] In the description of various embodiments, methods and / or processes may be presented herein as a detailed set of steps. However, unless such methods or processes depend on a specific sequence of steps shown herein, the methods or processes shall not be limited to the detailed set of steps described herein, and those skilled in the art will readily understand that the set may be modified, but still remain within the spirit and scope of various embodiments.

[0270]

[0251] Some embodiments of the present disclosure include a system including one or more data processors. In some embodiments, the system includes a non-temporary computer-readable storage medium containing instructions that, when executed on one or more data processors, cause one or more data processors to perform some or all of one or more of the methods and / or some or all of one or more processes disclosed herein. Some embodiments of the present disclosure include a computer program product embodied as a tangible object in a non-temporary machine-readable storage medium, which includes instructions configured to cause one or more data processors to perform some or all of one or more of the methods and / or some or all of one or more processes disclosed herein.

[0271]

[0252] The description herein provides specific details to give an understanding of the embodiments. However, it is understood that embodiments can be carried out without such specific details. For example, circuits, systems, networks, processes and other elements may be shown as elements in the form of block diagrams so as not to obscure the embodiments with unnecessary details. In other cases, well-known circuits, processes, algorithms, structures and techniques may be shown without unnecessary details to avoid obscuring the embodiments.

[0272] Embodiment 1. A method for generating an indicator of the risk of sepsis, It must accept at least one input feature corresponding to the patient. Using a machine learning model to analyze at least one input feature and generate a predictive score indicating the probability that a patient will have or develop sepsis within a given period, based on at least one input feature. A method that includes at least one input feature related to sepsis.

[0273] 2. The method according to Embodiment 1, wherein at least one input feature includes a clinical parameter.

[0274] 3. The method according to Embodiment 1 or 2, wherein at least one input feature includes a biomarker.

[0275] 4. The method according to any one of embodiments 1 to 3, further comprising generating a diagnostic output based on a predictive score.

[0276] 5. The method according to any one of Embodiments 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 8.

[0277] 6. The method according to any one of Embodiments 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 9.

[0278] 7. The method according to any one of Embodiments 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 10.

[0279] 8. The method according to any one of Embodiments 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 11.

[0280] 9. The method according to any one of Embodiments 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 12.

[0281] 10. The method according to any one of Embodiments 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 13.

[0282] 11. The method according to any one of Embodiments 5 to 10, wherein the group of input features listed in any one of Tables 1 to 6 is listed in terms of their relative importance to the prediction score.

[0283] 12. The method according to any one of Embodiments 1 to 11, wherein the relative importance to the prediction score is determined based on the rank of the features, the rank of the features is determined by ranking the input features based on their mean SHAP values, where a higher mean SHAP value represents a higher relative importance to the prediction score, and a lower SHAP value represents a lower relative importance to the prediction score.

[0284] 13. The average SHAP value of the input features in any one of the groups of input features listed in Tables 8-13 is: Determining the absolute magnitude of all SHAP values ​​for a given population regarding the input features, This involves summing the absolute magnitudes of all SHAP values ​​to obtain the total magnitude of the input features, The total sum is divided by the number of groups in a given set of input features. The method according to Embodiment 12, calculated by...

[0285] 14. The method according to any one of Embodiments 1 to 13, wherein at least one input feature comprises procalcitonin, and the procalcitonin concentration values ​​are divided into two groups, with the procalcitonin concentration values ​​of the first group having a lower relative importance than the procalcitonin concentration values ​​of the second group.

[0286] 15. The method according to Embodiment 14, wherein the splitting threshold between the first and second groups of procalcitonin concentration values ​​is 2.5 log10 pm / mL.

[0287] 16. The method according to Embodiment 15, wherein the relative importance of procalcitonin to the predictive score in each of the first and second groups is the SHAP median of procalcitonin concentration values ​​within each group.

[0288] 17. The method according to any one of Embodiments 1 to 13, wherein at least one input feature comprises procalcitonin, and the procalcitonin concentration values ​​are divided into at least three groups, the procalcitonin concentration values ​​of the first group having a lower relative importance than the procalcitonin concentration values ​​of the second group, and the procalcitonin concentration values ​​of the third group having a higher relative importance than the procalcitonin concentration values ​​of both the first and second groups.

[0289] 18. The method according to Embodiment 17, wherein the first division threshold between the first and second procalcitonin concentration groups is 1.7 log10 pg / mL, and the second division threshold between the second and third procalcitonin concentration groups is 3.4 log10 pg / mL.

[0290] 19. The method according to Embodiment 18, wherein the relative importance of procalcitonin to the predictive score in each of the first, second, and third groups is the SHAP median of the procalcitonin concentration values ​​within each group.

[0291] 20. The method according to any one of Embodiments 1 to 19, wherein the group of input features listed in Table 1 includes Glasgow Coma Scale, FiO2, procalcitonin, systolic blood pressure, and platelet count.

[0292] 21. The method according to Embodiment 20, wherein the Glasgow Coma Scale has a higher relative importance than FiO2, FiO2 has a higher relative importance than procalcitonin, procalcitonin has a higher relative importance than systolic blood pressure, and systolic blood pressure has a higher relative importance than platelet count.

[0293] 22. The method according to any one of Embodiments 1 to 19, wherein the group of input features listed in Table 1 includes Glasgow Coma Scale, FiO2, procalcitonin, systolic blood pressure, platelet count, bilirubin, blood urea nitrogen, respiratory rate, diastolic blood pressure, and albumin.

[0294] 23. The method according to any one of Embodiments 1 to 19, wherein the group of input features listed in Table 1 includes Glasgow Coma Scale, FiO2, procalcitonin, systolic blood pressure, platelet count, bilirubin, blood urea nitrogen, respiratory rate, diastolic blood pressure, albumin, creatinine, age, blood oxygen saturation, lactate, and C-reactive protein.

[0295] 24. The method according to any one of Embodiments 1 to 19, wherein the group of input features listed in Table 1 includes Glasgow Coma Scale, FiO2, procalcitonin, systolic blood pressure, platelet count, bilirubin, blood urea nitrogen, respiratory rate, diastolic blood pressure, albumin, creatinine, age, blood oxygen saturation, lactate, C-reactive protein, potassium, neutrophil count, temperature, total carbon dioxide, white blood cell count, lymphocyte count, sodium, heart rate, and chloride.

[0296] 25. The method according to any one of Embodiments 1 to 19, wherein the group of input features includes procalcitonin and C-reactive protein.

[0297] 26. The method according to Embodiment 25, wherein either or both procalcitonin and C-reactive protein are replaced with one or more other biomarkers listed in Table 2, according to the correlation values ​​associated in Table 3.

[0298] 27. The method according to Embodiment 26, wherein one or more other biomarkers replacing either or both procalcitonin and C-reactive protein have a correlation value less than XX.

[0299] 28. The decision that the predicted score falls into one of the four categories, Based on the category into which the predicted score is placed, the system generates a diagnostic output that includes a low, moderate, high, or very high probability that the patient will have or develop sepsis within a given period. The method according to any one of embodiments 1 to 27, further comprising the above.

[0300] 29. The method according to Embodiment 28, wherein the prediction score is in the range between 0 and 1.

[0301] 30. The four categories include low, medium, high, and very high categories. The method according to Embodiment 28 or 29, wherein a low category is defined as having a predicted score below a first threshold, a medium category is defined as having a predicted score between a first threshold and a second threshold, a high category is defined as having a predicted score between a second threshold and a third threshold, and a very high category is defined as having a predicted score above a third threshold.

[0302] 31. The method according to Embodiment 30, wherein the predictive score is below a first threshold of 0.122, and the diagnostic output indicates that the patient has a low probability of having or developing sepsis within a given period.

[0303] 32. The method according to Embodiment 30, wherein the predictive score is greater than or equal to a first threshold of 0.122 and less than a second threshold of 0.306, and the diagnostic output indicates that the patient has a moderate probability of having or developing sepsis within a given period.

[0304] 33. The method according to Embodiment 30, wherein the predictive score is greater than or equal to a second threshold of 0.306 and less than a third threshold of 0.872, and the diagnostic output indicates that the patient has a high probability of having or developing sepsis within a given period.

[0305] 34. The method according to Embodiment 30, wherein the predictive score is greater than or equal to a third threshold of 0.872, and the diagnostic output indicates that the patient has a very high probability of having or developing sepsis within a given period.

[0306] 35. Calibrating the predicted score using a calibration model to generate a calibrated score, To generate risk categories as diagnostic output that takes calibrated scores into account. The method according to any one of embodiments 1 to 34, further comprising the above.

[0307] 36. Generating diagnostic output is The calibrated score is determined to fall into one of the four risk categories, Based on the categories into which the calibrated scores are placed, risk categories are generated that include a low, moderate, high, or very high probability that a patient will have or develop sepsis within a given period. The method according to embodiment 35, including the method described in embodiment 35.

[0308] 37. The method according to embodiment 36, wherein the calibrated score is in the range between 0 and 1.

[0309] 38. The method according to any one of embodiments 35 to 37, wherein the risk category may be selected from the group consisting of a low category, a medium category, a high category and a very high category, the low category being specified as having a calibrated score below a first threshold, the medium category being specified as having a calibrated score between a first threshold and a second threshold, the high category being specified as having a calibrated score between a second threshold and a third threshold, and the very high category being specified as having a calibrated score above a third threshold.

[0310] 39. The method according to Embodiment 38, wherein the calibrated score is below a first threshold of 0.122, and the risk category indicates that the patient has a low probability of having or developing sepsis within a given period.

[0311] 40. The method according to Embodiment 38, wherein the calibrated score is above a first threshold of 0.122 and below a second threshold of 0.306, and the risk category indicates that the patient has a moderate probability of having or developing sepsis within a given period.

[0312] 41. The method according to Embodiment 38, wherein the calibrated score is above a second threshold of 0.306 and below a third threshold of 0.872, and the risk category indicates that the patient has a high probability of having or developing sepsis within a given period.

[0313] 42. The method according to Embodiment 38, wherein the calibrated score is above a third threshold of 0.872, and the risk category indicates that the patient has a very high probability of having or developing sepsis within a given period of time.

[0314] 43. The method according to any one of Embodiments 1 to 42, wherein the prescribed period is 12 hours, 18 hours, 24 hours, 36 hours, 48 ​​hours, or 72 hours or less, or longer.

[0315] 44. The method according to any one of Embodiments 1 to 43, further comprising training at least one machine learning model using training data, wherein the training data includes multiple biomarkers and healthy profiles for multiple patients, and multiple target diagnoses for multiple patients.

[0316] 45. The method according to any one of embodiments 4 to 44, wherein the diagnostic output is an automatically derived label based on a predictive score generated by at least one machine learning model.

[0317] 46. ​​The method according to any one of embodiments 4 to 45, wherein the diagnostic output is pronounced by a physician taking into account a predictive score generated by at least one machine learning model.

[0318] 47. The method according to any one of embodiments 4 to 46, wherein the analysis is performed without the value of one clinical parameter out of a plurality of clinical parameters and / or the value of one biomarker out of one or more biomarkers in order to generate a predictive score, and the method further comprises generating a diagnostic output based on the predictive score.

[0319] 48. The method according to any one of Embodiments 1 to 47, further comprising identifying and imputing the value of one of the input features whose value related to the other input feature is missing.

[0320] 49. One of the input features has a missing value, and the method is: Using a decision-making module that triggers logic, the system determines whether to perform the analysis without missing values ​​for one input feature in order to generate a predictive score. The method according to any one of embodiments 1 to 47, further comprising the above.

[0321] 50. One of the input features has a missing value, and the method is: This involves imputing missing values ​​in one input feature to generate an imputed value for the missing input feature, Analyzing data including complementary values ​​to generate a predictive score indicating the probability that a patient will have or develop sepsis within a given period, To generate diagnostic output based on the predicted score and The method according to any one of embodiments 4 to 47, further comprising the above.

[0322] 51. The method according to any one of Embodiments 48 to 50, wherein the imputation of missing values ​​is performed via a pre-trained template trained using training data, the training data comprising input features including multiple clinical parameters and multiple biomarkers for multiple patients, and multiple target diagnoses for multiple patients.

[0323] 52. The method according to any one of embodiments 48 to 51, wherein one input feature having missing values ​​is designated as a feature whose value is measured or obtained outside a specified time window for data that is acceptable for use in generating prediction scores.

[0324] 53. The method according to Embodiment 53, wherein the specified time window is the same for a first set of input features and different for at least two input features of a second set of input features.

[0325] 54. The method according to any one of Embodiments 1 to 53, wherein the input features include clinical parameters having numerical values ​​measured within a specified time window.

[0326] 55. The method according to any one of Embodiments 1 to 54, wherein the input feature comprises a biomarker having a biomarker-related measurement value acquired within a specified time window.

[0327] 56. The method according to any one of Embodiments 1 to 55, wherein the input features include clinical parameters, and the clinical parameters include at least one of demographic measurements, patient assessments, vital signs, blood test values, and chemical test values.

[0328] 57. The method according to any one of embodiments 48 to 50, further comprising identifying and supplementing a biomarker from a group of biomarkers, which has a missing value associated with that biomarker.

[0329] 58. The method according to any one of embodiments 48 to 57, wherein one input feature for which a value related to one input feature is missing is a biomarker from a plurality of biomarkers and / or a clinical parameter from a plurality of clinical parameters, and the biomarker and / or clinical parameter for which a value related to the biomarker and / or clinical parameter is missing.

[0330] 59. The method according to any one of Embodiments 1 to 58, wherein the input features include at least 20 clinical parameters and at least 2 biomarkers.

[0331] 60. The method according to any one of Embodiments 5 to 10, wherein the group of input features listed in any one of Tables 1 to 6 includes at least one biomarker feature and at least one clinical parameter feature, the values ​​of one or more biomarker features are divided into two groups, and the values ​​of one or more clinical parameter features are divided into two groups.

[0332] 61. The method according to Embodiment 60, wherein the biomarker features include at least one of procalcitonin, C-reactive protein, interleukin-6, pentraxin-3, interleukin-8, interleukin-1 receptor antagonist protein, and vascular cell adhesion molecule 1, and the concentration value of each biomarker feature in the first group has a relative importance lower than the concentration value of the biomarker features in the second group.

[0333] 62. The method according to Embodiment 60 or 61, wherein the clinical parameters include at least one of platelets, systolic blood pressure, diastolic blood pressure, albumin, and blood oxygen saturation, and the values ​​of the first group of each clinical parameter feature have a higher relative importance than the concentration values ​​of the second group of clinical parameter features.

[0334] 63. The method according to Embodiment 60 or 61, wherein the clinical parameters include at least one of respiratory rate, blood urea nitrogen, bilirubin, and creatinine, and the values ​​of the first group of each clinical parameter feature have a relative importance lower than the concentration values ​​of the second group of clinical parameter features.

[0335] 64. The method according to Embodiment 60, wherein the biomarker features include procalcitonin, and the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group.

[0336] 65. The method according to Embodiment 60, wherein the biomarker features include C-reactive proteins, and the C-reactive protein concentration value of the first group has a lower relative importance than the C-reactive protein concentration value of the second group.

[0337] 66. The method according to Embodiment 60, wherein the biomarker features include interleukin-6, and the interleukin-6 concentration value of the first group has a lower relative importance than the interleukin-6 concentration value of the second group.

[0338] 67. The method according to Embodiment 60, wherein the biomarker features include pentraxin 3, and the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group.

[0339] 68. The method according to Embodiment 60, wherein the biomarker features include interleukin-8, and the interleukin-8 concentration value of the first group has a lower relative importance than the interleukin-8 concentration value of the second group.

[0340] 69. The method according to Embodiment 60, wherein the biomarker features include interleukin-1 receptor antagonist proteins, and the interleukin-1 receptor antagonist protein concentration values ​​of the first group have a lower relative importance than the interleukin-1 receptor antagonist protein concentration values ​​of the second group.

[0341] 70. The method according to Embodiment 60, wherein the biomarker feature comprises vascular cell adhesion molecule 1, and the concentration value of vascular cell adhesion molecule 1 in the first group has a lower relative importance than the concentration value of vascular cell adhesion molecule 1 in the second group.

[0342] 71. The method according to Embodiment 60, wherein the biomarker features include procalcitonin and C-reactive protein, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, and the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group.

[0343] 72. The method according to Embodiment 60, wherein the biomarker features include procalcitonin and interleukin-6, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, and the interleukin-6 concentration value of the first group having a lower relative importance than the interleukin-6 concentration value of the second group.

[0344] 73. The method according to Embodiment 60, wherein the biomarker features include interleukin-6 and C-reactive protein, the interleukin-6 concentration value of the first group having a lower relative importance than the interleukin-6 concentration value of the second group, and the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group.

[0345] 74. The method according to Embodiment 60, wherein the biomarker features include procalcitonin, C-reactive protein, and interleukin-6, and the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, the C-reactive protein concentration value of the first group has a lower relative importance than the C-reactive protein concentration value of the second group, and the interleukin-6 concentration value of the first group has a lower relative importance than the interleukin-6 concentration value of the second group.

[0346] 75. The method according to Embodiment 60, wherein the biomarker features include pentraxin 3 and interleukin-1 receptor antagonist protein, the pentraxin 3 concentration value of the first group having a lower relative importance than the pentraxin 3 concentration value of the second group, and the interleukin-1 receptor antagonist protein concentration value of the first group having a lower relative importance than the interleukin-1 receptor antagonist protein concentration value of the second group.

[0347] 76. The method according to Embodiment 60, wherein the biomarker features include pentraxin 3 and interleukin 8, the pentraxin 3 concentration value of the first group having a lower relative importance than the pentraxin 3 concentration value of the second group, and the interleukin 8 concentration value of the first group having a lower relative importance than the interleukin 8 concentration value of the second group.

[0348] 77. The method according to Embodiment 60, wherein the biomarker features include procalcitonin and interleukin-8, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, and the interleukin-8 concentration value of the first group having a lower relative importance than the interleukin-8 concentration value of the second group.

[0349] 78. The method according to Embodiment 60, wherein the biomarker features include procalcitonin and pentraxin 3, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, and the pentraxin 3 concentration value of the first group having a lower relative importance than the pentraxin 3 concentration value of the second group.

[0350] 79. The method according to Embodiment 60, wherein the biomarker features include pentraxin 3 and interleukin 6, the pentraxin 3 concentration value of the first group having a lower relative importance than the pentraxin 3 concentration value of the second group, and the interleukin 6 concentration value of the first group having a lower relative importance than the interleukin 6 concentration value of the second group.

[0351] 80. The method according to Embodiment 60, wherein the clinical parameter features include platelets, and the platelet count of the first group has a higher relative importance than the platelet count of the second group.

[0352] 81. The method according to Embodiment 60, wherein the clinical parameter features include systolic blood pressure, and the systolic blood pressure values ​​of the first group have a higher relative importance than the systolic blood pressure values ​​of the second group.

[0353] 82. The method according to Embodiment 60, wherein the clinical parameter features include diastolic blood pressure, and the diastolic blood pressure values ​​of the first group have a higher relative importance than the diastolic blood pressure values ​​of the second group.

[0354] 83. The method according to Embodiment 60, wherein the clinical parameter features include blood oxygen saturation, and the blood oxygen saturation value of the first group has a higher relative importance than the blood oxygen saturation value of the second group.

[0355] 84. The method according to Embodiment 60, wherein the clinical parameter features include albumin, and the blood oxygen saturation value of the first group has a higher relative importance than the albumin value of the second group.

[0356] 85. The method according to Embodiment 60, wherein the clinical parameter features include respiratory rate, and the respiratory value of the first group has a lower relative importance than the respiratory value of the second group.

[0357] 86. The method according to Embodiment 60, wherein the clinical parameter features include blood urea nitrogen, and the blood urea nitrogen value of the first group has a lower relative importance than the blood urea nitrogen value of the second group.

[0358] 87. The method according to Embodiment 60, wherein the clinical parameter features include bilirubin, and the bilirubin value of the first group has a lower relative importance than the bilirubin value of the second group.

[0359] 88. The method according to Embodiment 60, wherein the clinical parameter features include creatinine, and the creatinine value of the first group has a lower relative importance than the creatinine value of the second group.

[0360] 89. The method according to Embodiment 60, wherein the clinical parameter features include creatinine and blood urea nitrogen, the creatinine value of group 1 has a lower relative importance than the creatinine value of group 2, and the blood urea nitrogen value of group 1 has a lower relative importance than the blood urea nitrogen value of group 2.

[0361] 90. The method according to Embodiment 60, wherein the clinical parameter features include platelets, creatinine, and blood urea nitrogen, and the clinical parameter features include platelets, the platelet value of group 1 has a higher relative importance than the platelet value of group 2, the creatinine value of group 1 has a lower relative importance than the creatinine value of group 2, and the blood urea nitrogen value of group 1 has a lower relative importance than the blood urea nitrogen value of group 2.

[0362] 91. The method according to Embodiment 60, wherein the clinical parameter features include respiratory rate, creatinine, and blood urea nitrogen, and the clinical parameter features include platelets, the respiratory value of the first group has a lower relative importance than the respiratory value of the second group, the creatinine value of the first group has a lower relative importance than the creatinine value of the second group, and the blood urea nitrogen value of the first group has a lower relative importance than the blood urea nitrogen value of the second group.

[0363] 92. The method according to Embodiment 60, wherein the clinical parameter features include platelet count and respiratory rate, the platelet count of the first group has a higher relative importance than the platelet count of the second group, and the respiratory rate of the first group has a lower relative importance than the respiratory rate of the second group.

[0364] 93. The method according to Embodiment 60, wherein the clinical parameter features include platelets and systolic blood pressure, the clinical parameter features include platelets, the platelet count of the first group has a higher relative importance than the platelet count of the second group, and the systolic blood pressure count of the first group has a higher relative importance than the systolic blood pressure count of the second group.

[0365] 94. The method according to Embodiment 60, wherein the clinical parameter features include platelets, respiratory rate, and systolic blood pressure, and the clinical parameter features include platelets, the platelet count of the first group has a higher relative importance than the platelet count of the second group, the respiratory count of the first group has a lower relative importance than the respiratory count of the second group, and the systolic blood pressure count of the first group has a higher relative importance than the systolic blood pressure count of the second group.

[0366] 95. The method according to Embodiment 60, wherein the relative importance of one or more features to the prediction score in each of the first and second groups is the SHAP median of the feature values ​​in each group.

[0367] 96. The method according to Embodiment 95, wherein the relative importance of one or more biomarker features to the prediction score in each of the first and second groups is the SHAP median of the biomarker feature values ​​within each group.

[0368] 97. The method according to Embodiment 95, wherein the relative importance of one or more clinical parameter features to the predictive score in each of the first and second groups is the SHAP median of the clinical parameter feature values ​​in each group.

[0369] 98. The method according to Embodiment 60, wherein the relative importance of one or more features to the prediction score in each of the first and second groups is the SHAP mean of the feature values ​​within each group.

[0370] 99. The method according to Embodiment 98, wherein the relative importance of one or more biomarker features to the prediction score in each of the first and second groups is the SHAP mean of the biomarker feature values ​​within each group.

[0371] 100. The method according to Embodiment 98, wherein the relative importance of one or more clinical parameter features to the predictive score in each of the first and second groups is the SHAP mean of the biomarker feature values ​​within each group.

[0372] 101. The method according to any one of Embodiments 61 to 100, wherein the splitting threshold between the first and second groups of procalcitonin concentration values ​​is 2.24 log10 pg / mL.

[0373] 102. The method according to any one of Embodiments 61 to 100, wherein the split threshold between the first and second groups of pentraxin 3 concentration values ​​is 3.98 log10 pg / mL.

[0374] 103. The method according to any one of Embodiments 61 to 100, wherein the partition threshold between the first and second groups of interleukin-8 concentration values ​​is 1.41 log10 pg / mL.

[0375] 104. The method according to any one of Embodiments 61 to 100, wherein the partition threshold between the first and second groups of interleukin-1 receptor antagonist protein concentrations is 3.48 log10 pg / mL.

[0376] 105. The method according to any one of Embodiments 61 to 100, wherein the partition threshold between the first and second groups of interleukin-6 concentration values ​​is 2.07 log10 pg / mL.

[0377] 106. The method according to any one of Embodiments 61 to 100, wherein the partition threshold between the first and second groups of C-reactive protein concentration values ​​is 7.63 log10 pg / mL.

[0378] 107. The method according to any one of embodiments 61 to 100, wherein the partition threshold between the first and second groups of vascular cell adhesion molecule 1 concentration values ​​is 6.17 log10 pg / mL.

[0379] 108. The method according to any one of embodiments 61 to 100, wherein the partition threshold between the first and second groups of platelet concentration values ​​is 162 10^9 / L.

[0380] 109. The method according to any one of embodiments 61 to 100, wherein the dividing threshold between the first and second groups of systolic blood pressure values ​​is 101 mm Hg.

[0381] 110. The method according to any one of embodiments 61 to 100, wherein the dividing threshold between the first and second groups of respiratory values ​​is 101 breaths per minute.

[0382] 111. The method according to any one of embodiments 61 to 100, wherein the partition threshold between the first and second groups of blood urea nitrogen levels is 24 mg / dL.

[0383] 112. The method according to any one of Embodiments 61 to 100, wherein the partition threshold between the first and second bilirubin levels is 1.19 mg / dL.

[0384] 113. The method according to any one of Embodiments 61 to 100, wherein the split threshold between the first and second groups of creatinine levels is 1.56 mg / dL.

[0385] 114. The method according to any one of embodiments 61 to 100, wherein the dividing threshold between the first and second groups of diastolic blood pressure values ​​is 54.5 mm Hg.

[0386] 115. The method according to any one of embodiments 61 to 100, wherein the dividing threshold between the first and second blood oxygen saturation groups is 91%.

[0387] 116. The method according to any one of Embodiments 5 to 10, wherein the group of input features listed in any one of Tables 1 to 6 includes at least one biomarker feature and at least one clinical parameter feature, the values ​​of one or more biomarker features are divided into at least three groups, and the values ​​of one or more clinical parameter features are divided into at least three groups.

[0388] 117. The method according to Embodiment 116, wherein the biomarker features include at least one of procalcitonin, C-reactive protein, interleukin-6, pentraxin-3, interleukin-8, interleukin-1 receptor antagonist protein, and vascular cell adhesion molecule 1, and the concentration value of the first group of each biomarker feature has a relative importance lower than the concentration value of the second group of biomarker features, and the concentration value of the third group of biomarker features has a relative importance higher than the concentration values ​​of both the first and second groups.

[0389] 118. The method according to Embodiment 118, wherein the clinical parameters include at least one of platelets, systolic blood pressure, diastolic blood pressure, albumin, and pulse oximetry, wherein the values ​​of the first group of each clinical parameter feature have a higher relative importance than the concentration values ​​of the second group of clinical parameter features, and the concentration values ​​of the third group of clinical parameter features have a lower relative importance than the clinical parameter values ​​of both the first and second groups.

[0390] 119. The method according to Embodiment 118, wherein the clinical parameters include at least one of respiratory rate, blood urea nitrogen, bilirubin, and creatinine, and the value of each clinical parameter feature in the first group has a relative importance lower than the concentration value of each clinical parameter feature in the second group, and each clinical parameter feature in the third group has a relative importance higher than the clinical parameter feature values ​​of both the first and second groups.

[0391] 120. The method according to Embodiment 116, wherein the biomarker feature includes procalcitonin, the procalcitonin concentration value of group 1 has a lower relative importance than the procalcitonin concentration value of group 2, and the procalcitonin concentration value of group 3 has a higher relative importance than the procalcitonin concentration values ​​of both groups 1 and 2.

[0392] 121. The method according to Embodiment 116, wherein the biomarker features include C-reactive proteins, the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group, and the C-reactive protein concentration value of the third group having a higher relative importance than the C-reactive protein concentration values ​​of both the first and second groups.

[0393] 122. The method according to Embodiment 116, wherein the biomarker features include interleukin-6, the interleukin-6 concentration value of the first group has a lower relative importance than the interleukin-6 concentration value of the second group, and the interleukin-6 concentration value of the third group has a higher relative importance than both the interleukin-6 concentration values ​​of the first and second groups.

[0394] 123. The method according to Embodiment 116, wherein the biomarker feature includes pentraxin 3, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, and the pentraxin 3 concentration value of the third group has a higher relative importance than both the pentraxin 3 concentration values ​​of the first and second groups.

[0395] 124. The method according to Embodiment 116, wherein the biomarker features include interleukin-8, the interleukin-8 concentration value of the first group has a lower relative importance than the interleukin-8 concentration value of the second group, and the interleukin-8 concentration value of the third group has a higher relative importance than both the interleukin-8 concentration values ​​of the first and second groups.

[0396] 125. The method according to Embodiment 116, wherein the biomarker features include interleukin-1 receptor antagonist proteins, the interleukin-1 receptor antagonist protein concentration values ​​of the first group have a lower relative importance than the interleukin-1 receptor antagonist protein concentration values ​​of the second group, and the interleukin-1 receptor antagonist protein concentration values ​​of the third group have a higher relative importance than the interleukin-1 receptor antagonist protein concentration values ​​of both the first and second groups.

[0397] 126. The method according to Embodiment 116, wherein the biomarker feature comprises vascular cell adhesion molecule 1, the concentration value of vascular cell adhesion molecule 1 in the first group has a lower relative importance than the concentration value of vascular cell adhesion molecule 1 in the second group, and the concentration value of vascular cell adhesion molecule 1 in the third group has a higher relative importance than the concentration values ​​of vascular cell adhesion molecule 1 in both the first and second groups.

[0398] 127. The method according to Embodiment 116, wherein the biomarker features include procalcitonin and C-reactive protein, the procalcitonin concentration value of group 1 has a lower relative importance than the procalcitonin concentration value of group 2, the procalcitonin concentration value of group 3 has a higher relative importance than both the procalcitonin concentration values ​​of group 1 and group 2, the C-reactive protein concentration value of group 1 has a lower relative importance than the C-reactive protein concentration value of group 2, and the C-reactive protein concentration value of group 3 has a higher relative importance than both the C-reactive protein concentration values ​​of group 1 and group 2.

[0399] 128. The method according to Embodiment 116, wherein the biomarker features include procalcitonin and interleukin-6, the procalcitonin concentration value of group 1 has a lower relative importance than the procalcitonin concentration value of group 2, the procalcitonin concentration value of group 3 has a higher relative importance than both the procalcitonin concentration values ​​of group 1 and group 2, the interleukin-6 concentration value of group 1 has a lower relative importance than the interleukin-6 concentration value of group 2, and the interleukin-6 concentration value of group 3 has a higher relative importance than both the interleukin-6 concentration values ​​of group 1 and group 2.

[0400] 129. The method according to Embodiment 116, wherein the biomarker features include C-reactive protein and interleukin-6, the C-reactive protein concentration value of group 1 has a lower relative importance than the C-reactive protein concentration value of group 2, the C-reactive protein concentration value of group 3 has a higher relative importance than both the C-reactive protein concentration values ​​of group 1 and group 2, the interleukin-6 concentration value of group 1 has a lower relative importance than the interleukin-6 concentration value of group 2, and the interleukin-6 concentration value of group 3 has a higher relative importance than both the interleukin-6 concentration values ​​of group 1 and group 2.

[0401] 130. The method according to Embodiment 116, wherein the biomarker features include procalcitonin, C-reactive protein, and interleukin-6, the procalcitonin concentration value of group 1 has a lower relative importance than the procalcitonin concentration value of group 2, the procalcitonin concentration value of group 3 has a higher relative importance than both the procalcitonin concentration values ​​of group 1 and group 2, the C-reactive protein concentration value of group 1 has a lower relative importance than the C-reactive protein concentration value of group 2, the C-reactive protein concentration value of group 3 has a higher relative importance than both the C-reactive protein concentration values ​​of group 1 and group 2, the interleukin-6 concentration value of group 1 has a lower relative importance than the interleukin-6 concentration value of group 2, and the interleukin-6 concentration value of group 3 has a higher relative importance than both the interleukin-6 concentration values ​​of group 1 and group 2.

[0402] 131. The method according to Embodiment 116, wherein the biomarker features include pentraxin 3 and interleukin-1 receptor antagonist protein, the pentraxin 3 concentration value of group 1 has a lower relative importance than the pentraxin 3 concentration value of group 2, the pentraxin 3 concentration value of group 3 has a higher relative importance than both the pentraxin 3 concentration values ​​of group 1 and group 2, the interleukin-1 receptor antagonist protein concentration value of group 1 has a lower relative importance than the interleukin-1 receptor antagonist protein concentration value of group 2, and the interleukin-1 receptor antagonist protein concentration value of group 3 has a higher relative importance than both the interleukin-1 receptor antagonist protein concentration values ​​of group 1 and group 2.

[0403] 132. The method according to Embodiment 116, wherein the biomarker features include pentraxin 3 and interleukin 8, the pentraxin 3 concentration value of group 1 has a lower relative importance than the pentraxin 3 concentration value of group 2, the pentraxin 3 concentration value of group 3 has a higher relative importance than both the pentraxin 3 concentration values ​​of group 1 and group 2, the interleukin 8 concentration value of group 1 has a lower relative importance than the interleukin 8 concentration value of group 2, and the interleukin 8 concentration value of group 3 has a higher relative importance than both the interleukin 8 concentration values ​​of group 1 and group 2.

[0404] 133. The method according to Embodiment 116, wherein the biomarker features include procalcitonin and interleukin-8, the procalcitonin concentration value of group 1 has a lower relative importance than the procalcitonin concentration value of group 2, the procalcitonin concentration value of group 3 has a higher relative importance than both the procalcitonin concentration values ​​of group 1 and group 2, the interleukin-8 concentration value of group 1 has a lower relative importance than the interleukin-8 concentration value of group 2, and the interleukin-8 concentration value of group 3 has a higher relative importance than both the interleukin-8 concentration values ​​of group 1 and group 2.

[0405] 134. The method according to Embodiment 116, wherein the biomarker features include procalcitonin and pentraxin 3, the procalcitonin concentration value of group 1 has a lower relative importance than the procalcitonin concentration value of group 2, the procalcitonin concentration value of group 3 has a higher relative importance than both the procalcitonin concentration values ​​of group 1 and group 2, the pentraxin 3 concentration value of group 1 has a lower relative importance than the pentraxin 3 concentration value of group 2, and the pentraxin 3 concentration value of group 3 has a higher relative importance than both the pentraxin 3 concentration values ​​of group 1 and group 2.

[0406] 135. The method according to Embodiment 116, wherein the biomarker features include pentraxin 3 and interleukin 6, the pentraxin 3 concentration value of group 1 has a lower relative importance than the pentraxin 3 concentration value of group 2, the pentraxin 3 concentration value of group 3 has a higher relative importance than both the pentraxin 3 concentration values ​​of group 1 and group 2, the interleukin 6 concentration value of group 1 has a lower relative importance than the interleukin 6 concentration value of group 2, and the interleukin 6 concentration value of group 3 has a higher relative importance than both the interleukin 6 concentration values ​​of group 1 and group 2.

[0407] 136. The method according to Embodiment 116, wherein the clinical parameter features include platelets, the platelet value of the first group has a higher relative importance than the platelet value of the second group, and the platelet value of the third group has a lower relative importance than both the platelet values ​​of the first and second groups.

[0408] 137. The method according to Embodiment 116, wherein the clinical parameter features include systolic blood pressure, the systolic blood pressure values ​​of the first group have a higher relative importance than the systolic blood pressure values ​​of the second group, and the systolic blood pressure values ​​of the third group have a lower relative importance than the systolic blood pressure values ​​of both the first and second groups.

[0409] 138. The method according to Embodiment 116, wherein the clinical parameter features include diastolic blood pressure, the diastolic blood pressure values ​​of the first group have a higher relative importance than the diastolic blood pressure values ​​of the second group, and the diastolic blood pressure values ​​of the third group have a lower relative importance than the diastolic blood pressure values ​​of both the first and second groups.

[0410] 139. The method according to Embodiment 116, wherein the clinical parameter features include pulse oximetry, the pulse oximetry values ​​of the first group having a higher relative importance than the pulse oximetry values ​​of the second group, and the pulse oximetry values ​​of the third group having a lower relative importance than the pulse oximetry values ​​of both the first and second groups.

[0411] 140. The method according to Embodiment 116, wherein the clinical parameter features include albumin, the albumin value of the first group has a higher relative importance than the albumin value of the second group, and the albumin value of the third group has a lower relative importance than both the albumin values ​​of the first and second groups.

[0412] 141. The method according to Embodiment 116, wherein the clinical parameter features include respiratory rate, the respiratory value of the first group has a higher relative importance than the respiratory value of the second group, and the respiratory value of the third group has a lower relative importance than both the respiratory values ​​of the first and second groups.

[0413] 142. The method according to Embodiment 116, wherein the clinical parameter features include blood urea nitrogen, the blood urea nitrogen value of the first group having a higher relative importance than the blood urea nitrogen value of the second group, and the blood urea nitrogen value of the third group having a lower relative importance than the blood urea nitrogen values ​​of both the first and second groups.

[0414] 143. The method according to Embodiment 116, wherein the clinical parameter features include bilirubin, the bilirubin value of group 1 has a higher relative importance than the bilirubin value of group 2, and the bilirubin value of group 3 has a lower relative importance than the bilirubin values ​​of both groups 1 and 2.

[0415] 144. The method according to Embodiment 116, wherein the clinical parameter feature includes creatinine, the creatinine value of group 1 has a higher relative importance than the creatinine value of group 2, and the creatinine value of group 3 has a lower relative importance than the creatinine values ​​of both groups 1 and 2.

[0416] 145. The method according to Embodiment 116, wherein the clinical parameter features include creatinine and blood urea nitrogen, the creatinine value of group 1 has a higher relative importance than the creatinine value of group 2, the creatinine value of group 3 has a lower relative importance than both the creatinine values ​​of group 1 and group 2, the blood urea nitrogen value of group 1 has a higher relative importance than the blood urea nitrogen value of group 2, and the blood urea nitrogen value of group 3 has a lower relative importance than both the blood urea nitrogen values ​​of group 1 and group 2.

[0417] 146. The method according to Embodiment 116, wherein the clinical parameter features include platelets, creatinine, and blood urea nitrogen, the platelet value of group 1 having a higher relative importance than the platelet value of group 2, the platelet value of group 3 having a lower relative importance than both the platelet values ​​of group 1 and group 2, the creatinine value of group 1 having a higher relative importance than the creatinine value of group 2, the creatinine value of group 3 having a lower relative importance than both the creatinine values ​​of group 1 and group 2, the blood urea nitrogen value of group 1 having a higher relative importance than the blood urea nitrogen value of group 2, and the blood urea nitrogen value of group 3 having a lower relative importance than both the blood urea nitrogen values ​​of group 1 and group 2.

[0418] 147. The method according to Embodiment 116, wherein the clinical parameter features include respiratory rate, creatinine, and blood urea nitrogen, the respiratory value of group 1 having a higher relative importance than the respiratory value of group 2, the respiratory value of group 3 having a lower relative importance than both the respiratory values ​​of group 1 and group 2, the creatinine value of group 1 having a higher relative importance than the creatinine value of group 2, the creatinine value of group 3 having a lower relative importance than both the creatinine values ​​of group 1 and group 2, the blood urea nitrogen value of group 1 having a higher relative importance than the blood urea nitrogen value of group 2, and the blood urea nitrogen value of group 3 having a lower relative importance than both the blood urea nitrogen values ​​of group 1 and group 2.

[0419] 148. The method according to Embodiment 116, wherein the clinical parameter features include platelet count and respiratory rate, the platelet count of group 1 has a higher relative importance than the platelet count of group 2, the platelet count of group 3 has a lower relative importance than both the platelet counts of group 1 and group 2, the respiratory rate of group 1 has a higher relative importance than the respiratory rate of group 2, and the respiratory rate of group 3 has a lower relative importance than both the respiratory rates of group 1 and group 2.

[0420] 149. The method according to Embodiment 116, wherein the clinical parameter features include platelet count and systolic blood pressure, wherein the platelet count of group 1 has a higher relative importance than the platelet count of group 2, the platelet count of group 3 has a lower relative importance than both the platelet counts of group 1 and group 2, the systolic blood pressure of group 1 has a higher relative importance than the systolic blood pressure of group 2, and the systolic blood pressure of group 3 has a lower relative importance than both the systolic blood pressure of group 1 and group 2.

[0421] 150. The method according to Embodiment 116, wherein the clinical parameter features include platelet count, respiratory rate, and systolic blood pressure, wherein the platelet count of group 1 has a higher relative importance than the platelet count of group 2, the platelet count of group 3 has a lower relative importance than both the platelet counts of group 1 and group 2, the respiratory count of group 1 has a higher relative importance than the respiratory count of group 2, the respiratory count of group 3 has a lower relative importance than both the respiratory counts of group 1 and group 2, the systolic blood pressure of group 1 has a higher relative importance than the systolic blood pressure of group 2, and the systolic blood pressure of group 3 has a lower relative importance than both the systolic blood pressure of group 1 and group 2.

[0422] 151. The method according to Embodiment 116, wherein the relative importance of one or more features to the prediction score in each of the first, second, and third groups is the SHAP median of the feature values ​​within each group.

[0423] 152. The method according to Embodiment 151, wherein the relative importance of one or more biomarker features to the prediction score in each of the first, second, and third groups is the SHAP median of the biomarker feature values ​​within each group.

[0424] 153. The method according to Embodiment 151, wherein the relative importance of one or more clinical parameter features to the predictive score in each of the first, second, and third groups is the SHAP median of the clinical parameter feature values ​​within each group.

[0425] 154. The method according to Embodiment 116, wherein the relative importance of one or more features to the prediction score in each of the first, second, and third groups is the SHAP mean of the feature values ​​within each group.

[0426] 155. The method according to Embodiment 154, wherein the relative importance of one or more biomarker features to the prediction score in each of the first, second, and third groups is the SHAP mean of the biomarker feature values ​​within each group.

[0427] 156. The method according to Embodiment 154, wherein the relative importance of one or more clinical parameter features to the predictive score in each of the first, second, and third groups is the SHAP mean of the clinical parameter feature values ​​within each group.

[0428] 157. The method according to any one of embodiments 116 to 156, wherein the split threshold between the first and second procalcitonin concentration groups is 2.05 log10 pg / mL, and the second split threshold between the second and third procalcitonin concentration groups is 2.85 log10 pg / mL.

[0429] 158. The method according to any one of Embodiments 116 to 156, wherein the split threshold between the first and second groups of pentraxin 3 concentration values ​​is 3.61 log10 pg / mL, and the second split threshold between the second and third groups of pentraxin 3 concentration values ​​is 3.98 log10 pg / mL.

[0430] 159. The method according to any one of embodiments 116 to 156, wherein the partition threshold between the first and second groups of interleukin-8 concentration values ​​is 1.36 log10 pg / mL, and the second partition threshold between the second and third groups of interleukin-8 concentration values ​​is 1.97 log10 pg / mL.

[0431] 160. The method according to any one of embodiments 116 to 156, wherein the split threshold between the first and second groups of interleukin-1 receptor antagonist protein concentrations is 3.15 log10 pg / mL, and the second split threshold between the second and third groups of interleukin-1 receptor antagonist protein concentrations is 3.57 log10 pg / mL.

[0432] 161. The method according to any one of embodiments 116 to 156, wherein the split threshold between the first and second groups of interleukin-6 concentration values ​​is 3.15 log10 pg / mL, and the second split threshold between the second and third groups of interleukin-6 concentration values ​​is 3.57 log10 pg / mL.

[0433] 162. The method according to any one of Embodiments 116 to 156, wherein the partition threshold between the first and second groups of C-reactive protein concentration values ​​is 7.31 log10 pg / mL, and the second partition threshold between the second and third groups of C-reactive protein concentration values ​​is 7.91 log10 pg / mL.

[0434] 163. The method according to any one of embodiments 116 to 156, wherein the division threshold between the first and second groups of vascular cell adhesion molecule 1 concentration values ​​is 6.05 log10 pg / mL, and the second division threshold between the second and third groups of vascular cell adhesion molecule 1 concentration values ​​is 6.17 log10 pg / mL.

[0435] 164. The method according to any one of embodiments 116 to 156, wherein the division threshold between the first and second groups of platelet concentration values ​​is 139 10^9 / L, and the second division threshold between the second and third groups of platelet concentration values ​​is 188 10^9 / L.

[0436] 165. The method according to any one of embodiments 116 to 156, wherein the dividing threshold between the first and second groups of systolic blood pressure values ​​is 91 mm Hg, and the second dividing threshold between the second and third groups of systolic blood pressure values ​​is 111 mm Hg.

[0437] 166. The method according to any one of embodiments 116 to 156, wherein the dividing threshold between the first and second groups of respiratory values ​​is 23.5 breaths per minute, and the second dividing threshold between the second and third groups of respiratory values ​​is 28 breaths per minute.

[0438] 167. The method according to any one of embodiments 116 to 156, wherein the dividing threshold between the first and second groups of blood urea nitrogen levels is 21 mg / dL, and the second dividing threshold between the second and third groups of blood urea nitrogen levels is 41 mg / dL.

[0439] 168. The method according to any one of Embodiments 116 to 156, wherein the split threshold between the first and second bilirubin groups is 0.88 mg / dL, and the second split threshold between the second and third bilirubin groups is 1.5 mg / dL.

[0440] 169. The method according to any one of embodiments 116 to 156, wherein the split threshold between the first and second creatinine groups is 1.15 mg / dL, and the second split threshold between the second and third creatinine groups is 1.73 mg / dL.

[0441] 170. The method according to any one of embodiments 116 to 156, wherein the dividing threshold between the first and second groups of diastolic blood pressure values ​​is 52 mm Hg, and the second dividing threshold between the second and third groups of diastolic blood pressure values ​​is 66 mm Hg.

[0442] 171. The method according to any one of embodiments 116 to 156, wherein the dividing threshold between the first and second groups of blood oxygen saturation values ​​is 81 mm Hg, and the second dividing threshold between the second and third groups of blood oxygen saturation values ​​is 92.5 mm Hg.

[0443] 172. A method comprising generating a diagnostic output based on a predictive score from the method of any one of Embodiments 4 to 171, classifying a patient as having sepsis or at risk of developing sepsis within a given period of time, the method further comprising administering a therapeutically effective dose of treatment for sepsis to the individual.

[0444] 173. The method according to Embodiment 172, wherein the treatment comprises at least one of an antibiotic or an intravenous fluid based on the generated diagnostic output.

[0445] 174. A method for treating an individual having a score based on a plurality of input features indicating the probability that a patient has or will develop sepsis, based on the method of any one of Embodiments 1 to 171, wherein the treatment comprises administering an effective amount of one or more sepsis therapies to the individual.

[0446] 175. A method comprising measuring a plurality of input features in Tables 8, 9, 10, 11, 12, or 13 from an individual and / or a sample from an individual, wherein, optionally, the measurements of at least a first input feature and at least a second input feature are sequential, and the measurement of the second input feature is based on the result of the measurement of the first input feature.

[0447] 176. The method of Embodiment 175, comprising measuring the majority of the multiple input features in Table 8 for an individual.

[0448] 177. The method according to Embodiment 175 or 176, wherein the individual is suspected of having sepsis, suspected of developing sepsis, or at risk of developing sepsis, or the individual has one or more symptoms of sepsis.

[0449] 178. The method according to any one of embodiments 175 to 177, wherein the individual is suspected to develop sepsis within 24 hours of measurement of the individual.

[0450] 179. The method according to any one of Embodiments 175 to 178, wherein the sample is blood, urine, tissue, skin, saliva, sputum and / or mucus.

[0451] 180. The method according to any one of embodiments 175 to 179, wherein the individual has an infectious disease.

[0452] 181. The method according to Embodiment 180, wherein the infectious agent is a virus, bacteria, fungus, or prion.

[0453] 182. The method according to Embodiment 180 or 181, wherein the infectious disease is SARS-CoV-2, influenza, meningitis, pneumonia, tuberculosis, or hepatitis.

[0454] 183. The method according to any one of Embodiments 175 to 182, wherein the individual is over approximately 65 years of age, under approximately 1 year of age, is immunocompromised, has a chronic medical condition, has recently suffered from a serious illness or been hospitalized, and / or has previously suffered from sepsis.

[0455] 184. The method according to Embodiment 183, wherein the chronic medical condition is diabetes, lung disease, cancer, heart disease, hepatitis, or kidney disease.

[0456] 185. A method comprising measuring one, two, three, four, five, six, seven, eight, nine, ten, one, two, ten, one, two, ten, one, or more input features of an individual in one or more of the tables 8, 9, 10, 11, 12, or 13.

[0457] 186. The method according to Embodiment 185, wherein the individual has one or more symptoms of sepsis.

[0458] 187. The method according to Embodiment 185 or 186, wherein the individual is suspected of having sepsis, suspected of developing sepsis, or at risk of developing sepsis.

[0459] 188. The method according to any one of Embodiments 185 to 187, wherein the individual is suspected to develop sepsis within 24 hours of measurement of a sample obtained from the individual.

[0460] 189. The method according to any one of embodiments 185 to 188, wherein the individual has an infectious disease.

[0461] 190. The method according to Embodiment 189, wherein the infectious agent is a virus, bacteria, fungus, or prion.

[0462] 191. The method according to Embodiment 189 or 190, wherein the infectious disease is SARS-CoV-2, influenza, meningitis, pneumonia, tuberculosis, or hepatitis.

[0463] 192. The method according to any one of Embodiments 185 to 191, wherein the individual is over approximately 65 years of age, under approximately 1 year of age, immunocompromised, has a chronic medical condition, has recently suffered from or been hospitalized for a serious illness, or has previously suffered from sepsis.

[0464] 193. The method according to Embodiment 192, wherein the chronic medical condition is diabetes, lung disease, cancer, heart disease, or kidney disease.

[0465] 194. The method according to any one of Embodiments 1 to 171, wherein the plurality of input features include at least one input feature from a first group, at least one input feature from a second group, and at least one input feature from a third group.

[0466] 195. The first group includes input features selected from the group consisting of age, sex at birth, race, ethnicity, patient's medical history, patient's current illness or symptoms, neurological evaluation, clinical decision support alerts, clinician or other medical records, diagnostic codes, procedures performed on the patient, patient's current drug therapy, interventions, patient treatment environment, medical imaging data or evaluations, electrocardiograms, endoscopy, systolic blood pressure, diastolic blood pressure, body temperature, respiratory rate, heart rate, blood oxygen saturation, inhaled oxygen fraction, and combinations thereof. The second group includes input features selected from a group consisting of white blood cell count, lymphocyte count, neutrophil count, platelet count, creatinine, blood urea nitrogen, potassium, chloride, total carbon dioxide, sodium, albumin, bilirubin, lactate, glucose, calcium, hematocrit, hemoglobin concentration, laboratory-derived biomarkers, clinical severity scale and overall score, biopsy data, microbiological test data, and combinations thereof. The method according to Embodiment 194, wherein the third group includes input features selected from the group consisting of procalcitonin, C-reactive protein, protein biomarkers, genomic biomarkers, gene expression or transcriptome biomarkers, composite biomarker scores, and combinations thereof.

[0467] 196. The first group is, Vital signs (e.g., body temperature, heart rate, blood pressure, respiratory rate, oxygen saturation), Demographic attributes (e.g., age, sex, race, ethnicity), medical history, Current illness or symptoms, Neurological assessment (Glasgow Coma Scale, full summary of no response (FOUR)), Clinical decision support alerts, Clinician or other medical records, Diagnostic code (e.g., ICD-10), Procedure (including CPT code), Current drug therapies (including NDC codes), intervention, Patient treatment environment (ICU, ED, hospital bed, outpatient treatment, etc.), Imaging methods (e.g., X-ray, CT scan, MRI, fMRI, ultrasound), Electrical recording diagrams (e.g., EEG, EKG), Endoscopic examination, and These combinations The method according to Embodiment 194 or 195, comprising input features selected from the group consisting of the following.

[0468] 197. The second group is, Clinical test values, Laboratory-derived biomarkers (e.g., monocyte distribution width, Cytovale IntelliSep), Clinical severity scales and overall scores (e.g., Sequential Organ Failure Assessment (SOFA) score, SOFA component score, Charlson Comorbidity Index, Acute Physiological and Chronic Health Assessment (APACHE) I and II), biopsy, Microbiological testing (culture, PCR testing, antigen testing), and These combinations The method according to any one of embodiments 194 to 196, comprising an input feature selected from the group consisting of the following.

[0469] 198. The third group is, Protein biomarkers, Genome biomarkers, Gene expression or transcriptome biomarkers, Composite biomarker score, and These combinations The method according to any one of embodiments 194 to 196, comprising an input feature selected from the group consisting of the following.

[0470] 199. Protein markers include angiopoietin-1, angiopoietin-2, C-reactive protein, cystatin C, D-dimer, E-selectin, fractalkine, FLT3-ligand, GCSF, GDF15, GMCSF, granzyme-B, IFN-α, IFN-γ, IL1-β, IL1-RA, IL-2, IL-4, IL-6, IL-7, IL-8, IL-10, IL-15, IP-10, and lactate dehydrogenase (LDH). The method according to Embodiment 198, comprising a selection from the group consisting of lipopolysaccharide-binding protein (LBP), leptin, MCP1, MIP1-α, MIP1-β, MIP3-α, NGAL, pancreatic stone protein, PDL-1, pentraxin-3, procalcitonin, protein C, S100b, TGF-α, thrombomodulin, tissue factor, TNF-α, TRAIL, TREM-1, troponin, VCAM-1, VEG-F, and combinations thereof.

[0471] 200. Gene expression or transcriptome biomarkers include CEACAM4, LAMP1, PLAC8, PLA2G7, CEACAM1, ZDHHC19, C9orf95, GNA15, BATF, C3AR1, KIAA1370, TGFBI, MTCH1, RPGRIP1, HLA-DPB1, nuclear factor kappa B (NF-κB), heme oxygenase-1 (HO-1), FCGR2C, CRP, IL-6, IL-10, IL-12, TNF, LY6E, LAMP3, ISG15, USP18, RSAD2, IFI44L, NCOA7, CMPK2, IFI44, BATF2, GBP4, IFIH1, GMPR, CXCL10, IFI27, ISG15, CCL2, OAS3, P2RY14, GCH1, CD274, ID01, ARL4A, BPGM, TNFAIP3, TNFAIP6, IL1R1, PROK2, RGS1, C XCR6, RTP4, TLR3, FZD5, KCNJ2, IRAK2, RGS16, CXCL8, OLR1, CCRL2, SPP1, ST3GAL4, CST7, IL1R2, IL4R, GALM, CTLA4, SOCS1, BCL2L1, TGM2, SLC39A8, CA2, PNP, GATA1, SLC1A5, RGS16, CD44, PDGFC, CSF2RA, CXCL1, CSF3R, SOCS3, IL18R1, PIM1, LEPR, IL1B, CSF2RB, ITGB3, STAT1, FAS, EFNA1, IER3, EETS2, SLC2 A3, BCL2A1, PFKFB3, CEBPB, BIRC2, ATF3, TUBB2A, G0S2, MERTK, BIK, QSOX1, PYGL, TGFA, P4HA2, LDHA, IRS2, TSPO, HGF, GADD45A, TGFBR3, CD38, PRF1, FAS LG, TIMP3, ANKH, LGALS3, PPP2R5B, GPX1, BNIP3L, BCL2L1, KLF7, FOSL2, ADM, MXI1, SELENBP1, PGF, FOXO3, NEDD4L, SLC2A1, SIAH2, MMP9, CXCR3, CD8A, IF NG, CD8B, JAK2, CCL4, ICAM1, WARS1, KRT1, GPR65, DYRK3, ACHE, ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4,The method according to Embodiment 198 or 199, selected from the group consisting of TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, CD24, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, DDX6, SENP5, RAPGEF1, DTX2, RELB, DYRK2, CCNB1IP1, TDRD9, ZAP70, ARL14EP, MDC1, ADGRE3, and combinations thereof.

[0472] 201. The composite biomarker score is selected from the group consisting of SeptiCyte LAB, Septicyte RAPID, Inflammatix TruVerity, and combinations thereof, according to any one of Embodiments 198 to 200.

Claims

1. A method for generating an indicator of the risk of sepsis, It must receive at least one input feature corresponding to the patient. Using a machine learning model to analyze the at least one input feature and generate a predictive score indicating the probability that the patient will have or develop sepsis within a predetermined period, based on the at least one input feature. A method comprising, wherein at least one input feature is related to sepsis.

2. The method according to claim 1, wherein the at least one input feature includes a clinical parameter.

3. The method according to claim 1 or 2, wherein the at least one input feature includes a biomarker.

4. The method according to any one of claims 1 to 3, further comprising generating a diagnostic output based on the predicted score.

5. The method according to any one of claims 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 8.

6. The method according to any one of claims 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 9.

7. The method according to any one of claims 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 10.

8. The method according to any one of claims 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 11.

9. The method according to any one of claims 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 12.

10. The method according to any one of claims 1 to 4, wherein the clinical parameters and / or biomarkers are selected from the group of input features listed in Table 13.

11. The method according to any one of claims 5 to 10, wherein the group of input features listed in any one of Tables 1 to 6 is listed in terms of their relative importance to the prediction score.

12. The method according to any one of claims 1 to 11, wherein the relative importance to the predicted score is determined based on the rank of the features, the rank of the features is determined by ranking the input features based on the average SHAP value, a high average SHAP value represents a high relative importance to the predicted score, and a low SHAP value represents a low relative importance to the predicted score.

13. The average SHAP value of the input features in the group of input features listed in any one of Tables 8 to 13 is: Determining the absolute magnitude of all SHAP values ​​for a given population with respect to the aforementioned input features, The absolute magnitudes of all SHAP values ​​are summed up to obtain the total sum of the input features, The sum of the above is divided by the number of groups of the given group for the input features. The method according to claim 12, calculated by...

14. The method according to any one of claims 1 to 13, wherein the at least one input feature comprises procalcitonin, the procalcitonin concentration values ​​are divided into two groups, and the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group.

15. The method according to claim 14, wherein the partition threshold between the first group and the second group of procalcitonin concentration values ​​is 2.5 log10 pm / mL.

16. The method according to claim 15, wherein the relative importance of procalcitonin to the prediction score in each of the first and second groups is the SHAP median of the procalcitonin concentration values ​​in each of the respective groups.

17. The method according to any one of claims 1 to 13, wherein the at least one input feature comprises procalcitonin, the procalcitonin concentration values ​​are divided into at least three groups, the procalcitonin concentration values ​​of the first group have a lower relative importance than the procalcitonin concentration values ​​of the second group, and the procalcitonin concentration values ​​of the third group have a higher relative importance than the procalcitonin concentration values ​​of both the first and second groups.

18. The method according to claim 17, wherein the first division threshold between the first group and the second group of procalcitonin concentration values ​​is 1.7 log10 pg / mL, and the second division threshold between the second group and the third group of procalcitonin concentration values ​​is 3.4 log10 pg / mL.

19. The method according to claim 18, wherein the relative importance of procalcitonin to the prediction score in each of the first, second, and third groups is the SHAP median of the procalcitonin concentration values ​​in each of the groups.

20. The aforementioned group of input features listed in Table 1 is the Glasgow Coma Scale, FiO ２ The method according to any one of claims 1 to 19, comprising procalcitonin, systolic blood pressure, and platelet count.

21. The method according to claim 20, wherein the Glasgow Coma Scale has a higher relative importance than FiO2, FiO2 has a higher relative importance than procalcitonin, procalcitonin has a higher relative importance than systolic blood pressure, and systolic blood pressure has a higher relative importance than platelet count.

22. The aforementioned group of input features listed in Table 1 is the Glasgow Coma Scale, FiO ２ The method according to any one of claims 1 to 19, comprising procalcitonin, systolic blood pressure, platelet count, bilirubin, blood urea nitrogen, respiratory rate, diastolic blood pressure, and albumin.

23. The aforementioned group of input features listed in Table 1 is the Glasgow Coma Scale, FiO ２ The method according to any one of claims 1 to 19, comprising procalcitonin, systolic blood pressure, platelet count, bilirubin, blood urea nitrogen, respiratory rate, diastolic blood pressure, albumin, creatinine, age, blood oxygen saturation, lactate, and C-reactive protein.

24. The aforementioned group of input features listed in Table 1 is the Glasgow Coma Scale, FiO ２ The method according to any one of claims 1 to 19, comprising procalcitonin, systolic blood pressure, platelet count, bilirubin, blood urea nitrogen, respiratory rate, diastolic blood pressure, albumin, creatinine, age, blood oxygen saturation, lactate, C-reactive protein, potassium, neutrophil count, temperature, total carbon dioxide, white blood cell count, lymphocyte count, sodium, heart rate, and chloride.

25. The method according to any one of claims 1 to 19, wherein the group of input features includes procalcitonin and C-reactive protein.

26. The method according to claim 25, wherein either or both procalcitonin and C-reactive protein are replaced with one or more other biomarkers listed in Table 2, according to the correlation values ​​associated with Table 3.

27. The method according to claim 26, wherein one or more other biomarkers replacing either or both procalcitonin and C-reactive protein have a correlation value smaller than XX.

28. The aforementioned prediction score is determined to fall into one of the four categories, Based on the category into which the predictive score is placed, a diagnostic output is generated that includes a low, moderate, high, or very high probability that the patient will have or develop sepsis within the predetermined period. The method according to any one of claims 1 to 27, further comprising:

29. The method according to claim 28, wherein the prediction score is in the range of 0 and 1.

30. The four categories mentioned above include low category, medium category, high category and very high category, The method according to claim 28 or 29, wherein the low category is specified as having the predicted score below a first threshold, the medium category is specified as having the predicted score between the first threshold and a second threshold, the high category is specified as having the predicted score between the second threshold and a third threshold, and the very high category is specified as having the predicted score above the third threshold.

31. The method according to claim 30, wherein the predictive score is less than the first threshold of 0.122, and the diagnostic output indicates that the patient has a low probability of having or developing sepsis within the predetermined period.

32. The method according to claim 30, wherein the predictive score is greater than or equal to the first threshold of 0.122 and less than the second threshold of 0.306, and the diagnostic output indicates that the patient has a moderate probability of having or developing sepsis within the predetermined period.

33. The method according to claim 30, wherein the predictive score is greater than or equal to the second threshold of 0.306 and less than the third threshold of 0.872, and the diagnostic output indicates that the patient has a high probability of having or developing sepsis within the predetermined period.

34. The method according to claim 30, wherein the predictive score is greater than or equal to the third threshold of 0.872, and the diagnostic output indicates that the patient has a very high probability of having or developing sepsis within the predetermined period.

35. Calibrating the predicted score using a calibration model to generate a calibrated score, To generate a risk category as a diagnostic output that takes the aforementioned calibrated score into consideration. The method according to any one of claims 1 to 34, further comprising:

36. Generating the aforementioned diagnostic output means The calibrated score is determined to fall into one of the four risk categories, Based on the category into which the calibrated score is placed, the risk category is generated which includes a low, moderate, high, or very high probability that the patient will have or develop sepsis within the predetermined period. The method according to claim 35, including the method described in claim 35.

37. The method according to claim 36, wherein the calibrated score is in the range of 0 and 1.

38. The method according to any one of claims 35 to 37, wherein the risk category may be selected from a group consisting of a low category, a medium category, a high category and a very high category, the low category is specified as having the calibrated score below a first threshold, the medium category is specified as having the calibrated score between the first threshold and a second threshold, the high category is specified as having the calibrated score between the second threshold and a third threshold, and the very high category is specified as having the calibrated score above the third threshold.

39. The method according to claim 38, wherein the calibrated score is less than the first threshold of 0.122, and the risk category indicates that the patient has a low probability of having or developing sepsis within the predetermined period.

40. The method according to claim 38, wherein the calibrated score is greater than the first threshold of 0.122 and less than the second threshold of 0.306, and the risk category indicates that the patient has a moderate probability of having or developing sepsis within the predetermined period.

41. The method according to claim 38, wherein the calibrated score is greater than the second threshold of 0.306 and less than the third threshold of 0.872, and the risk category indicates that the patient has a high probability of having or developing sepsis within the predetermined period.

42. The method according to claim 38, wherein the calibrated score is above the third threshold of 0.872, and the risk category indicates that the patient has a very high probability of having or developing sepsis within the predetermined period.

43. The method according to any one of claims 1 to 42, wherein the predetermined period is 12 hours, 18 hours, 24 hours, 36 hours, 48 ​​hours, or 72 hours or less, or longer than thereto.

44. The method according to any one of claims 1 to 43, further comprising training the at least one machine learning model using training data, wherein the training data includes a plurality of biomarkers and healthy profiles for a plurality of patients and a plurality of target diagnoses for the plurality of patients.

45. The method according to any one of claims 4 to 44, wherein the diagnostic output is an automatically derived label based on the prediction score generated by the at least one machine learning model.

46. The method according to any one of claims 4 to 45, wherein the diagnostic output is pronounced by a physician taking into consideration the predictive score generated by the at least one machine learning model.

47. The method according to any one of claims 4 to 46, wherein the analysis is performed without a value for one of the plurality of clinical parameters and / or a value for one of the one or more biomarkers in order to generate the predictive score, and the method further comprises generating the diagnostic output based on the predictive score.

48. The method according to any one of claims 1 to 47, further comprising identifying and imputing the value of one of the aforementioned input features for which a value related to the aforementioned input feature is missing.

49. One of the input features has a missing value, and the method is A decision-making module that triggers logic is used to determine whether to perform the analysis without the missing values ​​of one input feature in order to generate the prediction score. The method according to any one of claims 1 to 47, further comprising:

50. One of the input features has a missing value, and the method is The process involves imputing the missing value of one of the input features to generate an imputed value for the missing input feature, The data, including the aforementioned supplementary values, is analyzed to generate a predictive score indicating the probability that the patient will have or develop sepsis within a predetermined period. To generate the diagnostic output based on the predicted score. The method according to any one of claims 4 to 47, further comprising:

51. The method according to any one of claims 48 to 50, wherein the imputation of missing values ​​is performed via a pre-trained template trained using training data, the training data comprising input features including a plurality of clinical parameters and a plurality of biomarkers for a plurality of patients and a plurality of target diagnoses for the plurality of patients.

52. The method according to any one of claims 48 to 51, wherein the one input feature having missing values ​​is designated as a feature whose value is measured or obtained outside a specified time window for data that is acceptable for use in generating the prediction score.

53. The method according to claim 53, wherein the specified time window is the same for a first set of input features and different for at least two input features of a second set of input features.

54. The method according to any one of claims 1 to 53, wherein the input features include clinical parameters having numerical values ​​measured within a specified time window.

55. The method according to any one of claims 1 to 54, wherein the input feature is a biomarker having a measurement value related to the biomarker, which is acquired within a specified time window.

56. The method according to any one of claims 1 to 55, wherein the input feature includes a clinical parameter, and the clinical parameter includes at least one of a demographic measurement, patient evaluation, vital signs, blood test value, and chemical test value.

57. The method according to any one of claims 48 to 50, further comprising identifying and supplementing a biomarker from the plurality of biomarkers that has a missing value related to the biomarker.

58. The method according to any one of claims 48 to 57, wherein the one input feature for which a value related to the one input feature is missing is a biomarker from the plurality of biomarkers and / or a clinical parameter from the plurality of clinical parameters, and includes a biomarker and / or a clinical parameter for which a value related to the biomarker and / or the clinical parameter is missing.

59. The method according to any one of claims 1 to 58, wherein the input features include at least 20 clinical parameters and at least 2 biomarkers.

60. The method according to any one of claims 5 to 10, wherein the group of input features listed in any one of Tables 1 to 6 includes at least one biomarker feature and at least one clinical parameter feature, the values ​​of the one or more biomarker features are divided into two groups, and the values ​​of the one or more clinical parameter features are divided into two groups.

61. The method according to claim 60, wherein the biomarker features include at least one of procalcitonin, C-reactive protein, interleukin-6, pentraxin-3, interleukin-8, interleukin-1 receptor antagonist protein, and vascular cell adhesion molecule 1, and the concentration value of the first group of each biomarker feature has a relative importance lower than the concentration value of the second group of biomarker features.

62. The method according to claim 60 or 61, wherein the clinical parameters include at least one of platelets, systolic blood pressure, diastolic blood pressure, albumin, and blood oxygen saturation, and the values ​​of the first group of each clinical parameter feature have a higher relative importance than the concentration values ​​of the second group of the clinical parameter features.

63. The method according to claim 60 or 61, wherein the clinical parameters include at least one of respiratory rate, blood urea nitrogen, bilirubin, and creatinine, and the values ​​of the first group of each clinical parameter feature have a relative importance lower than the concentration values ​​of the second group of the clinical parameter feature.

64. The method according to claim 60, wherein the biomarker feature includes procalcitonin, and the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group.

65. The method according to claim 60, wherein the biomarker feature comprises a C-reactive protein, and the C-reactive protein concentration value of the first group has a lower relative importance than the C-reactive protein concentration value of the second group.

66. The method according to claim 60, wherein the biomarker feature comprises interleukin-6, and the interleukin-6 concentration value of the first group has a lower relative importance than the interleukin-6 concentration value of the second group.

67. The method according to claim 60, wherein the biomarker feature comprises pentraxin 3, and the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group.

68. The method according to claim 60, wherein the biomarker feature comprises interleukin-8, and the interleukin-8 concentration value of the first group has a lower relative importance than the interleukin-8 concentration value of the second group.

69. The method according to claim 60, wherein the biomarker feature comprises an interleukin-1 receptor antagonist protein, and the interleukin-1 receptor antagonist protein concentration value of the first group has a lower relative importance than the interleukin-1 receptor antagonist protein concentration value of the second group.

70. The method according to claim 60, wherein the biomarker feature comprises vascular cell adhesion molecule 1, and the concentration value of vascular cell adhesion molecule 1 in the first group has a lower relative importance than the concentration value of vascular cell adhesion molecule 1 in the second group.

71. The method according to claim 60, wherein the biomarker features include procalcitonin and C-reactive protein, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, and the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group.

72. The method according to claim 60, wherein the biomarker features include procalcitonin and interleukin-6, the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, and the interleukin-6 concentration value of the first group has a lower relative importance than the interleukin-6 concentration value of the second group.

73. The method according to claim 60, wherein the biomarker features include interleukin-6 and C-reactive protein, the interleukin-6 concentration value of the first group having a lower relative importance than the interleukin-6 concentration value of the second group, and the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group.

74. The method according to claim 60, wherein the biomarker features include procalcitonin, C-reactive protein, and interleukin-6, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group, and the interleukin-6 concentration value of the first group having a lower relative importance than the interleukin-6 concentration value of the second group.

75. The method according to claim 60, wherein the biomarker features include pentraxin 3 and interleukin-1 receptor antagonist protein, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, and the interleukin-1 receptor antagonist protein concentration value of the first group has a lower relative importance than the interleukin-1 receptor antagonist protein concentration value of the second group.

76. The method according to claim 60, wherein the biomarker features include pentraxin 3 and interleukin 8, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, and the interleukin 8 concentration value of the first group has a lower relative importance than the interleukin 8 concentration value of the second group.

77. The method according to claim 60, wherein the biomarker features include procalcitonin and interleukin-8, the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, and the interleukin-8 concentration value of the first group has a lower relative importance than the interleukin-8 concentration value of the second group.

78. The method according to claim 60, wherein the biomarker features include procalcitonin and pentraxin 3, the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, and the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group.

79. The method according to claim 60, wherein the biomarker features include pentraxin 3 and interleukin 6, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, and the interleukin 6 concentration value of the first group has a lower relative importance than the interleukin 6 concentration value of the second group.

80. The method according to claim 60, wherein the clinical parameter features include platelets, and the platelet count of the first group has a higher relative importance than the platelet count of the second group.

81. The method according to claim 60, wherein the clinical parameter features include systolic blood pressure, and the systolic blood pressure values ​​of the first group have a higher relative importance than the systolic blood pressure values ​​of the second group.

82. The method according to claim 60, wherein the clinical parameter features include diastolic blood pressure, and the diastolic blood pressure values ​​of the first group have a higher relative importance than the diastolic blood pressure values ​​of the second group.

83. The method according to claim 60, wherein the clinical parameter features include blood oxygen saturation, and the blood oxygen saturation values ​​of the first group have a higher relative importance than the blood oxygen saturation values ​​of the second group.

84. The method according to claim 60, wherein the clinical parameter feature includes albumin, and the blood oxygen saturation value of the first group has a higher relative importance than the albumin value of the second group.

85. The method according to claim 60, wherein the clinical parameter features include respiratory rate, and the respiratory values ​​of the first group have a lower relative importance than the respiratory values ​​of the second group.

86. The method according to claim 60, wherein the clinical parameter feature includes blood urea nitrogen, and the blood urea nitrogen value of the first group has a lower relative importance than the blood urea nitrogen value of the second group.

87. The method according to claim 60, wherein the clinical parameter feature includes bilirubin, and the bilirubin value of the first group has a lower relative importance than the bilirubin value of the second group.

88. The method according to claim 60, wherein the clinical parameter feature includes creatinine, and the creatinine value of the first group has a lower relative importance than the creatinine value of the second group.

89. The method according to claim 60, wherein the clinical parameter features include creatinine and blood urea nitrogen, the creatinine value of the first group having a lower relative importance than the creatinine value of the second group, and the blood urea nitrogen value of the first group having a lower relative importance than the blood urea nitrogen value of the second group.

90. The method according to claim 60, wherein the clinical parameter features include platelets, creatinine, and blood urea nitrogen, and the clinical parameter features include platelets, the platelet value of the first group has a higher relative importance than the platelet value of the second group, the creatinine value of the first group has a lower relative importance than the creatinine value of the second group, and the blood urea nitrogen value of the first group has a lower relative importance than the blood urea nitrogen value of the second group.

91. The method according to claim 60, wherein the clinical parameter features include respiratory rate, creatinine, and blood urea nitrogen, and the clinical parameter features include platelets, the respiratory value of the first group has a lower relative importance than the respiratory value of the second group, the creatinine value of the first group has a lower relative importance than the creatinine value of the second group, and the blood urea nitrogen value of the first group has a lower relative importance than the blood urea nitrogen value of the second group.

92. The method according to claim 60, wherein the clinical parameter features include platelets and respiratory rate, and the clinical parameter features include platelets, the platelet value of the first group has a higher relative importance than the platelet value of the second group, and the respiratory rate of the first group has a lower relative importance than the respiratory rate of the second group.

93. The method according to claim 60, wherein the clinical parameter features include platelets and systolic blood pressure, and the clinical parameter features include platelets, the platelet count of the first group has a higher relative importance than the platelet count of the second group, and the systolic blood pressure count of the first group has a higher relative importance than the systolic blood pressure count of the second group.

94. The method according to claim 60, wherein the clinical parameter features include platelets, respiratory rate, and systolic blood pressure, and the clinical parameter features include platelets, the platelet value of the first group has a higher relative importance than the platelet value of the second group, the respiratory value of the first group has a lower relative importance than the respiratory value of the second group, and the systolic blood pressure value of the first group has a higher relative importance than the systolic blood pressure value of the second group.

95. The method according to claim 60, wherein the relative importance of one or more features with respect to the prediction score in each of the first group and the second group is the SHAP median of the feature values ​​in each of the respective groups.

96. The method according to claim 95, wherein the relative importance of one or more biomarker features to the prediction score in each of the first group and the second group is the SHAP median of the biomarker feature values ​​in each of the respective groups.

97. The method according to claim 95, wherein the relative importance of one or more clinical parameter features to the prediction score in each of the first group and the second group is the SHAP median of the clinical parameter feature values ​​in each of the respective groups.

98. The method according to claim 60, wherein the relative importance of one or more features to the prediction score in each of the first group and the second group is the SHAP mean of the feature values ​​in each of the respective groups.

99. The method according to claim 98, wherein the relative importance of one or more biomarker features to the prediction score in each of the first group and the second group is the SHAP mean of the biomarker feature values ​​in each of the respective groups.

100. The method according to claim 98, wherein the relative importance of one or more clinical parameter features to the prediction score in each of the first group and the second group is the SHAP mean of the biomarker feature values ​​in each of the respective groups.

101. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of procalcitonin concentration values ​​is 2.24 log10 pg / mL.

102. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of pentraxin 3 concentration values ​​is 3.98 log10 pg / mL.

103. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of interleukin-8 concentration values ​​is 1.41 log10 pg / mL.

104. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of interleukin-1 receptor antagonist protein concentrations is 3.48 log10 pg / mL.

105. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of interleukin-6 concentration values ​​is 2.07 log10 pg / mL.

106. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of C-reactive protein concentration values ​​is 7.63 log10 pg / mL.

107. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of vascular cell adhesion molecule 1 concentration values ​​is 6.17 log10 pg / mL.

108. The method according to any one of claims 61 to 100, wherein the partition threshold between the first group and the second group of platelet concentration values ​​is 162 10^9 / L.

109. The method according to any one of claims 61 to 100, wherein the dividing threshold between the first group and the second group of systolic blood pressure values ​​is 101 mm Hg.

110. The method according to any one of claims 61 to 100, wherein the dividing threshold between the first group and the second group of respiratory values ​​is 101 breaths per minute.

111. The method according to any one of claims 61 to 100, wherein the dividing threshold for the blood urea nitrogen level between the first group and the second group is 24 mg / dL.

112. The method according to any one of claims 61 to 100, wherein the partition threshold for bilirubin levels between the first group and the second group is 1.19 mg / dL.

113. The method according to any one of claims 61 to 100, wherein the dividing threshold for creatinine levels between the first group and the second group is 1.56 mg / dL.

114. The method according to any one of claims 61 to 100, wherein the dividing threshold for diastolic blood pressure values ​​between the first group and the second group is 54.5 mm Hg.

115. The method according to any one of claims 61 to 100, wherein the dividing threshold between the first group and the second group of blood oxygen saturation values ​​is 91%.

116. The method according to any one of claims 5 to 10, wherein the group of input features listed in any one of Tables 1 to 6 includes at least one biomarker feature and at least one clinical parameter feature, the values ​​of the one or more biomarker features are divided into at least three groups, and the values ​​of the one or more clinical parameter features are divided into at least three groups.

117. The method according to claim 116, wherein the biomarker features include at least one of procalcitonin, C-reactive protein, interleukin-6, pentraxin-3, interleukin-8, interleukin-1 receptor antagonist protein, and vascular cell adhesion molecule 1, and the concentration value of the first group of each biomarker feature has a relative importance lower than the concentration value of the second group of the biomarker features, and the concentration value of the third group of the biomarker features has a relative importance higher than the concentration values ​​of both the first and second groups.

118. The method according to claim 118, wherein the clinical parameters include at least one of platelets, systolic blood pressure, diastolic blood pressure, albumin, and pulse oximetry, the value of the first group of each clinical parameter feature has a higher relative importance than the concentration value of the second group of the clinical parameter feature, and the concentration value of the third group of the clinical parameter feature has a lower relative importance than the clinical parameter values ​​of both the first and second groups.

119. The method according to claim 118, wherein the clinical parameters include at least one of respiratory rate, blood urea nitrogen, bilirubin, and creatinine, the value of each clinical parameter feature in the first group having a relative importance lower than the concentration value of the clinical parameter feature in the second group, and each clinical parameter feature in the third group having a relative importance higher than the clinical parameter feature values ​​of both the first and second groups.

120. The method according to claim 116, wherein the biomarker feature includes procalcitonin, the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, and the procalcitonin concentration value of the third group has a higher relative importance than the procalcitonin concentration values ​​of both the first and second groups.

121. The method according to claim 116, wherein the biomarker feature comprises C-reactive protein, the C-reactive protein concentration value of the first group has a lower relative importance than the C-reactive protein concentration value of the second group, and the C-reactive protein concentration value of the third group has a higher relative importance than the C-reactive protein concentration values ​​of both the first and second groups.

122. The method according to claim 116, wherein the biomarker feature comprises interleukin-6, the interleukin-6 concentration value of the first group has a lower relative importance than the interleukin-6 concentration value of the second group, and the interleukin-6 concentration value of the third group has a higher relative importance than both the interleukin-6 concentration values ​​of the first and second groups.

123. The method according to claim 116, wherein the biomarker feature comprises pentraxin 3, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, and the pentraxin 3 concentration value of the third group has a higher relative importance than both the pentraxin 3 concentration values ​​of the first and second groups.

124. The method according to claim 116, wherein the biomarker feature comprises interleukin-8, the interleukin-8 concentration value of the first group has a lower relative importance than the interleukin-8 concentration value of the second group, and the interleukin-8 concentration value of the third group has a higher relative importance than both the interleukin-8 concentration values ​​of the first and second groups.

125. The method according to claim 116, wherein the biomarker feature comprises interleukin-1 receptor antagonist protein, the interleukin-1 receptor antagonist protein concentration value of the first group has a lower relative importance than the interleukin-1 receptor antagonist protein concentration value of the second group, and the interleukin-1 receptor antagonist protein concentration value of the third group has a higher relative importance than the interleukin-1 receptor antagonist protein concentration values ​​of both the first and second groups.

126. The method according to claim 116, wherein the biomarker feature comprises vascular cell adhesion molecule 1, the concentration value of vascular cell adhesion molecule 1 in the first group has a lower relative importance than the concentration value of vascular cell adhesion molecule 1 in the second group, and the concentration value of vascular cell adhesion molecule 1 in the third group has a higher relative importance than the concentration values ​​of vascular cell adhesion molecule 1 in both the first and second groups.

127. The method according to claim 116, wherein the biomarker features include procalcitonin and C-reactive protein, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, the procalcitonin concentration value of the third group having a higher relative importance than both the procalcitonin concentration values ​​of the first and second groups, the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group, and the C-reactive protein concentration value of the third group having a higher relative importance than both the C-reactive protein concentration values ​​of the first and second groups.

128. The method according to claim 116, wherein the biomarker features include procalcitonin and interleukin-6, the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, the procalcitonin concentration value of the third group has a higher relative importance than both the procalcitonin concentration values ​​of the first and second groups, the interleukin-6 concentration value of the first group has a lower relative importance than the interleukin-6 concentration value of the second group, and the interleukin-6 concentration value of the third group has a higher relative importance than both the interleukin-6 concentration values ​​of the first and second groups.

129. The method according to claim 116, wherein the biomarker features include C-reactive protein and interleukin 6, the C-reactive protein concentration value of the first group has a lower relative importance than the C-reactive protein concentration value of the second group, the C-reactive protein concentration value of the third group has a higher relative importance than both the C-reactive protein concentration values ​​of the first and second groups, the interleukin 6 concentration value of the first group has a lower relative importance than the interleukin 6 concentration value of the second group, and the interleukin 6 concentration value of the third group has a higher relative importance than both the interleukin 6 concentration values ​​of the first and second groups.

130. The method according to claim 116, wherein the biomarker features include procalcitonin, C-reactive protein, and interleukin-6, the procalcitonin concentration value of the first group having a lower relative importance than the procalcitonin concentration value of the second group, the procalcitonin concentration value of the third group having a higher relative importance than the procalcitonin concentration values ​​of both the first and second groups, the C-reactive protein concentration value of the first group having a lower relative importance than the C-reactive protein concentration value of the second group, the C-reactive protein concentration value of the third group having a higher relative importance than the C-reactive protein concentration values ​​of both the first and second groups, the interleukin-6 concentration value of the first group having a lower relative importance than the interleukin-6 concentration value of the second group, and the interleukin-6 concentration value of the third group having a higher relative importance than the interleukin-6 concentration values ​​of both the first and second groups.

131. The method according to claim 116, wherein the biomarker features include pentraxin 3 and interleukin-1 receptor antagonist protein, the pentraxin 3 concentration value of the first group having a lower relative importance than the pentraxin 3 concentration value of the second group, the pentraxin 3 concentration value of the third group having a higher relative importance than both the pentraxin 3 concentration values ​​of the first and second groups, the interleukin-1 receptor antagonist protein concentration value of the first group having a lower relative importance than the interleukin-1 receptor antagonist protein concentration value of the second group, and the interleukin-1 receptor antagonist protein concentration value of the third group having a higher relative importance than both the interleukin-1 receptor antagonist protein concentration values ​​of the first and second groups.

132. The method according to claim 116, wherein the biomarker features include pentraxin 3 and interleukin 8, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, the pentraxin 3 concentration value of the third group has a higher relative importance than both the pentraxin 3 concentration values ​​of the first and second groups, the interleukin 8 concentration value of the first group has a lower relative importance than the interleukin 8 concentration value of the second group, and the interleukin 8 concentration value of the third group has a higher relative importance than both the interleukin 8 concentration values ​​of the first and second groups.

133. The method according to claim 116, wherein the biomarker features include procalcitonin and interleukin-8, the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, the procalcitonin concentration value of the third group has a higher relative importance than both the procalcitonin concentration values ​​of the first and second groups, the interleukin-8 concentration value of the first group has a lower relative importance than the interleukin-8 concentration value of the second group, and the interleukin-8 concentration value of the third group has a higher relative importance than both the interleukin-8 concentration values ​​of the first and second groups.

134. The method according to claim 116, wherein the biomarker features include procalcitonin and pentraxin 3, the procalcitonin concentration value of the first group has a lower relative importance than the procalcitonin concentration value of the second group, the procalcitonin concentration value of the third group has a higher relative importance than both the procalcitonin concentration values ​​of the first and second groups, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, and the pentraxin 3 concentration value of the third group has a higher relative importance than both the pentraxin 3 concentration values ​​of the first and second groups.

135. The method according to claim 116, wherein the biomarker features include pentraxin 3 and interleukin 6, the pentraxin 3 concentration value of the first group has a lower relative importance than the pentraxin 3 concentration value of the second group, the pentraxin 3 concentration value of the third group has a higher relative importance than both the pentraxin 3 concentration values ​​of the first and second groups, the interleukin 6 concentration value of the first group has a lower relative importance than the interleukin 6 concentration value of the second group, and the interleukin 6 concentration value of the third group has a higher relative importance than both the interleukin 6 concentration values ​​of the first and second groups.

136. The method according to claim 116, wherein the clinical parameter features include platelets, the platelet value of the first group has a higher relative importance than the platelet value of the second group, and the platelet value of the third group has a lower relative importance than both the platelet values ​​of the first and second groups.

137. The method according to claim 116, wherein the clinical parameter feature includes systolic blood pressure, the systolic blood pressure value of the first group has a higher relative importance than the systolic blood pressure value of the second group, and the systolic blood pressure value of the third group has a lower relative importance than the systolic blood pressure values ​​of both the first and second groups.

138. The method according to claim 116, wherein the clinical parameter feature includes diastolic blood pressure, the diastolic blood pressure value of the first group has a higher relative importance than the diastolic blood pressure value of the second group, and the diastolic blood pressure value of the third group has a lower relative importance than the diastolic blood pressure values ​​of both the first and second groups.

139. The method according to claim 116, wherein the clinical parameter features include pulse oximetry, the pulse oximetry values ​​of the first group have a higher relative importance than the pulse oximetry values ​​of the second group, and the pulse oximetry values ​​of the third group have a lower relative importance than the pulse oximetry values ​​of both the first and second groups.

140. The method according to claim 116, wherein the clinical parameter feature includes albumin, the albumin value of the first group has a higher relative importance than the albumin value of the second group, and the albumin value of the third group has a lower relative importance than both the albumin values ​​of the first and second groups.

141. The method according to claim 116, wherein the clinical parameter features include respiratory rate, the respiratory values ​​of the first group have a higher relative importance than the respiratory values ​​of the second group, and the respiratory values ​​of the third group have a lower relative importance than both the respiratory values ​​of the first and second groups.

142. The method according to claim 116, wherein the clinical parameter feature includes blood urea nitrogen, the blood urea nitrogen value of the first group has a higher relative importance than the blood urea nitrogen value of the second group, and the blood urea nitrogen value of the third group has a lower relative importance than the blood urea nitrogen values ​​of both the first and second groups.

143. The method according to claim 116, wherein the clinical parameter feature includes bilirubin, the bilirubin value of the first group has a higher relative importance than the bilirubin value of the second group, and the bilirubin value of the third group has a lower relative importance than the bilirubin values ​​of both the first and second groups.

144. The method according to claim 116, wherein the clinical parameter feature includes creatinine, the creatinine value of the first group has a higher relative importance than the creatinine value of the second group, and the creatinine value of the third group has a lower relative importance than the creatinine values ​​of both the first and second groups.

145. The method according to claim 116, wherein the clinical parameter features include creatinine and blood urea nitrogen, the creatinine value of the first group having a higher relative importance than the creatinine value of the second group, the creatinine value of the third group having a lower relative importance than both the creatinine values ​​of the first and second groups, the blood urea nitrogen value of the first group having a higher relative importance than the blood urea nitrogen value of the second group, and the blood urea nitrogen value of the third group having a lower relative importance than both the blood urea nitrogen values ​​of the first and second groups.

146. The method according to claim 116, wherein the clinical parameter features include platelets, creatinine, and blood urea nitrogen, the platelet value of the first group having a higher relative importance than the platelet value of the second group, the platelet value of the third group having a lower relative importance than both the platelet values ​​of the first and second groups, the creatinine value of the first group having a higher relative importance than the creatinine value of the second group, the creatinine value of the third group having a lower relative importance than both the creatinine values ​​of the first and second groups, the blood urea nitrogen value of the first group having a higher relative importance than the blood urea nitrogen value of the second group, and the blood urea nitrogen value of the third group having a lower relative importance than both the blood urea nitrogen values ​​of the first and second groups.

147. The method according to claim 116, wherein the clinical parameter features include respiratory rate, creatinine, and blood urea nitrogen, the respiratory value of the first group having a higher relative importance than the respiratory value of the second group, the respiratory value of the third group having a lower relative importance than both the respiratory values ​​of the first and second groups, the creatinine value of the first group having a higher relative importance than the creatinine value of the second group, the creatinine value of the third group having a lower relative importance than both the creatinine values ​​of the first and second groups, the blood urea nitrogen value of the first group having a higher relative importance than the blood urea nitrogen value of the second group, and the blood urea nitrogen value of the third group having a lower relative importance than both the blood urea nitrogen values ​​of the first and second groups.

148. The method according to claim 116, wherein the clinical parameter features include platelet count and respiratory rate, the platelet count of the first group having a higher relative importance than the platelet count of the second group, the platelet count of the third group having a lower relative importance than both the platelet counts of the first and second groups, the respiratory rate of the first group having a higher relative importance than the respiratory rate of the second group, and the respiratory rate of the third group having a lower relative importance than both the respiratory rates of the first and second groups.

149. The method according to claim 116, wherein the clinical parameter features include platelets and systolic blood pressure, the platelet count of the first group having a higher relative importance than the platelet count of the second group, the platelet count of the third group having a lower relative importance than both the platelet counts of the first and second groups, the systolic blood pressure of the first group having a higher relative importance than the systolic blood pressure of the second group, and the systolic blood pressure of the third group having a lower relative importance than both the systolic blood pressure of the first and second groups.

150. The method according to claim 116, wherein the clinical parameter features include platelets, respiratory rate, and systolic blood pressure, wherein the platelet count of the first group has a higher relative importance than the platelet count of the second group, the platelet count of the third group has a lower relative importance than both the platelet counts of the first and second groups, the respiratory count of the first group has a higher relative importance than the respiratory count of the second group, the respiratory count of the third group has a lower relative importance than both the respiratory counts of the first and second groups, the systolic blood pressure of the first group has a higher relative importance than the systolic blood pressure of the second group, and the systolic blood pressure of the third group has a lower relative importance than both the systolic blood pressure of the first and second groups.

151. The method according to claim 116, wherein the relative importance of one or more features with respect to the prediction score in each of the first, second, and third groups is the SHAP median of the feature values ​​in each of the respective groups.

152. The method according to claim 151, wherein the relative importance of one or more biomarker features to the prediction score in each of the first, second, and third groups is the SHAP median of the biomarker feature values ​​in each of the groups.

153. The method according to claim 151, wherein the relative importance of one or more clinical parameter features to the prediction score in each of the first, second, and third groups is the SHAP median of the clinical parameter feature values ​​in each of the groups.

154. The method according to claim 116, wherein the relative importance of one or more features with respect to the prediction score in each of the first, second, and third groups is the SHAP mean of the feature values ​​in each of the respective groups.

155. The method according to claim 154, wherein the relative importance of one or more biomarker features to the prediction score in each of the first, second, and third groups is the SHAP mean of the biomarker feature values ​​in each of the respective groups.

156. The method according to claim 154, wherein the relative importance of one or more clinical parameter features with respect to the prediction score in each of the first, second, and third groups is the SHAP mean of the clinical parameter feature values ​​in each of the respective groups.

157. The method according to any one of claims 116 to 156, wherein the division threshold between the first group and the second group of procalcitonin concentration values ​​is 2.05 log10 pg / mL, and the second division threshold between the second group and the third group of procalcitonin concentration values ​​is 2.85 log10 pg / mL.

158. The method according to any one of claims 116 to 156, wherein the division threshold between the first group and the second group of pentraxin 3 concentration values ​​is 3.61 log10 pg / mL, and the second division threshold between the second group and the third group of pentraxin 3 concentration values ​​is 3.98 log10 pg / mL.

159. The method according to any one of claims 116 to 156, wherein the partition threshold between the first group and the second group of interleukin-8 concentration values ​​is 1.36 log10 pg / mL, and the second partition threshold between the second group and the third group of interleukin-8 concentration values ​​is 1.97 log10 pg / mL.

160. The method according to any one of claims 116 to 156, wherein the partition threshold between the first group and the second group of interleukin-1 receptor antagonist protein concentrations is 3.15 log10 pg / mL, and the second partition threshold between the second group and the third group of interleukin-1 receptor antagonist protein concentrations is 3.57 log10 pg / mL.

161. The method according to any one of claims 116 to 156, wherein the partition threshold between the first group and the second group of interleukin-6 concentration values ​​is 3.15 log10 pg / mL, and the second partition threshold between the second group and the third group of interleukin-6 concentration values ​​is 3.57 log10 pg / mL.

162. The method according to any one of claims 116 to 156, wherein the partition threshold between the first group and the second group of C-reactive protein concentration values ​​is 7.31 log10 pg / mL, and the second partition threshold between the second group and the third group of C-reactive protein concentration values ​​is 7.91 log10 pg / mL.

163. The method according to any one of claims 116 to 156, wherein the division threshold between the first group and the second group of vascular cell adhesion molecule 1 concentration values ​​is 6.05 log10 pg / mL, and the second division threshold between the second group and the third group of vascular cell adhesion molecule 1 concentration values ​​is 6.17 log10 pg / mL.

164. The method according to any one of claims 116 to 156, wherein the division threshold between the first group and the second group of platelet concentration values ​​is 139 10^9 / L, and the second division threshold between the second group and the third group of platelet concentration values ​​is 188 10^9 / L.

165. The method according to any one of claims 116 to 156, wherein the dividing threshold between the first group and the second group of systolic blood pressure values ​​is 91 mm Hg, and the second dividing threshold between the second group and the third group of systolic blood pressure values ​​is 111 mm Hg.

166. The method according to any one of claims 116 to 156, wherein the dividing threshold between the first group and the second group of respiratory values ​​is 23.5 breaths per minute, and the second dividing threshold between the second group and the third group of respiratory values ​​is 28 breaths per minute.

167. The method according to any one of claims 116 to 156, wherein the dividing threshold for blood urea nitrogen levels between the first group and the second group is 21 mg / dL, and the second dividing threshold for blood urea nitrogen levels between the second group and the third group is 41 mg / dL.

168. The method according to any one of claims 116 to 156, wherein the division threshold for bilirubin levels between the first group and the second group is 0.88 mg / dL, and the second division threshold for bilirubin levels between the second group and the third group is 1.5 mg / dL.

169. The method according to any one of claims 116 to 156, wherein the dividing threshold for creatinine levels between the first group and the second group is 1.15 mg / dL, and the second dividing threshold for creatinine levels between the second group and the third group is 1.73 mg / dL.

170. The method according to any one of claims 116 to 156, wherein the dividing threshold between the first group and the second group of diastolic blood pressure values ​​is 52 mm Hg, and the second dividing threshold between the second group and the third group of diastolic blood pressure values ​​is 66 mm Hg.

171. The method according to any one of claims 116 to 156, wherein the dividing threshold between the first group and the second group of blood oxygen saturation values ​​is 81 mm Hg, and the second dividing threshold between the second group and the third group of blood oxygen saturation values ​​is 92.5 mm Hg.

172. A method comprising generating a diagnostic output based on the predictive score from the method of any one of claims 4 to 171, classifying the patient as having sepsis or at risk of developing sepsis within a predetermined period, the method further comprising administering a therapeutically effective dose of treatment for sepsis to the individual.

173. The method according to claim 172, wherein the treatment comprises at least one of an antibiotic or an intravenous fluid based on the generated diagnostic output.

174. A method for treating an individual having a score based on a plurality of input features indicating the probability that the patient has or will develop sepsis, based on the method according to any one of claims 1 to 171, wherein the treatment comprises administering an effective amount of one or more sepsis therapies to the individual.

175. A method comprising measuring a plurality of input features in Tables 8, 9, 10, 11, 12, or 13 from an individual and / or a sample from an individual, wherein optionally, the measurements of at least a first input feature and at least a second input feature are sequential, and the measurement of the second input feature is based on the result of the measurement of the first input feature.

176. The method according to claim 175, comprising measuring the majority of the multiple input features in Table 8 for the aforementioned individual.

177. The method according to claim 175 or 176, wherein the individual is suspected of having sepsis, suspected of developing sepsis, or at risk of developing sepsis, or the individual has one or more symptoms of sepsis.

178. The method according to any one of claims 175 to 177, wherein the individual is suspected to develop sepsis within 24 hours of the measurement of the individual.

179. The method according to any one of claims 175 to 178, wherein the sample is blood, urine, tissue, skin, saliva, sputum and / or mucus.

180. The method according to any one of claims 175 to 179, wherein the individual has an infectious disease.

181. The method according to claim 180, wherein the infectious disease is a virus, bacteria, fungus, or prion.

182. The method according to claim 180 or 181, wherein the infectious disease is SARS-CoV-2, influenza, meningitis, pneumonia, tuberculosis, or hepatitis.

183. The method according to any one of claims 175 to 182, wherein the individual is over approximately 65 years of age, less than approximately 1 year of age, is immunocompromised, has a chronic medical condition, has recently suffered from a serious illness or been hospitalized, and / or has previously suffered from sepsis.

184. The method according to claim 183, wherein the chronic medical condition is diabetes, lung disease, cancer, heart disease, hepatitis, or kidney disease.

185. A method comprising measuring one, two, three, four, five, six, seven, eight, nine, ten, one, two, one, or more input features of an individual in one or more of the tables 8, nine, ten, one, one, one, one, or one of the tables 8, nine, ten, one

186. The method according to claim 185, wherein the individual has one or more symptoms of sepsis.

187. The method according to claim 185 or 186, wherein the individual is suspected of having sepsis, suspected of developing sepsis, or at risk of developing sepsis.

188. The method according to any one of claims 185 to 187, wherein the individual is suspected to develop sepsis within 24 hours of the measurement on a sample obtained from the individual.

189. The method according to any one of claims 185 to 188, wherein the individual has an infectious disease.

190. The method according to claim 189, wherein the infectious disease is a virus, bacteria, fungus, or prion.

191. The method according to claim 189 or 190, wherein the infectious disease is SARS-CoV-2, influenza, meningitis, pneumonia, tuberculosis, or hepatitis.

192. The method according to any one of claims 185 to 191, wherein the individual is over approximately 65 years of age, under approximately 1 year of age, immunocompromised, has a chronic medical condition, has recently suffered from a serious illness or been hospitalized, or has previously suffered from sepsis.

193. The method according to claim 192, wherein the chronic medical condition is diabetes, lung disease, cancer, heart disease, or kidney disease.

194. The method according to any one of claims 1 to 171, wherein the plurality of input features include at least one input feature from a first group, at least one input feature from a second group, and at least one input feature from a third group.

195. The first group includes the input features selected from the group consisting of age, sex at birth, race, ethnicity, the patient's medical history, the patient's current illness or symptoms, neurological evaluation, clinical decision support alerts, clinician or other medical records, diagnostic codes, procedures performed on the patient, the patient's current drug therapy, interventions, patient treatment environment, medical imaging data or evaluations, electrocardiograms, endoscopic examinations, systolic blood pressure, diastolic blood pressure, body temperature, respiratory rate, heart rate, blood oxygen saturation, inhaled oxygen fraction, and combinations thereof. The second group includes the input features selected from the group consisting of white blood cell count, lymphocyte count, neutrophil count, platelet count, creatinine, blood urea nitrogen, potassium, chloride, total carbon dioxide, sodium, albumin, bilirubin, lactate, glucose, calcium, hematocrit, hemoglobin concentration, laboratory-derived biomarkers, clinical severity scale and overall score, biopsy data, microbiological test data and combinations thereof, The method according to claim 194, wherein the third group comprises the input feature selected from the group consisting of procalcitonin, C-reactive protein, protein biomarker, genomic biomarker, gene expression or transcriptome biomarker, composite biomarker score and combinations thereof.

196. The first group mentioned above is, Vital signs (e.g., temperature, heart rate, blood pressure, respiratory rate, oxygen saturation), Demographic attributes (e.g., age, sex, race, ethnicity), medical history, Current illness or symptoms, Neurological assessment (Glasgow Coma Scale, Full Summary of No Response (FOUR)), Clinical decision support alerts, Clinician or other medical records, Diagnostic code (e.g., ICD-10), Procedure (including CPT code), Current drug therapies (including NDC codes), intervention, Patient treatment environment (ICU, ED, hospital bed, outpatient treatment, etc.), Imaging methods (e.g., X-ray, CT scan, MRI, fMRI, ultrasound), Electrical recording diagrams (e.g., EEG, EKG), Endoscopic examination, and These combinations The method according to claim 194 or 195, comprising the input feature quantities selected from the group consisting of the following.

197. The second group mentioned above is, Clinical test values, Laboratory-derived biomarkers (e.g., monocyte distribution width, Cytovale IntelliSep), Clinical severity scales and overall scores (e.g., Sequential Organ Failure Assessment (SOFA) score, SOFA component score, Charlson Comorbidity Index, Acute Physiological and Chronic Health Assessment (APACHE) I and II), biopsy, Microbiological testing (culture, PCR testing, antigen testing), and These combinations The method according to any one of claims 194 to 196, comprising the input feature quantities selected from the group consisting of the following.

198. The third group mentioned above is, Protein biomarkers, Genome biomarkers, Gene expression or transcriptome biomarkers, Composite biomarker score, and These combinations The method according to any one of claims 194 to 196, comprising the input feature quantities selected from the group consisting of the following.

199. The aforementioned protein markers are angiopoietin-1, angiopoietin-2, C-reactive protein, cystatin C, D-dimer, E-selectin, fractalkine, FLT3-ligand, GCSF, GDF15, GMCSF, granzyme-B, IFN-α, IFN-γ, IL1-β, IL1-RA, IL-2, IL-4, IL-6, IL-7, IL-8, IL-10, IL-15, IP-10, and lactate dehydrogenase (LDH). The method according to claim 198, selected from the group consisting of lipopolysaccharide-binding protein (LBP), leptin, MCP1, MIP1-α, MIP1-β, MIP3-α, NGAL, pancreatic stone protein, PDL-1, pentraxin-3, procalcitonin, protein C, S100b, TGF-α, thrombomodulin, tissue factor, TNF-α, TRAIL, TREM-1, troponin, VCAM-1, VEG-F, and combinations thereof.

200. The gene expression or transcriptome biomarkers mentioned above are CEACAM4, LAMP1, PLAC8, PLA2G7, CEACAM1, ZDHHC19, C9orf95, GNA15, BATF, C3AR1, KIAA1370, TGFBI, MTCH1, RPGRIP1, HLA-DPB1, nuclear factor kappa B (NF-κB), heme oxygenase-1 (HO-1), FCGR2C, CRP, IL-6, IL-10, IL-12, TNF, LY6E, LAMP3, ISG15, USP18, RSAD2, IFI44L, NCOA7, CMPK2, I FI44, BATF2, GBP4, IFIH1, GMPR, CXCL10, IFI27, ISG15, CCL2, OAS3, P2RY14 , GCH1, CD274, ID01, ARL4A, BPGM, TNFAIP3, TNFAIP6, IL1R1, PROK2, RGS1, CX CR6, RTP4, TLR3, FZD5, KCNJ2, IRAK2, RGS16, CXCL8, OLR1, CCRL2, SPP1, ST3 GAL4, CST7, IL1R2, IL4R, GALM, CTLA4, SOCS1, BCL2L1, TGM2, SLC39A8, CA2, P NP, GATA1, SLC1A5, RGS16, CD44, PDGFC, CSF2RA, CXCL1, CSF3R, SOCS3, IL18 R1, PIM1, LEPR, IL1B, CSF2RB, ITGB3, STAT1, FAS, EFNA1, IER3, EETS2, SLC2 A3, BCL2A1, PFKFB3, CEBPB, BIRC2, ATF3, TUBB2A, G0S2, MERTK, BIK, QSOX1, PYGL, TGFA, P4HA2, LDHA, IRS2, TSPO, HGF, GADD45A, TGFBR3, CD38, PRF1, FAS LG, TIMP3, ANKH, LGALS3, PPP2R5B, GPX1, BNIP3L, BCL2L1, KLF7, FOSL2, ADM , MXI1, SELENBP1, PGF, FOXO3, NEDD4L, SLC2A1, SIAH2, MMP9, CXCR3, CD8A, IF NG, CD8B, JAK2, CCL4, ICAM1, WARS1, KRT1, GPR65, DYRK3, ACHE, ARG1, LCN2, LTF, OLFM4, HLA-DMB, YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4,The method according to claim 198 or 199, selected from the group consisting of TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, CD24, STX1A, KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, DDX6, SENP5, RAPGEF1, DTX2, RELB, DYRK2, CCNB1IP1, TDRD9, ZAP70, ARL14EP, MDC1, ADGRE3, and combinations thereof.

201. The method according to any one of claims 198 to 200, wherein the composite biomarker score is selected from the group consisting of SeptiCyte LAB, Septicyte RAPID, Inflammatix TruVerity, and combinations thereof.

Images