A system for assessing and mitigating the potential spread of infectious diseases in dialysis patients.

Abstract

A method for mitigating the spread of infectious diseases among dialysis patients is provided. The method includes: receiving individual treatment data from a medical facility via a predictive system, indicating dialysis treatment information associated with a patient undergoing dialysis; receiving individual laboratory data from a blood testing laboratory via the predictive system, indicating blood analysis information associated with the patient; determining a disease analysis result for the patient via the predictive system based on inputting the individual treatment data and the individual laboratory data into a disease prediction machine learning (ML) model, wherein the disease analysis result indicates the likelihood of the patient contracting an infectious disease; and providing instructions to the medical facility via the predictive system instructing one or more response actions based on the disease analysis result.

Description

[0001] Cross-references to related applications

[0002] This application claims priority to U.S. Provisional Patent Application Serial No. 63 / 008,626, filed April 10, 2020, entitled “SYSTEMFOR ASSESSING AND MITIGATING POTENTIAL SPREAD OF INFECTIOUS DISEASE AMONG DIALYSIS PATIENTS”, the contents of which are expressly incorporated herein by reference. Technical Field

[0003] This disclosure generally relates to systems, apparatus and methods related to healthcare. Background Technology

[0004] Patients with kidney failure or partial kidney failure typically undergo dialysis to remove toxins and excess fluid from their blood. The COVID-19 pandemic has exacerbated the difficulties faced by patients on dialysis. For example, the COVID-19 pandemic is and continues to challenge healthcare systems worldwide, adding complexity to those undergoing dialysis, particularly those with end-stage renal disease (ESKD). In the United States, dialysis patients visit dialysis clinics up to three times a week, and most ESKD patients receive outpatient hemodialysis (HD) treatment. In such circumstances, maintaining social distancing can be difficult, and enhanced infection control measures (such as temperature screening, universal mask use, and isolation treatment / teams / clinics) are necessary.

[0005] Patients with ESKD (Extra-Skimmed Kidney Disease) are typically older and have multiple comorbidities, making them at higher risk of intensive care and death if affected by COVID-19. Early reports from the United States showed a COVID-19 mortality rate of 11% in ESKD patients, higher than the 3% reported in the national population. This is not surprising, as reports from Asia and Europe show COVID-19 mortality rates of 16% to 23% in ESKD patients. Despite the high mortality rate, a compromised immune response may make dialysis patients more likely to be asymptomatic when infected with COVID-19. The most common symptoms of COVID-19 in both the general population and ESKD patients are fever (11%–66% in dialysis patients; 82% in the general population) and cough (37%–57% in dialysis patients; 62% in the general population). The lower frequency of signs and symptoms indicative of COVID-19 in dialysis patients may make COVID-19 outbreaks more challenging, especially for dialysis providers trying to prevent the spread of the disease to other dialysis patients.

[0006] Dialysis providers routinely obtain patient / clinical data (e.g., treatment data) for each patient during their dialysis treatment. In addition, dialysis patients typically have blood drawn periodically (e.g., monthly) to obtain laboratory data for each patient, monitoring their health status and assessing whether their treatment plan is effective or needs adjustment.

[0007] Therefore, reliable data collected during dialysis treatment (usually three times a week) and / or regular blood draws can provide an opportunity to detect whether a patient has COVID-19 and / or other illnesses. It is precisely because of these and other considerations that the current improvements may be useful. Summary of the Invention

[0008] The summary of this invention is provided to introduce selected concepts in a simplified form, which will be further described below. This summary is not intended to necessarily identify key or essential features of this disclosure. This disclosure may include the following aspects and embodiments.

[0009] In one exemplary embodiment, this application provides a method for mitigating the spread of infectious diseases among dialysis patients. The method includes: receiving individual treatment data from a medical facility via a prediction system, indicating dialysis treatment information associated with a patient undergoing dialysis treatment; receiving individual laboratory data from a blood testing laboratory via the prediction system, indicating blood analysis information associated with the patient; determining a disease analysis result for the patient via the prediction system based on inputting the individual treatment data and the individual laboratory data into a disease prediction machine learning (ML) model, wherein the disease analysis result indicates the likelihood of the patient contracting an infectious disease; and providing instructions to the medical facility via the prediction system instructing one or more response actions based on the disease analysis result.

[0010] In some cases, the method further includes: receiving group treatment data via a prediction system that indicates dialysis treatment information associated with multiple patients undergoing dialysis treatment; receiving group laboratory data via a prediction system that indicates blood analysis information associated with multiple patients undergoing dialysis treatment; and training a disease prediction ML model via the prediction system based on the group treatment data and the group laboratory data.

[0011] In some examples, the method further includes receiving group physician data via a prediction system that indicates clinical or treatment records associated with multiple patients undergoing dialysis treatment, wherein the disease prediction ML model is further trained based on the group physician data.

[0012] In some variations, the method further includes: receiving individual physician data from a medical facility via a prediction system that indicates clinical or treatment records associated with a patient undergoing dialysis treatment, and wherein the disease analysis results are further determined based on inputting the individual physician data into a disease prediction ML model.

[0013] In some cases, the method further includes: obtaining group patient data indicating patient demographics and history associated with multiple patients undergoing dialysis treatment via a prediction system, wherein the disease prediction ML model is further trained based on the group patient data.

[0014] In some examples, the method further includes: receiving individual patient data via a prediction system that indicates clinical or treatment records associated with a patient undergoing dialysis treatment, and wherein the disease analysis results are further determined based on inputting the individual patient data into a disease prediction ML model.

[0015] In some variations, the method further includes: obtaining geographic disease data indicating newly reported cases of infectious diseases within a geographic region associated with a patient through a prediction system, wherein the disease prediction ML model is further trained based on the geographic disease data.

[0016] In some cases, the regional disease data indicates newly reported cases at the medical facility.

[0017] In some examples, the group treatment data is associated with the medical facility. The method further includes: training a second disease prediction ML model for a second medical facility different from the stated medical facility using a prediction system; and selecting patients to use the disease prediction ML model based on individual treatment data received from the stated medical facility.

[0018] In some variations, the method further includes: receiving one or more inaccurate feedbacks from the medical facility instructing the disease prediction ML model; and retraining the disease prediction ML model based on the feedback, wherein the patient's disease analysis outcome is determined based on inputting the individual treatment data and the individual laboratory data into the retrained disease prediction machine learning (ML) model.

[0019] In some cases, the disease prediction ML model is an extreme gradient boosting (XGBoost) model or a deep learning model.

[0020] In some examples, the one or more response actions include: adjusting the patient's schedule to reassign the patient to an isolation team for future dialysis treatment; initializing a treatment plan for the patient; allocating personal protective equipment (PPE) to the medical facility; or adjusting the patient's dialysis treatment parameters.

[0021] In another exemplary embodiment, a prediction system is provided. The prediction system includes one or more processors; and a non-transitory computer-readable medium having processor-executable instructions stored thereon. The processor-executable instructions, when executed by the one or more processors, facilitate: receiving individual treatment data from a medical facility indicating dialysis treatment information associated with a patient undergoing dialysis treatment; receiving individual laboratory data from a blood testing laboratory indicating blood analysis information associated with the patient; determining a disease analysis result for the patient based on inputting the individual treatment data and the individual laboratory data into a disease prediction machine learning (ML) model, wherein the disease analysis result indicates the likelihood of the patient contracting an infectious disease; and providing instructions to the medical facility instructing one or more response actions based on the disease analysis result.

[0022] In certain circumstances, the processor-executable instructions, when executed by the one or more processors, further facilitate: receiving group treatment data indicating dialysis treatment information associated with multiple patients undergoing dialysis treatment; receiving group laboratory data indicating blood analysis information associated with multiple patients undergoing dialysis treatment; and training a disease prediction machine learning model based on the group treatment data and the group laboratory data.

[0023] In some examples, the processor-executable instructions, when executed by the one or more processors, further facilitate: receiving group physician data indicating clinical or treatment records associated with multiple patients undergoing dialysis treatment, wherein the disease prediction ML model is further trained based on the group physician data.

[0024] In some variations, the processor-executable instructions, when executed by the one or more processors, further facilitate: receiving individual physician data from a medical facility indicating clinical or treatment records associated with a patient undergoing dialysis treatment, and wherein the disease analysis results are further determined based on inputting the individual physician data into a disease prediction ML model.

[0025] In some cases, the processor-executable instructions, when executed by the one or more processors, further facilitate: obtaining group patient data indicating patient demographics and histories associated with multiple patients undergoing dialysis treatment, wherein the disease prediction ML model is further trained based on the group patient data.

[0026] In some examples, the processor-executable instructions, when executed by the one or more processors, further facilitate: receiving individual patient data indicating clinical or treatment records associated with a patient undergoing dialysis treatment, and wherein the disease analysis results are further determined based on inputting the individual patient data into a disease prediction ML model.

[0027] In some variations, the processor-executable instructions, when executed by the one or more processors, further facilitate: obtaining geographic disease data indicating newly reported cases of infectious diseases within a geographic region associated with a patient, wherein the disease prediction ML model is further trained based on the geographic disease data.

[0028] Another exemplary embodiment of this disclosure provides a non-transitory computer-readable medium having processor-executable instructions stored thereon, wherein, when executed, the processor-executable instructions facilitate: receiving individual treatment data from a medical facility indicating dialysis treatment information associated with a patient undergoing dialysis treatment; receiving individual laboratory data from a blood testing laboratory indicating blood analysis information associated with the patient; determining a disease analysis result for the patient based on inputting the individual treatment data and the individual laboratory data into a disease prediction machine learning (ML) model, wherein the disease analysis result indicates the likelihood of the patient contracting an infectious disease; and providing instructions to the medical facility instructing one or more response actions based on the disease analysis result.

[0029] Further features and aspects will now be described in more detail with reference to the accompanying drawings. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of an exemplary medical system for providing treatment data according to one or more examples of this application.

[0031] Figure 2 This is a block diagram of an exemplary disease prediction and detection environment based on one or more examples of this application.

[0032] Figure 3 This illustrates one or more examples according to this application. Figure 2 A diagram illustrating an exemplary embodiment of a system within a disease prediction and detection environment.

[0033] Figure 4 This is a block diagram illustrating an exemplary embodiment of a computing device for a prediction system according to one or more examples of this application.

[0034] Figure 5 This is based on one or more examples of the use of this application. Figure 2 A flowchart illustrating an exemplary process for predicting and detecting diseases within a disease prediction and detection environment.

[0035] Figure 6 This is another flowchart of another exemplary process for using a prediction system to predict and detect diseases, according to one or more examples of this application. Detailed Implementation

[0036] Exemplary embodiments of this application utilize information periodically obtained from dialysis patients (e.g., treatment data and / or laboratory data) to assess and mitigate the spread of infectious diseases among dialysis patients. The spread of infectious diseases is particularly concerning in dialysis clinics and hospitals where dialysis patients constantly rotate in and out. Dialysis patients may have weakened immune systems due to chronic kidney disease undergoing dialysis treatment, and infected patients may expose many other vulnerable patients to the disease. In some cases, infectious diseases or contact-transmitted diseases may include, but are not limited to, diseases spread by at least one of the following: (a) direct and / or indirect contact; (b) droplets; (c) airborne transmission; (d) public transportation.

[0037] Specifically, exemplary embodiments of this application are capable of predicting whether a dialysis patient is likely to be infected with a disease based on analysis of treatment and laboratory data already collected for these dialysis patients, thereby allowing for early preventative measures to avoid the spread of infectious diseases. In one exemplary embodiment of this application, the disease of interest is coronavirus COVID-19 (also known as SARS-CoV-2 or abbreviated as "COVID"), which can be transmitted by asymptomatic or symptomatic individuals infected with the disease. The system (e.g., a prediction system) according to this exemplary embodiment of the application predicts whether a dialysis patient is infected with COVID based on periodically obtained treatment and / or laboratory data, and provides response actions to be taken based on determining whether the corresponding dialysis patient is infected or likely to be infected. Response actions may include, for example, adjusting the patient's schedule to reassign the patient to an "isolation team" for dialysis treatment, in which the patient is isolated from other patients and / or scheduling a COVID test for the patient. Other response actions may include, for example, initiating treatment protocols for patients identified as infected or potentially infected (which may include, for example, administering medications such as antiviral drugs to patients), and / or ordering or distributing personal protective equipment (PPE) to these medical facilities based on the number of infected or potentially infected patients treated by these facilities. Another response action may include, for example, adjusting dialysis treatment parameters for patients identified as infected or potentially infected (e.g., specifying a lower ultrafiltration rate for patients identified as infected or potentially infected).

[0038] Figure 1 This is an exemplary medical system for providing treatment data according to one or more examples of this application. For example, Figure 1 The medical system shown is a hemodialysis system; however, other extracorporeal medical systems, such as other types of dialysis systems (e.g., peritoneal dialysis (PD) systems), are also considered and can be configured to provide treatment data for detecting and / or predicting whether dialysis patients have contracted COVID and / or another disease. Figure 1 The hemodialysis system and / or other types of medical systems may be used to measure, determine, acquire, and / or obtain treatment data and / or other data associated with Patient 10. Treatment data for Patient 10 may include, but is not limited to: blood pressure (standing and / or sitting), weight, body temperature, respiratory rate, pulse rate, interdialysis weight gain (IDWG), days since last treatment, hematocrit (HCT) level, hemoglobin (HGB) level, blood volume (e.g., absolute blood volume (ABV)), oxygen saturation value, online clearance rate (OLC; a measure of dialysis adequacy), mean small molecule clearance rate (KECN), and / or other data associated with Patient 10. As explained below, treatment data may be used to determine whether Patient 10 and / or other dialysis patients have contracted COVID and / or another disease.

[0039] Figure 1 A patient 10 undergoing hemodialysis treatment using a hemodialysis machine 12 is depicted. The hemodialysis system also includes an optical blood monitoring system 14. An access needle or catheter 16 is inserted into the patient 10's access site, such as in the arm, and connected to an extracorporeal tubing 18 leading to a peristaltic pump 20 and a dialyzer 22 (or blood filter). The dialyzer 22 removes toxins and excess fluid from the patient's blood. Dialyzed blood is returned from the dialyzer 22 via an extracorporeal tubing 24 and a return needle or catheter 26. In some parts of the world, extracorporeal blood flow may be additionally administered with heparin infusions to prevent clotting. Excess fluid and toxins are removed by clean dialysate fluid supplied to the dialyzer 22 via tubing 28, and waste fluid is removed via tubing 30 for disposal. In the United States, a typical hemodialysis treatment session takes approximately three to five hours. Additionally and / or alternatively, patients in intensive care units (ICUs) may also receive hemodialysis treatment and / or other dialysis / blood monitoring treatments.

[0040] The optical blood monitoring system 14 includes a display device 35 and a sensor device 34. The sensor device 34 may be, for example, a sensor clip assembly that clips onto a blood chamber 32, wherein the blood chamber 32 is disposed in an extracorporeal blood circuit. The processor (e.g., a controller) of the optical blood monitoring system 14 may be implemented in the display device 35 or the sensor clip assembly 34, or both the display device 35 and the sensor clip assembly 34 may include a corresponding processor for performing corresponding operations associated with the optical blood monitoring system.

[0041] Blood chamber 32 may be connected in series with an external tubing 18 upstream of dialyzer 22. Blood from peristaltic pump 20 flows into blood chamber 32 through tubing 18. Sensor device 34 includes an emitter that emits light of a specific wavelength and a detector for receiving the emitted light after it has passed through blood chamber 32. For example, the emitter may include an LED emitter that emits light of approximately 810 nm (isoabsorbed by red blood cells), approximately 1300 nm (isoabsorbed by water), and approximately 660 nm (sensitive to oxyhemoglobin), and the detector may include a silicon photodetector for detecting light of approximately 660 and 810 nm wavelengths, and an indium gallium arsenide photodetector for detecting light of approximately 1300 nm wavelength. Blood chamber 32 includes a lens or viewing window that allows light to pass through and allows blood to flow within it.

[0042] An example of an optical blood monitoring system having a sensor clip assembly configured to measure hematocrit and oxygen saturation of extracorporeal blood flowing through a blood chamber is described in U.S. Patent No. 9,801,993 entitled “SENSOR CLIP ASSEMBLY FOR AN OPTICAL MONITORING SYSTEM”, the entire contents of which are incorporated herein by reference.

[0043] The processor of the optical blood monitoring system 14 uses the light intensity measured by the detector to determine the HCT value of the blood flowing through the blood chamber 32. The processor uses one or more models, algorithms, and / or equations to calculate changes in HCT, HGB, oxygen saturation, and blood volume (e.g., absolute blood volume (ABV)) associated with the blood flowing through the blood chamber 32 attached to the sensor device 34. Furthermore, the processor determines additional information such as treatment data for the patient 10.

[0044] Figure 1 The hemodialysis system described in the text can be such as Figure 2 This is one of several hemodialysis systems in a dialysis clinic, such as the medical facility described. Patients can visit the dialysis clinic regularly for treatment, for example, on a Monday-Wednesday-Friday or Tuesday-Thursday-Saturday schedule.

[0045] It should be understood that Figure 1 The hemodialysis system depicted herein is merely exemplary. The principles discussed herein can be applied to other medical systems in which treatment data can be obtained.

[0046] Figure 2 This is a block diagram of an exemplary disease prediction and detection environment 200 according to one or more examples of this application. Environment 200 includes one or more medical facilities 110 (e.g., dialysis clinics or hospitals), a blood testing laboratory 120, and a prediction system 130.

[0047] One or more dialysis patients 111 receive dialysis treatment at one or more medical facilities 110 via one or more dialysis machines 112, and the one or more medical facilities 110 may further include a computing device 113 that communicates with the one or more dialysis machines 112 to obtain patient treatment data. The dialysis patient 111 may include... Figure 1 The patient 10 shown is illustrated, and the dialysis machine 112 may include and / or Figure 1 The hemodialysis machine 12 shown and described, other types of dialysis machines (e.g., PD machines and / or other types of hemodialysis machines), and / or other types of medical systems. The computing device 113 can also obtain treatment data and / or additional patient-reported data via a user interface of the computing device or via communication with one or more other computing devices at one or more medical facilities 110. Additional patient-reported data may include, for example, patient-reported symptoms such as cough, diarrhea, and fever. The computing device 113 is configured to communicate via a network to provide the obtained treatment data and / or additional reported data to the prediction system 130.

[0048] Blood testing laboratory 120 receives and analyzes patient blood samples obtained from one or more patients 111 via periodic blood draws. Based on the analysis of the blood samples, blood testing laboratory 120 generates laboratory data that is communicated via a network to prediction system 130. Laboratory data may include, but is not limited to, albumin levels, sodium levels, creatinine levels, transferrin saturation (TSAT) levels, potassium levels, phosphorus levels, ferritin levels, urea reduction rate, calcium levels, calcium (serum albumin-corrected) levels, bicarbonate levels, intact parathyroid hormone (PTH) levels, platelet count, blood urea nitrogen levels, white blood cell count, hemoglobin (HGB) count, neutrophil count, lymphocyte count, monocyte count, eosinophil count, and / or basophil count.

[0049] Additional data from other data sources 140 can also be communicated to the prediction system 130. For example, the prediction system 130 can obtain county-level and / or clinical-level data on reported COVID cases and associated mortality rates. For example, other data sources 140 can provide county-level incidence rates, such as new cases per population group over a period of time (e.g., a 4-day span and / or a 2-week span). Additionally and / or alternatively, other data sources can provide clinical-level case data, such as new cases reported over a period of time (e.g., the past 14 days and / or 28 days). In some cases, the prediction system 130 can further obtain information such as vaccine statistics and / or vaccine status (e.g., whether the patient has received a disease-associated vaccine) for patients undergoing dialysis.

[0050] The prediction system 130 ingests treatment data, reported data, laboratory data, and / or other data, for example, by storing the data in a database 132. The prediction system 130 also includes a computing system 131 connected to the database 132, which processes the data using a disease prediction model (e.g., a disease prediction dataset and / or algorithm) to generate analytical results indicating whether a corresponding patient is suspected of having an infection. In some cases, the disease prediction model is a machine learning (ML) and / or artificial intelligence (AI) dataset, model, and / or algorithm, such as a supervised ML model (e.g., an extreme gradient boosting (XGBoost) model) and / or a deep learning model.

[0051] In different exemplary embodiments, the analysis results may indicate whether a patient may be positive or negative for a disease (e.g., COVID), may be used to indicate whether a patient may be positive or negative with an associated confidence value, or may provide patients with various different classifications and / or classifiers (e.g., labels) (e.g., strongly suspected positive cases, uncertain cases, strongly suspected negative cases, etc.).

[0052] The analysis results can be written back to database 132, which may be, for example, an Oracle database. The analysis results can then be further communicated via prediction system 130 to one or more medical facilities 110, enabling the facilities to take appropriate response actions. For example, in response to determining that a particular patient has been identified as positive (or suspected positive), computing device 113 can notify a healthcare provider to order follow-up testing for the patient and / or adjust the patient's schedule by reassigning the patient to an isolation team. Notifying the healthcare provider may include pushing the analysis results to electronic medical record (EMR) software, which generates alerts to inform the healthcare provider of the risk. Ordering follow-up testing and / or adjusting patient schedules can also be implemented via EMR software, for example, by generating tasks regarding ordering testing and / or assigning patients to designated isolation teams. EMR software can also be used for other response actions, such as initializing treatment protocols, ordering or allocating PPE, and / or adjusting dialysis treatment parameters. In some cases, prediction system 130 may receive information from the EMR system executing the EMR software. In other cases, the prediction system 130 may be an EMR system that executes EMR software.

[0053] In some examples, the predictive system 130 may determine response actions to be taken (e.g., notifying a healthcare provider, ordering follow-up tests for the patient, and / or adjusting the patient's schedule) and send instructions to one or more healthcare facilities 110 and / or to other entities concerned with performing the response actions. These instructions may be sent in lieu of the analysis results or attached to the analysis results.

[0054] It should be understood that Figure 2 The environment 200 depicted herein is merely exemplary. The principles discussed herein also apply to other types of environments and / or system configurations, entities, and devices.

[0055] Figure 3This is a diagram illustrating an exemplary embodiment of a system within a disease prediction and detection environment 200 according to one or more examples of this application. For example, the prediction system 130 communicates with one or more entities within environment 200, such as a medical facility 110, a blood testing laboratory 120, a patient information (e.g., EMR) system 220, and / or a county and / or state record system 230. The prediction system 130 obtains (e.g., receives and / or retrieves) information from these entities and / or other entities (e.g., from other data sources 140). For example, the prediction system 130 may obtain treatment data and / or physician data from medical facility 110 (e.g., dialysis machine 112 and / or hemodialysis machine 12), laboratory data from blood testing laboratory 120, patient data from patient information (e.g., EMR) system 220, and / or local disease data from local and / or state systems 230.

[0056] The above describes treatment data and laboratory data. Physician data may include, but is not limited to, clinical and / or treatment records from dialysis operators (e.g., doctors, nurses, and / or technicians) who are performing and / or assisting in dialysis treatment for patients (e.g., patients 10 and / or 111). For example, using a computing device (e.g., computing device 113), the dialysis operator can provide feedback to the patient receiving dialysis treatment. Predictive system 130 can obtain this feedback from medical facility 110.

[0057] Patient data may include, but is not limited to, patient demographics and / or history. For example, patient data may include the age, sex, and body mass index (BMI) of multiple patients. Additionally and / or alternatively, patient data may include multiple patients' recent hospital and / or emergency room (ER) visits and recent discomfort, infection, and / or illness. Predictive system 130 may obtain patient demographics and / or history from patient information system 220. In some cases, patient information system 220 is an EMR system. In other cases, as described above, predictive system 130 may be an EMR system and may already possess patient data.

[0058] Geographic disease data may include, but is not limited to, information about places, counties, facilities, states, and / or countries associated with one or more diseases. For example, geographic disease data may indicate county-level incidence rates for diseases such as COVID. County-level incidence rates may be the number of new cases per population group over a period of time (e.g., a four-day or two-week time span). Additionally and / or alternatively, geographic disease data may indicate the population and / or new cases reported over a period of time (e.g., a period of fourteen or twenty-eight days) for a specific geographic area, such as a particular medical facility (e.g., Medical Facility 110), an entire county, an entire state, and / or an entire country.

[0059] Based on information obtained from medical facilities 110, blood testing laboratories 120, local and / or state systems 230, patient information systems 220, and / or other systems, the prediction system 130 can generate and / or determine a disease prediction ML model 210. For example, the prediction system 130 can use treatment data, physician data, laboratory data, patient data, and / or geographic disease data to train the disease prediction ML model 210. In other words, the prediction system 130 can divide the obtained data into training data and test data (e.g., a 60 / 40, 70 / 30, or 80 / 20 division between training and test data). The prediction system 130 can use the training data to train the disease prediction ML model and use the test data to determine the accuracy of the trained disease prediction ML model. The prediction system 130 can use specific thresholds (e.g., a 95% threshold and / or an 80% threshold) to determine the trained disease prediction ML model based on the accuracy of the trained model.

[0060] The results using 95% and 80% thresholds are described below. For example, using a 95% threshold, the trained disease prediction ML data achieved 60% accuracy in identifying predicted positive patients who were actually COVID-positive, with 2% of all positive patients identified as COVID, 99.9% of negative patients labeled as negative, 10.8 times more patients identified as COVID compared to a randomly sampled test dataset, and 0.2% of the test population labeled (e.g., test data). Using an 80% threshold, the trained disease prediction ML data achieved 36% accuracy in identifying predicted positive patients who were actually COVID-positive, with 23% of all positive patients identified as COVID, 97.6% of negative patients labeled as negative, 6.5 times more patients identified as COVID compared to a randomly sampled test dataset, and 3.5% of the test population labeled (e.g., test data).

[0061] In some variations, the prediction system 130 may use supervised learning (e.g., XGBoost and / or Light Gradient Boosting Machine (LightGBM)) to train the disease prediction ML model. In other variations, the prediction system 130 may use a deep learning model (e.g., Long Short-Term Memory (LSTM)) to train the disease prediction ML model. In still other variations, the prediction system 130 may use an unsupervised learning model to train the disease prediction ML model. Additionally and / or alternatively, the prediction system 130 may use a combination of these and / or others as an ensemble to train the disease prediction ML model.

[0062] After training the disease prediction ML model, prediction system 130 stores the trained disease prediction ML data in a database such as database 132. Additionally and / or alternatively, prediction system 130 uses the trained disease prediction ML model to determine whether a patient (e.g., patient 10 and / or 111) has contracted a specific disease such as COVID. For example, prediction system 130 obtains treatment data and / or laboratory data associated with the patient. Prediction system 130 may input the treatment data and / or laboratory data into the trained disease prediction ML model to generate outputs such as analysis results, as described above.

[0063] The prediction system 130 includes one or more computing devices (e.g., computing device 131), computing platforms, cloud computing platforms, systems, databases (e.g., database 132), servers, and / or other devices and / or uses thereof to implement the prediction system. In some variations, the prediction system 130 may be implemented as an engine, software function, and / or application. In other words, the functionality of the prediction system 130 may be implemented as software instructions stored in a storage device (e.g., memory) and executed by one or more processors.

[0064] Figure 4 This is a block diagram illustrating an exemplary embodiment of a computing device 131 of a prediction system 130 according to one or more examples of this application. The computing device 131 may include a processor 310 and a memory 320. The processor 310 may receive control signals and / or send control signals to the prediction system 130 and / or other systems and / or other devices within the environment 200. Communication between the processor 310 and other systems may be bidirectional, whereby the system may acknowledge control signals and / or provide information associated with the system and / or requested operations. Furthermore, a user input interface 315 and a display 302 may be configured to receive and / or display input from an operator. For example, the prediction system 130 may use a supervised ML model, and the operator may use the user input interface 315 and / or the display 302 to train the supervised ML model. Examples of components that may be used within the user input interface 315 include a keyboard, buttons, a microphone, a touchscreen, a gesture recognition device, a display screen, and a speaker. A power supply 325 may allow the computing device 131 to receive power and, in some variations, may be a separate power supply.

[0065] Processor 310 may be a central processing unit (CPU), controller, and / or logic device that executes computer-executable instructions for performing the functions, processes, and / or methods described herein. According to various examples, processor 310 may be a commercially available processor, such as those manufactured by Intel, AMD, Motorola, and Freescale. However, processor 310 may be any type of processor, multiprocessor, or controller, whether commercially available or specially manufactured.

[0066] Memory 320 may include a computer-readable and writable non-volatile data storage medium configured to store non-transitory instructions and data. Furthermore, memory 320 may include processor memory that stores data during operation of processor 310. In some examples, processor memory includes relatively high-performance volatile random access memory such as dynamic random access memory (DRAM), static RAM (SRAM), or synchronous DRAM. However, processor memory may include any means for storing data, such as non-volatile memory, with sufficient throughput and storage capacity to support the functions described herein. Furthermore, the examples are not limited to specific memories, memory systems, or data storage systems.

[0067] Instructions stored in memory 320 may include executable programs or other code that can be executed by processor 310. These instructions may be persistently stored as coded signals, and may cause processor 310 to perform the functions described herein. Memory 320 may include information recorded on or in a medium, and this information may be processed by processor 310 during the execution of the instructions. Databases may be stored in memory 320 and accessible by processor 310. For example, a trained disease prediction ML model may be stored in memory 320.

[0068] The computing device 131 may include a network interface 306 for communicating with other systems and devices within the environment 200 and / or prediction system 130. In some cases, the network interface 306 may include wireless capabilities for wireless communication with other systems and devices. In other cases, the network interface 306 may use direct communication to communicate with other systems and devices within the environment 200 and / or prediction system 130.

[0069] Figure 5 This is based on one or more examples of the use of this application Figure 2 A flowchart of an exemplary process 500 for predicting and detecting diseases in a disease prediction and detection environment.

[0070] In stage 501, a healthcare professional draws blood from the patient to obtain a blood sample. This may be done, for example, at a dialysis clinic (e.g., medical facility 110) and may be repeated periodically (e.g., weekly or monthly). The patient's blood sample is then sent to a blood testing laboratory (e.g., laboratory 120) for analysis. In stage 503, the blood testing laboratory analyzes the patient's blood sample to determine the patient's laboratory data. In stage 505, the blood testing laboratory provides the laboratory data to a prediction system 130 (or to a healthcare provider who inputs the laboratory data into the prediction system). For example, laboratory instruments may directly communicate reportable laboratory results to a laboratory information management system that communicates with the prediction system or healthcare provider, or laboratory technicians may input laboratory data into a computing device in the blood testing laboratory, which uploads the laboratory data to the prediction system via a communication network. The prediction system may store the laboratory data in a database along with laboratory data from other patients and historical laboratory data.

[0071] In stage 507, the medical facility obtains treatment data and / or patient-reported data from patients during dialysis treatment. Treatment data may be obtained automatically, for example, via dialysis machines (e.g., dialysis machines 12 and / or 112) and communicated to a computing device of the medical facility (e.g., computing device 113). Treatment data may also be obtained based on medical providers inputting treatment data into the computing device of the medical facility. Patient-reported data (e.g., patient data) may be provided to the computing device of the medical facility by the patient and / or by the medical provider. In stage 509, the medical facility provides the treatment data and / or patient data to the prediction system, for example, via network communication. In some cases, patients and / or medical providers may provide treatment data and / or patient-reported data directly to the prediction system via a network.

[0072] In Phase 511, the predictive system analyzes laboratory data, treatment data, reported data, and / or other data to generate analytical results. The analysis in Phase 511 may be based on the use of a disease prediction model (e.g., a disease prediction ML model). In some examples, the disease prediction model is associated with one or more diseases or infections, such as COVID. In other words, the disease prediction model can make predictions indicating whether one or more patients are considered to have COVID. For example, the output of the disease prediction model may indicate a classification (e.g., patient is positive for COVID, negative for COVID, most likely to be positive for COVID, probability of patient having COVID, etc.).

[0073] Disease prediction models can be updated as new data flows into the prediction system. Data fed into the model can be preprocessed automatically, and the model can periodically ingest data. The analysis generated by the prediction system for each patient can be in the form of an ordered list of risk scores and the relevant causes (variables) driving those scores. The analysis may also include a prediction for the patient, which may or may not include an associated confidence score. Confidence scores can be between 0 and 1. For example, a confidence value of 0.99 might indicate that the disease prediction model is 99% certain the patient has COVID. Once the prediction and any additional information related to it are generated, the information can be written to the prediction system's database, from which the analysis results can be disseminated to healthcare providers via automated processing (e.g., via an EMR system).

[0074] The dissemination of analysis results may be part of phase 513. Phase 513 includes the prediction system and / or healthcare facility performing one or more response actions based on the analysis results generated by the prediction system. As discussed above, response actions may include, for example, notifying healthcare providers, ordering follow-up testing, reassigning patients to isolation teams, initializing treatment protocols, ordering or distributing PPE, and / or adjusting dialysis treatment parameters.

[0075] In some cases, the predictive system can perform one or more response actions, such as notifying a healthcare provider (e.g., healthcare facility 110) of the analysis results (e.g., the output of a disease prediction model indicating that a patient has a specific disease). For example, the predictive system can provide instructions to display a prompt indicating the analysis results at a computing device (e.g., computing device 113). For example, the prompt could indicate that patient 111 may have COVID and provide the probability (e.g., 95%) that patient 111 has COVID.

[0076] In some examples, the predictive system and / or medical facility may execute one or more response actions to reassign patients to isolation teams based on analysis results (e.g., the output of a disease predictive model indicating that a patient has a specific disease). For example, the predictive system may provide instructions to the medical facility and / or another server with a scheduling program and / or application. These instructions may direct the scheduling program and / or application to isolate the patient for one or more dialysis treatment periods.

[0077] In some variations, the predictive system and / or medical facility may perform one or more response actions, such as customizing follow-up testing based on analysis results (e.g., the output of a disease predictive model indicating that a patient has a specific disease). For example, the predictive system may provide instructions to the medical facility and / or another system / server to instruct the medical facility / other system to perform follow-up testing due to analysis results such as indicating that a patient has COVID. In some cases, instructions to the medical facility may cause a computing device (e.g., computing device 113) to display a prompt indicating that the patient should undergo follow-up testing due to the analysis results. In other cases, the predictive system may provide instructions to scheduling programs and / or applications associated with medical facility 110 and / or another system to directly schedule one or more follow-up tests based on the analysis results.

[0078] In some examples, based on the analysis results (e.g., the output of a disease prediction model indicating that a patient has a specific disease), the prediction system and / or medical facility can perform one or more response actions, such as initializing a treatment plan and / or adjusting one or more dialysis treatment parameters. For example, the prediction system may provide instructions to the medical facility and / or another system / server instructing the medical facility / other system to perform subsequent tests based on an analysis result such as indicating that a patient has COVID. In some cases, instructions to the medical facility may cause a computing device (e.g., computing device 113) to display prompts indicating that the patient should receive a specific treatment plan and / or that the patient's dialysis treatment parameters should be adjusted due to the analysis results. In other cases, the prediction system may provide instructions to scheduling programs and / or applications associated with medical facility 110 and / or another system to schedule a specific treatment plan for the patient and / or change dialysis treatment parameters for the patient.

[0079] In some variations, the predictive system can perform one or more response actions, such as ordering more PPE based on the analysis results (e.g., the output of a disease prediction model indicating that a patient has a specific disease). For example, the predictive system can provide instructions to display a prompt at a computing device (e.g., computing device 113) instructing the healthcare facility to order more PPE based on the analysis results. In some cases, the predictive system can provide instructions to another system, such as a PPE provider, to order more PPE for the healthcare facility.

[0080] Training the COVID prediction model used in Phase 511 may include the following operations:

[0081] 1. Extract treatment data, laboratory data, and report data from databases / systems (such as the EMR database of an EMR system such as Patient Information System 220), and extract other data (such as geographic disease data) from other data sources (such as external, open data sources such as local and / or state systems 230).

[0082] 2. For example, summarize laboratory and treatment data for dialysis patients weekly. Use the most recent laboratory panel and dialysis treatment week for each corresponding patient, as well as the changes in recent results relative to previous weeks (historical data).

[0083] 3. Add other data (such as county-level population data on the number of reported COVID-19 positive cases) to the patient-level data based on the patient’s place of residence and / or the geographic area of ​​the medical facility where the patient receives dialysis treatment.

[0084] 4. A subset of the data (“training” data) is fed into a disease prediction ML model such as an XGBoost classifier, where patients who test positive for COVID are labeled as 1, and patients known or presumed to be negative are labeled as 0. Each patient is provided with a single observation consisting of the elements from steps 2-3 above. The XGBoost classifier takes these inputs and constructs numerous decision trees. Each decision tree is given random samples of training set variables and observations, and a series of thresholds are constructed to split the variables to maximize the information gained from each split. For example, the first split might be based on separating observations by temperatures above or below 98.6°F, followed by additional splits for each set of separated observations. The trees are iteratively constructed, and new trees are added to predict the errors of previous trees. Once all decision trees have been constructed, and the maximum allowed number of trees has been reached or performance no longer improves with adding more trees, this ensemble of decision trees effectively constitutes the final model. In some cases, the prediction system may assign greater significance to certain datasets compared to other available datasets (e.g., predictions may be more influenced by certain datasets and / or more decision trees may be used with certain datasets). For example, a predictive system may be more influenced by geographic disease data (e.g., the number of cases reported by a healthcare facility) and / or may use geographic disease data more frequently compared to other datasets such as treatment data. Additionally and / or alternatively, a predictive system may be more influenced by patient data (e.g., BMI) and / or may use patient data (e.g., BMI) more frequently than treatment data, but still less frequently than geographic disease data.

[0085] 5. The performance of the XGBoost classifier is validated using a separate portion of the data not used during training, such as a validation dataset. The validation dataset consists of patient-level observations fed into the model, where, for each patient, the data is passed through individual decision trees, all of which "vote" for the most likely classification, resulting in a predicted probability that the patient is in the positive category. Performance is measured by looking at multiple metrics, such as the number of correctly identified true positives (recall) and the number of actual positives predicted (precision). If the model performance does not meet acceptable targets, the model hyperparameters (e.g., the maximum number of trees or the number of splits a tree can construct) can be tuned (trying different values) to find the optimal parameters, and the model can then be retrained.

[0086] Once the XGBoost classifier has been trained and its performance validated, it is ready to be used in Phase 5.11 to generate analytical results for dialysis patients as inflows of treatment data, laboratory data, reported data, and / or other data into the prediction system. Predictions follow a similar pattern to the performance testing in Step 4, except that performance is not calculated at this stage because the ground truth is unknown. When the ground truth is indeed known, the patient's ground truth, combined with patient-level data, can be used for further training and refinement of the model.

[0087] Figure 6 This is another flowchart of another exemplary process 600 for using a prediction system to predict and detect diseases, according to one or more examples of this application. Process 600 may describe stages similar to those of process 500 described above, except that process 600 is described from the perspective of a back-end system such as prediction system 130.

[0088] During operation, at stage 602, the prediction system 130 receives treatment data from the medical facility 110 that indicates dialysis treatment information associated with a patient undergoing dialysis treatment. As described above, the treatment data may include, but is not limited to, blood pressure, weight, body temperature, respiratory rate, pulse rate, interdialysis weight gain, number of days since the last treatment, hematocrit (HCT) level, hemoglobin (HGB) level, blood volume (e.g., absolute blood volume (ABV)), oxygen saturation value, and / or other data associated with a specific patient.

[0089] At stage 604, the prediction system 130 receives laboratory data from the blood testing laboratory 120 that indicates blood analysis information associated with the patient. For example, as described above, the laboratory data may include, but is not limited to, albumin levels, sodium levels, creatinine levels, transferrin saturation (TSAT) levels, potassium levels, phosphorus levels, ferritin levels, urea reduction rate, calcium levels, calcium (corrected) levels, bicarbonate levels, intact parathyroid hormone (PTH) levels, platelet count, blood urea nitrogen levels, white blood cell count, hemoglobin count, neutrophil count, lymphocyte count, monocyte count, eosinophil count, and / or basophil count.

[0090] Additionally and / or alternatively, the prediction system 130 may receive further information associated with the patient, the medical facility 110, and / or the geographic area associated with the medical facility 110 and / or the patient. For example, the prediction system 130 may receive physician data (e.g., records about patients receiving dialysis treatment) from the medical facility 110, patient data (e.g., age, sex, BMI, recent hospitalization / ER visit, recent illness, infection, and / or other patient demographic or patient history data), and / or geographic disease data (e.g., the number of new cases or the number of new cases per population group for a specific geographic area associated with the patient and / or the medical facility 110).

[0091] In stage 606, the prediction system 130 determines the patient's disease analysis outcome based on inputting treatment data and laboratory data into a disease prediction machine learning (ML) model. For example, as described above, the prediction system 130 can use the acquired information to train and / or store the disease prediction ML model. After training the disease prediction ML model, in stage 606, the prediction system 130 can input patient-related information (e.g., the patient's laboratory data and / or the patient's treatment data) into the trained disease prediction ML model to determine the output from the trained model. The output can be a disease analysis outcome, such as indicating whether the patient may be positive or negative for a certain disease (e.g., COVID), indicating whether the patient is positive or negative with an associated confidence value, or providing various different classifications and / or classifiers for the patient.

[0092] Additionally and / or alternatively, the prediction system 130 may input additional and / or alternative information or data into the trained disease prediction ML model. For example, the prediction system 130 may input physician data, patient data, and / or geographic disease data into the trained disease prediction ML model. By inputting further data into the trained disease prediction ML model, the prediction system 130 can more accurately predict whether a patient has a specific disease such as COVID. For example, based on input treatment data and laboratory data, the prediction system 130 can determine that a patient has COVID with a 90% probability value (e.g., 0.9). Based on further input of geographic disease data and / or patient data, this probability value can be increased to 95% (e.g., 0.95). In some cases, even if certain aspects or types of data are missing (e.g., portions of patient data such as BMI), the prediction system 130 can still determine outputs such as predicting whether a patient has a specific disease.

[0093] At stage 608, the predictive system 130 provides instructions to the healthcare facility 110 to instruct one or more response actions based on the results of the disease analysis. Response actions may include, but are not limited to, ordering follow-up testing for patients, ordering or allocating more PPE, and / or adjusting patient schedules by reassigning patients to isolation teams.

[0094] In some cases, prior to stage 606, prediction system 130 may train a disease prediction ML model. For example, prediction system 130 may acquire information such as treatment data, physician data, laboratory data, patient data, geographic disease data, and / or additional information. Using the acquired information, prediction system 130 may train a disease prediction ML model that can determine whether a patient has one or more diseases such as COVID. In some examples, prediction system 130 may use treatment data and laboratory data to train the disease prediction ML model. In other examples, prediction system 130 may use treatment data, laboratory data, and further information such as physician data and / or patient data to train the disease prediction ML model. In still other examples, prediction system 130 may use all acquired information (e.g., treatment, physician, laboratory, patient, and geographic disease data) to train the disease prediction ML model.

[0095] In some examples, prediction system 130 can train multiple disease prediction ML models, such that each model is associated with a specific geographic region and / or medical facility. For example, prediction system 130 may obtain treatment data, laboratory data, physician data, geographic disease data, and / or other data associated with a specific medical facility (e.g., medical facility 110). Prediction system 130 can use data associated with (e.g., obtained from) a specific medical facility to train a disease prediction ML model for that facility. Furthermore, prediction system 130 can use data associated with another (e.g., a second) medical facility to train a second disease prediction ML model for that other medical facility. These two disease prediction ML models may have some similarities (e.g., some trees may be the same), but may also have some differences (e.g., some trees may be slightly different and / or significantly different). After training multiple disease prediction ML models, prediction system 130 and / or other systems can perform the processes 500 and / or 600 described above. Additionally, prediction system 130 can select a specific disease prediction ML model from the multiple trained disease prediction ML models for use with patients. For example, a specific medical facility 110 may be associated with a first disease prediction ML model, and the medical facility 110 may provide treatment data for a specific patient to the prediction system 130. In this case, the prediction system 130 may use the first disease prediction ML model to determine whether a specific patient is positive or negative for a disease such as COVID.

[0096] In some variations, prediction system 130 can retrain one or more disease prediction ML models. For example, prediction system 130 may receive and / or obtain information such as feedback. This feedback may instruct prediction system 130 to retrain a specific disease prediction ML model and / or indicate that a specific disease prediction ML model has been inaccurate once or multiple times. For example, geographic disease data can change over time (e.g., a specific geographic area may have 10 cases in the first month and then 2000 cases in the second month, or alternatively, different variants of the disease may be encountered in that specific geographic area after a period of time). Based on this change, a disease prediction ML model trained using data from the first month may not accurately detect whether a patient is positive or negative for COVID in the second month. Therefore, prediction system 130 may obtain feedback from healthcare facility 110 indicating that its disease prediction ML model is malfunctioning and / or producing inaccuracies. Based on this feedback, prediction system 130 can retrain the disease prediction ML model. Prediction system 130 can use more relevant (e.g., up-to-date) data to retrain the disease prediction ML model. For example, prediction system 130 can use data such as geographic disease data from the previous two weeks. After retraining the disease prediction ML model, prediction system 130 and / or other systems can perform the above processes 500 and / or 600.

[0097] In some cases, this retraining can be automated. For example, the prediction system 130 can retrain the disease prediction ML model based on a specific number of instances (e.g., 200) or a specific percentage (e.g., 10%) of cases that indicate the disease prediction ML model is inaccurate. In other cases, the prediction system 130 can retrain the disease prediction ML model based on operator feedback.

[0098] A COVID prediction model based on an exemplary implementation of the prediction system has been proven effective. Exemplary results obtained using the COVID prediction model described herein are described in more detail in U.S. Provisional Patent Application Serial No. 63 / 008,626, filed April 10, 2020, entitled “SYSTEM FOR ASSESSING AND MITIGATING POTENTIAL SPREAD OFINFECTIOUS DISEASE AMONG DIALYSIS PATIENTS,” the contents of which are expressly incorporated herein by reference.

[0099] Furthermore, it should be understood that the threshold score used to generate positive predictions for patients can be adjusted based on the balance between the model's ability to detect true positive cases and the expectation of avoiding false positives. An appendix to U.S. Provisional Patent Application Serial No. 63 / 008,626 illustrates the option to select a high true positive rate while still maintaining a low false positive rate.

[0100] In some examples, the PYTHON version can be used to build disease prediction ML models using XGBoost. The XGBoost PYTHON package uses input variables from the training model to construct multiple decision trees, giving each decision tree a random sample and establishing a series of thresholds for the splitting variables to maximize information gain. Decision trees are constructed iteratively, with new decision trees added to predict previous errors. Decision trees produced by the XGBoost ML model are able to handle missing values ​​without imputation by including the presence of missing values ​​when determining the split (e.g., splitting observations with temperatures greater than or equal to 98.0 degrees Fahrenheit with temperatures below 98.0 degrees Fahrenheit or missing temperatures). After achieving little or no further performance improvement using a validation dataset (also used for hyperparameter tuning), the ensemble of decision trees produced by the final ML model is evaluated using a test dataset.

[0101] In some cases, disease prediction ML models can be trained using multiple (e.g., 81) selected treatment / laboratory variables up to a separately defined prediction date (e.g., 3 days before the indexing date or testing date for HD patients with COVID) to predict the risk of COVID infection identified in the next 3 days or more. This can produce individual predictions at least 3 days before the symptoms that require testing. Test data can be randomly split into 60, 20, and 20 datasets for training, validation, and testing, respectively. The same number of COVID-negative patients can then be added to the training and validation datasets. A larger number of COVID-negative samples can be added to the test dataset used to evaluate the performance of the final model.

[0102] In some variations, the performance of a disease prediction ML model can be measured by the area under the receiver operating characteristic curve (AUROC) on the training, validation, and test datasets, as well as recall, precision, and lift on the test dataset. Additionally and / or alternatively, the area under the precision-recall curve (AUPRC) can also be used to evaluate the test dataset.

[0103] AUROC can measure the true positive rate and false positive rate of classification by the predictive model across probability thresholds. Recall (sensitivity) measures the true positive rate of classification by the predictive model at a specified threshold and can be calculated as follows: Recall = Number of true positives classified by the model / (Number of true positives classified by the model + Number of false negatives classified by the model).

[0104] Precision measures the number of positive predictions made by the model at a specified threshold and can be calculated as follows: Precision = Number of true positives classified by the model / (Number of true positives classified by the model + Number of false positives classified by the model). Lift measures the effectiveness of the model compared to random sampling and can be calculated as follows: Lift = Model Precision / Proportion of positives in the dataset. AUPRC measures the precision ratio across probability thresholds for the corresponding recall values. AUROC, AUPRC, recall, and precision metrics can produce scores from 0 (lowest) to 1 (highest). A cutoff threshold for classifying predictions can be selected to optimize recall, precision, and lift based on specific use cases.

[0105] An exemplary implementation of this application uses a disease prediction ML model that has been described, developed, used, tested, and successfully validated (e.g., using retrospective data and / or results), which appears to have suitable performance in identifying dialysis patients at risk of later-identified undetected COVID infection. This is described in more detail in “Machine Learning for Prediction of Hemodialysis Patients with an Undetected SARS-CoV-2 Infection” by Monaghan, Caitlin, et al. (https: / / kidney360.asnjournals.org / content / early / 2021 / 01 / 13 / KID.0003802020), which is incorporated herein by reference.

[0106] It should be understood that the various machine-implemented operations described herein can occur by one or more corresponding processors executing processor-executable instructions stored on a tangible, non-transitory computer-readable medium such as random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), and / or another electronic memory mechanism. Therefore, for example, operations performed by any of the means described herein can be performed based on instructions stored on the means and / or applications installed on the means, and via the means's software and / or hardware.

[0107] All references cited in this article, including publications, patent applications and patents, are incorporated into this article to the same extent as if each reference were individually and specifically indicated to be incorporated into this article and described in its entirety.

[0108] While the invention has been detailed and described in the accompanying drawings and foregoing description, such description should be considered illustrative or exemplary rather than restrictive. It should be understood that changes and modifications can be made by those skilled in the art within the scope of the appended claims. In particular, this application covers further embodiments having any combination of features from the various embodiments described above and below.

[0109] The terms used in the claims should be interpreted as having the broadest reasonable interpretation consistent with the foregoing description. For example, the articles “a” or “described” used when introducing an element should not be interpreted as excluding multiple elements. Similarly, references to “or” should be interpreted as inclusive, such that references to “A or B” do not exclude “A and B” unless it is clear from the context or the preceding description that only one of A and B is intentionally chosen. Furthermore, the expression “at least one of A, B, and C” should be understood as one or more of a set of elements consisting of A, B, and C, and should not be understood as requiring at least one of the listed elements A, B, and C, regardless of whether A, B, and C are related as a category or otherwise. Moreover, the expressions “A, B, and / or C” or “at least one of A, B, or C” should be interpreted as including any singular entity from the listed elements, such as A, any subset from the listed elements, such as A and B, or the entire list of elements A, B, and C.

[0110] Unless otherwise stated herein, the listing of numerical ranges herein is intended only as a shorthand method for individually referencing each individual value falling within that range, and each individual value is incorporated into the specification as if it were listed separately herein. Unless otherwise stated herein or clearly contradicted by the context, all methods described herein may be performed in any suitable order. Unless otherwise stated, the use of any and all examples or exemplary language (e.g., “such as”) provided herein is intended only to better illustrate the invention and does not constitute a limitation on the scope of the invention. No language in the specification should be construed as indicating that any unclaimed element is essential for the practice of the invention.

Claims

1. A method comprising: Individual treatment data is received from a medical facility through a predictive system, which indicates dialysis treatment information associated with a patient undergoing dialysis treatment; The system receives individual laboratory data from a blood testing laboratory through a predictive system, the individual laboratory data indicating blood analysis information associated with the patient; The prediction system determines the patient's disease analysis results by inputting the individual treatment data and the individual laboratory data into a first disease prediction machine learning model, wherein the disease analysis results indicate the patient's likelihood of contracting an infectious disease. The method further includes: Group treatment data is received through a predictive system, which indicates dialysis treatment information associated with multiple patients undergoing dialysis; and A predictive system receives group laboratory data, which indicates blood analysis information associated with multiple patients undergoing dialysis. Specifically, the first disease prediction machine learning model is trained by the prediction system based on the group's treatment data and the group's laboratory data; and Wherein, the group of treatment data is associated with the medical facility, and wherein, the method further includes: A second disease prediction machine learning model is trained using a prediction system for a second medical facility different from the stated medical facility; and Based on receiving individual treatment data from the medical facility associated with the first geographic region, the patient is selected to use the first disease prediction machine learning model instead of the second disease prediction machine learning model; and The predictive system provides instructions to medical facilities, which instruct one or more response actions based on the results of the disease analysis, including: adjusting patient schedules to reassign patients to isolation teams for future dialysis treatment.

2. The method according to claim 1, wherein, The method further includes: The system receives group physician data, which indicates clinical or treatment records associated with multiple patients undergoing dialysis. Furthermore, the first disease prediction machine learning model is trained based on the group of physician data.

3. The method according to claim 2, wherein, The method further includes: The system receives individual physician data from the medical facility through a predictive system. This individual physician data indicates clinical or treatment records associated with patients undergoing dialysis treatment. Furthermore, the disease analysis results are determined based on the input of the individual physician data into the first disease prediction machine learning model.

4. The method according to claim 1, wherein, The method further includes: Group patient data is obtained through a predictive system, which indicates patient demographics and history associated with multiple patients currently undergoing dialysis treatment. Furthermore, the first disease prediction machine learning model is trained based on the patient data of the group.

5. The method according to claim 4, wherein, The method further includes: Individual patient data is received through a predictive system, which indicates clinical or treatment records associated with patients undergoing dialysis. The determination of the disease analysis results is further based on inputting the individual patient data into the first disease prediction machine learning model.

6. The method according to claim 1, wherein, The method further includes: Geographic disease data is obtained through a predictive system, indicating newly reported cases of infectious diseases within a geographic area associated with a patient. Furthermore, the first disease prediction machine learning model is trained based on the regional disease data.

7. The method according to claim 6, wherein, The regional disease data indicates newly reported cases at the medical facilities within the first geographic region.

8. The method according to claim 1, wherein, The method further includes: Receive feedback from the medical facility, the feedback indicating one or more inaccuracies in the first disease prediction machine learning model; and The first disease prediction machine learning model is retrained based on the feedback. Specifically, the patient's disease analysis results are determined based on the input of the individual treatment data and the individual laboratory data into a retrained first disease prediction machine learning model.

9. The method according to claim 1, wherein, The first disease prediction machine learning model is an extreme gradient boosting model or a deep learning model.

10. The method according to claim 1, wherein, The one or more response actions include: initializing a treatment plan for the patient; allocating personal protective equipment to the medical facility; or adjusting the patient's dialysis treatment parameters.

11. A prediction system comprising: One or more processors; and A non-transitory computer-readable medium storing processor-executable instructions, wherein the processor-executable instructions, when executed by the one or more processors, facilitate: Receive individual treatment data from the medical facility, the individual treatment data indicating dialysis treatment information associated with a patient who is receiving dialysis treatment; Receive individual laboratory data from a blood testing laboratory, the individual laboratory data indicating blood analysis information associated with the patient; The patient's disease analysis results are determined based on the input of the individual treatment data and the individual laboratory data into a first disease prediction machine learning model, wherein the disease analysis results indicate the patient's likelihood of contracting an infectious disease; The processor-executable instructions, when executed by the one or more processors, further facilitate: Receive group treatment data, which indicates dialysis treatment information associated with multiple patients undergoing dialysis treatment; and Receive group laboratory data, which indicates blood analysis information associated with multiple patients undergoing dialysis treatment. The first disease prediction machine learning model is trained based on the group's treatment data and the group's laboratory data. Wherein, the group of treatment data is associated with the medical facility, and wherein, when the processor-executable instructions are executed by the one or more processors, they further facilitate: Train a second disease prediction machine learning model for a second medical facility different from the stated medical facility; and Based on receiving individual treatment data from the medical facility associated with the first geographic region, the patient is selected to use the first disease prediction machine learning model instead of the second disease prediction machine learning model; and Instructions are provided to the medical facility, indicating one or more response actions based on the results of the disease analysis, including: adjusting patient schedules to reassign patients to isolation teams for future dialysis treatment.

12. The prediction system according to claim 11, wherein, The processor-executable instructions, when executed by the one or more processors, further facilitate: Receive group physician data, which indicates clinical or treatment records associated with multiple patients undergoing dialysis. Furthermore, the first disease prediction machine learning model is trained based on the group of physician data.

13. The prediction system according to claim 11, wherein, The processor-executable instructions, when executed by the one or more processors, further facilitate: Receive individual physician data from the medical facility, the individual physician data indicating clinical or treatment records associated with a patient receiving dialysis treatment, and Furthermore, the disease analysis results are determined based on the input of the individual physician data into the first disease prediction machine learning model.

14. The prediction system according to claim 11, wherein, The processor-executable instructions, when executed by the one or more processors, further facilitate: Obtain group patient data, which indicates patient demographics and history associated with multiple patients currently receiving dialysis treatment. Furthermore, the first disease prediction machine learning model is trained based on the patient data of the group.

15. The prediction system according to claim 14, wherein, The processor-executable instructions, when executed by the one or more processors, further facilitate: Receive individual patient data, which indicates clinical or treatment records associated with a patient receiving dialysis treatment, and Furthermore, the disease analysis results are determined based on the input of the individual patient data into the first disease prediction machine learning model.

16. The prediction system according to claim 11, wherein, The processor-executable instructions, when executed by the one or more processors, further facilitate: Obtain geographic disease data, which indicates newly reported cases of infectious diseases within a geographic region associated with a patient. Furthermore, the first disease prediction machine learning model is trained based on the regional disease data.

17. A non-transitory computer-readable medium having processor-executable instructions stored thereon, wherein, The processor-executable instructions, when executed, facilitate: Receive individual treatment data from the medical facility, the individual treatment data indicating dialysis treatment information associated with a patient who is receiving dialysis treatment; Receive individual laboratory data from a blood testing laboratory, the individual laboratory data indicating blood analysis information associated with the patient; The patient's disease analysis results are determined based on the input of the individual treatment data and the individual laboratory data into a first disease prediction machine learning model, wherein the disease analysis results indicate the patient's likelihood of contracting an infectious disease; The processor-executable instructions, when executed, further facilitate: Receive group treatment data, which indicates dialysis treatment information associated with multiple patients undergoing dialysis treatment; and Receive group laboratory data, which indicates blood analysis information associated with multiple patients undergoing dialysis treatment. The first disease prediction machine learning model is trained based on the group's treatment data and the group's laboratory data. Wherein, the group of treatment data is associated with the medical facility, and wherein, when the processor-executable instructions are executed, they further facilitate: Train a second disease prediction machine learning model for a second medical facility different from the stated medical facility; and Based on receiving individual treatment data from the medical facility associated with the first geographic region, the patient is selected to use the first disease prediction machine learning model instead of the second disease prediction machine learning model; and Instructions are provided to the medical facility, indicating one or more response actions based on the results of the disease analysis, including: adjusting patient schedules to reassign patients to isolation teams for future dialysis treatment.

Images