A method for time trajectory typing based on peripheral blood lymphocyte counts of sepsis patients 72 hours after transfer to ICU and its application

Abstract

The present invention provides a method for time trajectory typing based on peripheral blood lymphocyte counts of sepsis patients 72 hours after they are transferred to the ICU and its application. The present invention studies and discloses for the first time the time trajectory typing of peripheral blood lymphocyte counts of sepsis patients 72 hours after they are transferred to the ICU, and establishes a prediction method for predicting the 28-day mortality rate of sepsis patients after they are transferred to the ICU on this basis. The AUC comparison results of the ROC curve of the prediction method with the peripheral blood lymphocyte count at a single time point, the single time trajectory phenotype and the mature standard method APACHE II score for predicting the 28-day mortality risk of patients indicate that the prediction method proposed in the present patent has better performance.

Description

Technical Field

[0001] The present invention relates to the field of medical detection technology, and in particular to a method for time trajectory typing based on peripheral blood lymphocyte counts of sepsis patients 72 hours after they are transferred to an ICU (intensive care unit), and an application thereof in predicting clinical prognosis. Background Art

[0002] Sepsis, as one of the most common and fatal diseases in intensive care units (ICUs), has a complex pathogenesis involving host immune response disorders and excessive inflammatory responses. In this process, changes in the number and function of lymphocytes play a crucial role.

[0003] Studies have shown that a decrease in lymphocyte count is closely associated with a poor prognosis in patients with sepsis. [1] Clinically, indicators used to evaluate the immune status and prognosis of patients with sepsis include the neutrophil-to-lymphocyte ratio (NLR), the neutrophil-to-platelet ratio (NPR), etc. [2] These indicators reflect the body's inflammatory response and immunosuppression level to a certain extent, but they have certain limitations: first, these indicators are usually based on the measurement values ​​at a single time point, ignoring the dynamic changes of the indicators during the disease process; second, simply ratioing different cell types makes it difficult to fully reflect the interaction of immune cells and the overall immune status; finally, the selection of cutoff values ​​is easily affected by the sample and measurement time point, and may lack robustness and universality.

[0004] With the development of medical information technology and the advancement of machine learning technology, it has become possible to use longitudinal data to dynamically monitor diseases and predict prognosis. Especially in the ICU scenario, immune indicator data such as lymphocyte counts of patients at different stages of the disease can be obtained, which provides an opportunity to analyze the dynamic changes in immune status during the course of the disease. However, existing studies mainly focus on the relationship between lymphocyte counts and prognosis at a single time point, or use traditional statistical methods to compare the differences in indicators at different time points. There is still a lack of research on analyzing the dynamic change pattern of immune status from the perspective of time trajectory. Therefore, it is imperative to develop a new method to more accurately predict the risk of patient death and comprehensively characterize the changes in immune status in patients with sepsis.

[0005] References:

[0006] [1] Li Lijuan, Liu Guoliang, Li Jun, Tao Xincao, Ruan Dongshi, Zhang Yunjian. Clinical value of peripheral blood lymphocyte count in assessing the prognosis of patients with sepsis[J]. Chinese Journal of Emergency Medicine, 2015, (Issue 11).

[0007] [2] Liu Dadong, Yu Zongying, Zhang Dehou, Zhang Jianguo, Zhang Yafeng, Wang Xu. The predictive value of the ratio of peripheral blood neutrophil count to lymphocyte and platelet count for 28-day mortality in patients with sepsis[J]. Chinese Journal of Critical Care Medicine, 2021, (Issue 1). Summary of the invention

[0008] Problems to be solved by the invention:

[0009] In view of the defects of the existing methods for predicting the risk of death of sepsis patients after being transferred to the ICU, the purpose of the present invention is to provide a method for time trajectory typing of peripheral blood lymphocyte counts in sepsis patients within the first 72 hours after being transferred to the ICU, and to establish a new model for evaluating the clinical prognosis of sepsis patients based on this method.

[0010] To achieve the above purpose, the specific technical solution of the present invention is as follows:

[0011] In the first aspect, a method for time trajectory typing based on peripheral blood lymphocyte counts of sepsis patients within the first 72 hours after transfer to the ICU is provided.

[0012] A retrospective cohort of 2,149 patients with sepsis treated in the ICU of the PUMCH cohort between 2010 and 2019 was selected, and 1,056 patients with sepsis were selected according to the inclusion and exclusion criteria as the derivation set. The data in the MIMIC IV v2.2 database were used to establish a validation set with patients having the same inclusion criteria as the derivation set.

[0013] LCMM was used to perform time series analysis on the collected lymphocyte count data, including data collection, LCMM modeling, determination of the optimal number of categories, and clustering result analysis. The clustering result analysis divided the patients into four different phenotypes: high-decreasing type (α type), stable-intermediate type (β type), high-increasing type (γ type), and stable-low type (δ type).

[0014] Secondly, the nomogram established based on the time trajectory classification of peripheral blood lymphocyte counts of sepsis patients 72 hours after transfer to the ICU can be used as a new model for predicting the risk of death within 28 days.

[0015] Binary logistic regression analysis determined that independent risk factors associated with death within 28 days included peripheral blood lymphocyte time trajectory phenotype within 72 hours after transfer to the ICU, patient age, APACHE II score, heart rate at transfer, peripheral blood NK cell count, and whether the patient was transferred due to lung infection.

[0016] A nomogram was made based on the independent risk factors associated with death within 28 days. The complex relationship between the probability of death within 28 days and the six independent risk factors was simply graphically represented to establish a nomogram, which can be used as a model to predict the risk of death within 28 days in patients with sepsis transferred to the ICU.

[0017] Comparison between the models established in the third and second aspects and three other common models

[0018] The ROC curve of the prediction model established according to the second aspect was compared with the ROC curves of the other three common models. The results showed that the AUC value of the ROC curve of the prediction model established according to the second aspect was higher, indicating that the prediction model established according to the second aspect is more effective than the other three common prediction models.

[0019] The other three common prediction models include: a prediction model based on the peripheral blood lymphocyte count when the patient is transferred to the ICU; a prediction model based on the time trajectory phenotype of peripheral blood lymphocytes within 72 hours after the patient is transferred to the ICU; and a prediction model based on the acute physiology and chronic health evaluation, APACHE II score when the patient is transferred to the ICU.

[0020] Compared with the prior art, the beneficial effects of this application are:

[0021] The method provided by the present invention can effectively classify the time trajectory of peripheral blood lymphocyte counts of sepsis patients in the first three days before being transferred to the ICU, and identify dynamic change patterns associated with different clinical prognoses. Compared with existing static indicators, time trajectory typing can more comprehensively and dynamically reflect the evolution of the patient's immune status over time. Combined with the results of peripheral blood lymphocyte subpopulation analysis when admitted to the ICU, it further reveals the biological basis of different trajectory categories. The present invention focuses on the dynamic change pattern of immune cells and provides a new, explainable, and immunologically significant sepsis prognosis prediction model, which is expected to guide the precise diagnosis and treatment of sepsis and improve patient prognosis.

[0022] This method has significant advantages over the currently commonly used single time point immune indicators and traditional statistical analysis methods:

[0023] 1. LCMM can identify hidden dynamic change patterns from longitudinal data. Even if the absolute count values ​​of different patients vary greatly, they can be classified into the same category based on the relative change trend, which has stronger robustness and generalization ability;

[0024] 2. The categories obtained by clustering time trajectories have clear clinical explanations. Each category corresponds to a specific pattern of immune status changes, which is convenient for guiding treatment decisions.

[0025] 3. Trajectory-based typing can capture the nonlinear changes of immune status over time. It is more comprehensive and accurate than indicators based on a single time point and can provide richer prognostic prediction information. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 , flow chart of the overall research method;

[0027] Figure 2 , case screening flow chart for the derivation set and validation set;

[0028] A is the case screening flowchart of the PUMCH cohort (derivation set),

[0029] B is the case screening flowchart of the MIMIC-IV cohort (validation set);

[0030] Figure 3 , Bayesian Information Criterion (BIC) scores for different numbers of clusters, where

[0031] A is the Bayesian Information Criterion (BIC) score of different cluster numbers of the PUMCH cohort (derivation set),

[0032] B is the Bayesian information criterion (BIC) score of different numbers of clusters in the MIMIC-IV cohort (validation set);

[0033] Figure 4 , Four Longitudinal lymphocyte count trajectories of the phenotypes during the first 72 hours after admission to the intensive care unit, including

[0034] A is the 4 phenotypes of the derivation set,

[0035] B is the 4 phenotypes of the validation set;

[0036] Figure 5 , 28-day case fatality survival curves of the four phenotypes of the PUMCH cohort;

[0037] Figure 6 , nomogram;

[0038] Figure 7 , Four Comparison of ROC curves and AUC values ​​of the models, among which,

[0039] AUC1 is the ROC curve and AUC value of the 28-day mortality prediction model proposed in this patent.

[0040] AUC2 is the ROC curve and AUC value of the prediction model established based on the peripheral blood lymphocyte count of patients when they were transferred to the ICU.

[0041] AUC3 is the ROC curve and AUC value of the prediction model established based on the time trajectory phenotype of peripheral blood lymphocytes within 72 hours after the patient was transferred to the ICU.

[0042] AUC4 is the ROC curve and AUC value of the prediction model established based on the acute physiology and chronic health assessment and APACHE II score when the patient was transferred to the ICU. DETAILED DESCRIPTION

[0043] The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention. Unless otherwise specified, the technical means used in the examples are conventional means well known to those skilled in the art.

[0044] Unless otherwise specified, the experimental methods used in the following examples are conventional methods.

[0045] All materials, reagents, etc. in the following examples, unless otherwise specified, can be obtained from commercial sources.

[0046] Example 1: Establishment of a method for time trajectory typing based on peripheral blood lymphocyte counts in sepsis patients within the first 72 hours after admission to the ICU

[0047] The overall research method flow chart of the method for predicting the 28-day mortality risk of patients with sepsis based on the method of time trajectory typing of peripheral blood lymphocyte counts in the first 72 hours after transfer to the ICU is shown in Figure 1 .

[0048] 1. Patient screening and data collection

[0049] Inclusion criteria: Patients with sepsis aged ≥18 years who were treated in the ICU for more than 48 hours and had more than 2 lymphocyte tests were selected.

[0050] Exclusion criteria: Patients with a history of immunosuppression were included to exclude factors that may interfere with lymphocyte counts.

[0051] A retrospective cohort-PUMCH cohort of 2149 patients with sepsis treated in the ICU between 2010 and 2019 was selected. According to the above inclusion and exclusion criteria, a total of 1,056 patients with sepsis were included in this study, of whom 173 died within 28 days after transfer to the ICU.

[0052] 2. Data collection and preprocessing

[0053] 2.1 Derivation Set

[0054] Patients with sepsis were enrolled consecutively, and demographic and clinical data were recorded. Peripheral blood samples were collected at 0, 24, 48, and 72 hours after admission to the ICU, and the absolute lymphocyte counts were tested to form longitudinal serial measurement data.

[0055] The demographic and clinical data of all subjects were collected at the time of enrollment, including patient age, gender, medical history, whether they were admitted to the emergency department, whether they were transferred due to lung infection, infection focus, comprehensive organ function [expressed by Sequential Organ Failure Assessment (SOFA)], disease severity [expressed by Acute Physiology and Chronic Health Evaluation II (APACHE) II)], basic vital signs (including body temperature, heart rate, mean arterial blood pressure, respiratory rate, blood lactate, superior vena cava oxygen saturation, arterial-venous carbon dioxide partial pressure difference, and the number of vasoactive drugs used), and various organ functions (including left ventricular ejection fraction, oxygenation index, blood creatinine, blood urea nitrogen, alanine aminotransferase, total bilirubin, direct bilirubin, and blood amylase); all subjects' outcome-related indicators were collected, including ICU stay, hospital stay, ventilator use time, and 28-day mortality; peripheral blood was collected from all subjects to analyze their immune and inflammatory status, including interleukin-6, interleukin-8, interleukin-10, tumor necrosis factor-α, neutrophil count, lymphocyte count, neutrophil-to-lymphocyte count ratio, high-sensitivity C-reactive protein, and procalcitonin; peripheral blood was collected to analyze immunoglobulin A, immunoglobulin G, immunoglobulin M, complement 3, complement 4, and lymphocyte subsets [including CD19 + Total B lymphocyte count, CD3 + Total T lymphocyte count, CD4 + T lymphocytes, CD8 + T lymphocytes and CD3 - CD16 + CD56 + Natural killer (NK) cells].

[0056] Blood samples, including complete blood count, C-reactive protein (CRP), procalcitonin, and blood gas analysis, were collected upon ICU admission and during routine examinations. Immunological parameters were measured in the laboratory of Peking Union Medical College Hospital. Peripheral blood mononuclear cells were stained with fluorescent monoclonal antibodies and then analyzed by flow cytometry using a three-color EPICS-XL flow cytometer (Beckman Coulter, Brea, CA) to detect T cells (CD3 + )、CD4 + and CD8 + T cell subsets, B cells (CD19 + ) and natural killer (NK) cells (CD3 - CD16 + CD56 +). Rate nephelometry (Array 360; Beckman Coulter) was used to measure the levels of immunoglobulin (Ig) A, IgG, IgM, and complement factors C3 and C4 in serum.

[0057] Figure 2 A is the case screening flowchart of the PUMCH cohort (derivation set).

[0058] 2.2 Validation Set

[0059] For external validation, we used data from the MIMIC IV v2.2 database, a publicly available database. MIMIC IV contains de-identified health data related to more than 200,000 patients admitted to the intensive care unit (ICU) at Beth Israel Deaconi Medical Center in Boston, Massachusetts, between 2008 and 2019. The database includes data on vital signs, medications, diagnostic codes, laboratory values, imaging reports, clinical notes, and mortality. For the validation cohort, MIMIC IV was queried using structured query language to identify patients who met the same inclusion criteria as the derivation set cohort.

[0060] MIMIC IV v2.2 database reference: Johnson AEW, Pollard TJ, Shen L, et al. MIMIC-IV (version 2.2). PhysioNet. 2023. https: / / doi.org / 10.13026 / 6mm1-ek67

[0061] Figure 2 B is the case screening flowchart of the MIMIC-IV cohort (validation set).

[0062] 3. Time trajectory classification

[0063] 3.1LCMM model

[0064] Latent Class Mixed Models (LCMM) in Machine Learning LCMM is a statistical model that allows researchers to group individuals based on the trajectories of their longitudinal data. The model assumes that there are unobserved (latent) heterogeneous subgroups that have different developmental trajectories. LCMM combines the characteristics of traditional mixed effects models and latent class models. It captures the heterogeneity between individuals by introducing latent classes in the mixed effects model. Its mathematical model form is as follows:

[0065] Y it =β0+β1X it +β2X it2 +b 0i +b 1i X it +∈it

[0066] Among them, Y it represents the observed value of individual i at time t,

[0067] X it is the time variable,

[0068] β0, β1, β2 are fixed effect parameters,

[0069] b 0i , b 1i is the random effect parameter,

[0070] ∈it is the residual term.

[0071] This method has significant advantages over the currently commonly used single time point immune indicators and traditional statistical analysis methods:

[0072] (1) LCMM can identify hidden dynamic change patterns from longitudinal data. Even if the absolute count values ​​of different patients vary greatly, they can be classified into the same category based on their relative change trends, which has stronger robustness and generalization ability.

[0073] (2) The categories obtained by clustering time trajectories have clear clinical interpretations. Each category corresponds to a specific pattern of immune status changes, which is convenient for guiding treatment decisions.

[0074] (3) Trajectory-based typing can capture the nonlinear changes of immune status over time, which is more comprehensive and accurate than indicators based on a single time point and can provide richer prognostic prediction information. In addition, as a semi-parametric model, LCMM has statistical advantages in parameter estimation and category inference, and can better balance model complexity and goodness of fit.

[0075] 3.2 Use LCMM to perform time series analysis on the collected lymphocyte count data

[0076] To classify the trend of lymphocyte count changes in patients, the steps are as follows:

[0077] (1) Data collection

[0078] Lymphocyte counts were collected from patients and combined with time points (0, 24, 48, and 72 hours after ICU admission) to form a longitudinal dataset.

[0079] (2) LCMM modeling

[0080] The latent class mixed model of peripheral blood lymphocyte count was constructed using the R language lcmm package (version: 2.1.0, https: / / cran.r-project.org / web / packages / lcmm / index.html), with patients as random effects, time (wave) and its quadratic term (wave2) as fixed effects and mixed effects. This was achieved using the following R language code (if the number of latent classes is specified as 4):

[0081] lcmm::lcmm(LymLevel~wave+I(wave^2),mixture=~wave,random=~wave,subject='icuaId',ng=4,data=datlong,B=m1,link="3-equi-splines")

[0082] Among them, LymLevel is the dependent variable lymphocyte count, wave is the time variable (0-3 corresponds to 0, 24, 48 and 72 h after transfer to the ICU, respectively), mixture and random specify the relationship between latent classes and individual random effects and time, ng is the specified number of latent classes, and link is the link function.

[0083] (3) Determination of the optimal number of categories

[0084] Set ng in the code to different parameters from 1 to 9, perform typing analysis in turn, and obtain the category of each individual through the BIC clustering curve. The results are shown in Figure 3 , ,For the PUMCH cohort and the MIMIC cohort, the lowest Bayesian Information Criterion (BIC) scores for different ,numbers of clusters are both 4, and the optimal number of ,classes is finally determined to be 4.

[0085] Figure 3 The x-axis represents the number of clusters, and the y-axis represents the Bayesian Information Criterion score.

[0086] (4) Clustering results analysis

[0087] According to the changing trend, the characteristics of lymphocyte change patterns in each category were analyzed, and the patients were divided into four different phenotypes: high-decreasing type (α type), stable-intermediate type (β type), high-increasing type (γ type) and stable-low type (δ type). The results are shown in Figure 4In both the PUMCH cohort and the MIMIC cohort, the α-type had the highest lymphocyte count at 0 hour, and the lymphocyte count dropped sharply from 0 to 72 hours; the β-type had a medium lymphocyte count at 0 hour, and the lymphocyte count remained basically stable from 0 to 72 hours; the γ-type had a high lymphocyte count at 0 hour, and the lymphocyte count increased sharply from 0 to 72 hours; the δ-type had the lowest lymphocyte count at 0 hour, and the lymphocyte count remained basically stable from 0 to 72 hours.

[0088] Figure 4 In the figure, the x-axis represents the time point of lymphocyte measurement (0, 24, 48, and 72 hours after ICU admission), and the y-axis represents the lymphocyte count (x10 9 / L); the shaded area around the smooth curve indicates the confidence interval, and the confidence level is 0.95; Abbreviations: MIMIC, Medical Information Mart for Intensive Care, PUMCH, Peking Union Medical College Hospital, ICU, Intensive Care Unit, Department of Critical Care Medicine.

[0089] After analysis, in the external validation set (MIMIC cohort), the LCMM model also optimized the lymphocyte count trajectory into 4 similar time trajectory phenotype results, which was consistent with the derivation set (PUMCH cohort).

[0090] Example 2: Establishing a method for predicting the 28-day mortality risk of sepsis patients based on the time trajectory typing of peripheral blood lymphocyte counts 72 hours after transfer to the ICU

[0091] 1. Independent risk factors for death within 28 days

[0092] 1.1 Multivariate logistic regression analysis of 28-day mortality in patients with sepsis

[0093] Table 1. Multivariate logistic regression analysis of 28-day mortality in patients with sepsis

[0094]

[0095]

[0096] 1.2 Independent risk factors for death within 28 days

[0097] Multivariate logistic regression analysis showed that independent risk factors associated with death within 28 days included: peripheral blood lymphocyte time trajectory phenotype within 72 hours after transfer to the ICU, patient age, APACHEII score, heart rate at ICU transfer, peripheral blood NK cell count, and whether the patient was transferred due to lung infection. The OR values ​​and P values ​​of each independent risk factor are shown in Table 2.

[0098] Table 2. Independent risk factors associated with death within 28 days

[0099]

[0100]

[0101] ★The OR value of the time trajectory phenotype of peripheral blood lymphocytes within 72 hours after transfer to the ICU was based on the δ phenotype

[0102] 2. Analysis of peripheral blood lymphocyte subsets

[0103] Peripheral blood lymphocyte subsets including CD19 were completed within 24 hours after the patient was admitted to the ICU. + Total B lymphocyte count, CD3 + Total T lymphocyte count, CD4 + T lymphocytes, CD8 + T lymphocytes and CD3 - CD16 + CD56 + Natural killer (NK) cells.

[0104] Comparison of survival curves of the four phenotypes

[0105] Survival curves are constructed using the Kaplan-Meier method, a nonparametric statistical technique used to estimate the probability of survival over time. In this procedure, the probability of survival at each time point is calculated by multiplying the probability of survival at the previous time point by the probability of not having the event at that time point. If multiple events occur at a time point, this will result in a decrease in the probability of survival. The log-rank test is then used to compare the survival curves between different groups to assess if there are statistically significant differences. From these analyses, we can obtain a graphical representation, the survival curve, which shows the proportion of a patient population that is alive over time.

[0106] The results are as follows Figure 5Figure 2. Comparison of the survival curves of the four phenotypes during 28 days of hospitalization in the intensive care unit. Abbreviations: MIMIC, Medical Information Mart for Intensive Care; PUMCH, Peking Union Medical College Hospital; ICU, Intensive Care Unit.

[0107] Figure 5 In the figure, the x-axis is the survival time (months) and the y-axis is the survival probability.

[0108] 4. Establishment of Nomogram

[0109] A nomogram was made based on the independent risk factors associated with death within 28 days, which simply visualized the complex relationship between the probability of death within 28 days and the six independent risk factors.

[0110] The first line is a score scale, ranging from 0 to 100;

[0111] The second row is the patient's age, with 10 years old corresponding to 0 points and 110 years old corresponding to 42 points, and the intervals in between are evenly divided;

[0112] The third row is the APACHE II score, where an APACHE II score of 0 corresponds to a score scale of 0, and an APACHE II score of 55 corresponds to a score scale of 98, and the intervals in between are evenly divided;

[0113] The fourth row is the heart rate upon admission to the ICU, with 50 beats corresponding to 0 points, 180 beats corresponding to 45 points, and the intervals in between divided equally;

[0114] The fifth row is the peripheral blood NK cell count, 800 cells / μL is 0 points, 0 cells / μL is 100 points, and the intervals in between are divided equally;

[0115] The sixth line is whether the patient was transferred due to lung infection, with 0 points for no and 33 points for yes;

[0116] The seventh line is the time series classification of peripheral blood lymphocyte count, where β type is 0 points, γ type is 2 points, α type is 13 points, and δ type is 22 points;

[0117] The eighth line predicts the 28-day mortality rate with a probability range of 0.1 to 0.8.

[0118] In the nomogram, the second to seventh rows are independent risk factors for death within 28 days. Different factor states correspond to different scale scores. The eighth row is the sum of the scores of each factor. The ninth row is the predicted probability of death within 28 days. The sum of the different scores of the second to seventh rows is projected to the corresponding position of the eighth row, which is the predicted probability of death within 28 days. The nomogram is as follows: Figure 6 shown.

[0119] Use of nomogram:

[0120] When receiving a sepsis patient transferred to the ICU in clinical work, it is necessary to predict the risk of death within 28 days. The patient's age, APACHEII score, heart rate, peripheral blood NK cell count, whether the patient was transferred due to lung infection, peripheral blood lymphocyte count time series typing and other clinical information can be collected. Then, the corresponding score is obtained according to the interval of each risk factor of the patient. After summarizing the scores, the total score of the patient at this stage is obtained. In the column of death probability, the specific probability value of the patient's current death within 28 days is obtained. A total score of 210 points corresponds to a predicted probability of death within 28 days of 0.22, a total score of 250 points corresponds to a predicted probability of death within 28 days of 0.50, and a total score of 280 points corresponds to a predicted probability of death within 28 days of 0.75.

[0121] For example, suppose an 80-year-old sepsis patient is transferred to the ICU due to lung infection. His heart rate is 110 beats / min, the peripheral blood lymphocyte time series typing is α type, the peripheral blood NK cell count is 180 / μL, and the APACHEII score is 25 when he is transferred to the ICU. According to the prediction model, the total score of the patient is 215 points, and the probability of death within 28 days is about 25%. The actual prognosis of the patient is: he is discharged after 35 days of hospitalization. The prediction results are basically consistent with the actual prognosis, which preliminarily confirms the reliability of the nomogram in individualized risk prediction.

[0122] Conclusion: The nomogram is a specific, visual and simple method to calculate and predict the risk of 28-day mortality in patients with sepsis admitted to the ICU based on some clinical characteristics.

[0123] 5. ROC curve of the method for predicting the risk of death within 28 days of sepsis patients based on the time trajectory typing of peripheral blood lymphocyte counts 72 hours after transfer to the ICU

[0124] The process of establishing the ROC curve involves generating a prediction score for each patient based on different prediction models, which represents the probability of the patient's death within 28 days. Next, using these prediction scores and the patient's actual survival status, the predictive performance of the model is evaluated by drawing the ROC curve. On the ROC curve, each point represents a specific prediction score threshold and corresponds to the true positive rate and false positive rate under that threshold. Finally, the area under the ROC curve (AUC) is calculated to quantify the predictive accuracy of the model, where the closer the AUC value is to 1, the stronger the predictive ability of the model. In this way, the performance of different models in predicting 28-day mortality can be compared and evaluated.

[0125] The results are as follows Figure 7 As shown in AUC1, AUC1 = 0.754 (0.716, 0.793).

[0126] Example 3. ROC curve of a method for predicting the risk of death of sepsis patients 28 days after transfer to the ICU based on peripheral blood lymphocyte count at the time of admission to the ICU (Comparative Example 1)

[0127] The results are as follows Figure 7 As shown in AUC2, AUC2 = 0.564 (0.515, 0.613).

[0128] Example 4. ROC curve of a method for predicting the risk of death 28 days after transfer to the ICU of sepsis patients using the 72-hour peripheral blood lymphocyte time trajectory phenotype (Comparative Example 2)

[0129] The results are as follows Figure 7 As shown in AUC3, AUC3 = 0.680 (0.634, 0.727).

[0130] Example 5. ROC curve of a method for predicting the risk of death of sepsis patients 28 days after transfer to the ICU using the admission APACHE II score (Comparative Example 3)

[0131] The results are as follows Figure 7 As shown in AUC4, AUC4 = 0.623 (0.578, 0.668).

[0132] In summary, the AUC comparison results of the ROC curve of the prediction method proposed in this patent and the single time point peripheral blood lymphocyte count, single time trajectory phenotype and mature standard method APACHE II score for predicting the risk of death of sepsis patients after transfer to the ICU suggest that the prediction method proposed in this patent is more effective.

[0133] Although the present invention has been described in detail above with general descriptions and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements may be made thereto based on the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all fall within the scope of protection claimed by the present invention.

Claims

1. A method for predicting the risk of death within 28 days of a sepsis patient based on the temporal trajectory typing of peripheral blood lymphocyte counts 72 hours after transfer to an ICU, the method comprising: (1) Create a nomogram based on independent risk factors for death within 28 days The independent risk factors were: peripheral blood lymphocyte count time trajectory typing within 72 hours after transfer to ICU, patient age, APACHEII score, heart rate at transfer to ICU, peripheral blood NK cell count, and whether the patient was transferred due to lung infection; The independent risk factors mentioned were determined by multivariate logistic regression analysis of 28-day mortality in patients with sepsis; The time trajectory typing of peripheral blood lymphocyte count within 72 hours after transfer to the ICU is determined by time series analysis of the collected lymphocyte count data using the LCMM model, and is divided into four different typings: α-type, i.e., high-declining type, β-type, i.e., stable-intermediate type, γ-type, i.e., high-increasing type, and δ-type, i.e., stable-low type; The α-type has the highest lymphocyte count at 0 hours, and the lymphocyte count drops sharply from 0 hours to 72 hours; the β-type has a medium lymphocyte count at 0 hours, and the lymphocyte count remains basically stable from 0 hours to 72 hours; the γ-type has a high lymphocyte count at 0 hours, and the lymphocyte count rises sharply from 0 hours to 72 hours; the δ-type has the lowest lymphocyte count at 0 hours, and the lymphocyte count remains basically stable from 0 hours to 72 hours; (2) Use the nomogram to predict the risk of death within 28 days after transfer to the ICU for patients with sepsis Clinical information such as the patient's age, APACHE II score, heart rate, peripheral blood NK cell count, whether the patient was transferred due to lung infection, and peripheral blood lymphocyte count time trajectory typing was collected. Then, the corresponding score was obtained according to the interval of each independent risk factor of the patient. The scores were summarized to obtain the patient's total score at this stage, and the specific probability value of the patient's death within 28 days was obtained in the death probability column.

2. The method according to claim 1, characterized in that The method for constructing the time trajectory typing of peripheral blood lymphocyte counts of sepsis patients 72 hours after being transferred to the ICU includes: (1) Data collection: Lymphocyte counts were collected at 0, 24, 48, and 72 hours after septic patients were transferred to the ICU and combined with time to form a longitudinal data set; (2) LCMM modeling; (3) Determine the optimal model category number 4 through the BIC clustering curve; (4) Clustering result analysis.

3. The method according to claim 1, characterized in that The nomogram is constructed by using the independent risk factors related to death within 28 days to create a nomogram, which simply visualizes the complex relationship between the probability of death within 28 days and the six independent risk factors. The nomogram includes: The first line is a score scale, ranging from 0 to 100; The second row is the patient's age, with 10 years old corresponding to 0 points and 110 years old corresponding to 42 points, and the intervals in between are evenly divided; The third row is the APACHE II score. An APACHE II score of 0 corresponds to a score scale of 0, and an APACHE II score of 55 corresponds to a score scale of 98. The intervals in between are evenly divided. The fourth row is the heart rate upon admission to the ICU, with 50 beats corresponding to 0 points, 180 beats corresponding to 45 points, and the intervals in between divided equally; The fifth row is the peripheral blood NK cell count, 800 cells / μL is 0 points, 0 cells / μL is 100 points, and the intervals in between are divided equally; The sixth line is whether the patient was transferred due to lung infection, with 0 points for no and 33 points for yes; The seventh line is the time series classification of peripheral blood lymphocyte count, where β type is 0 points, γ type is 2 points, α type is 13 points, and δ type is 22 points; The eighth line predicts the 28-day mortality rate, with a probability range of 0.1 to 0.

8. In the nomogram, the second to seventh rows are independent risk factors for death within 28 days. Different factor states correspond to different scores on the scale. The eighth row is the sum of the scores of each factor. The ninth row is the predicted probability of death within 28 days. The sum of the different scores of the second to seventh rows is projected to the corresponding position of the eighth row, which is the predicted probability of death within 28 days.

4. The method according to claim 1, characterized in that The AUC value of the ROC curve for predicting the risk of death within 28 days in patients with sepsis based on the time trajectory typing of peripheral blood lymphocyte counts 72 hours after transfer to the ICU was 0.754.

Images