Method and apparatus for renal transplant recipient infection prediction

By screening the study population, collecting data, and building a predictive model, the problem of predicting infection in kidney transplant recipients was solved, enabling accurate prediction and individualized management of infection in kidney transplant recipients and reducing the risk of infection.

CN119049735APending Publication Date: 2024-11-29WEST CHINA HOSPITAL SICHUAN UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411556611.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-04
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The lack of readily available and simple methods for predicting infection in kidney transplant recipients leads to inappropriate use of antibiotics during immunosuppressive therapy, increasing the risk of infection and making it difficult to design effective infection prevention and management strategies.

Method used

By screening the study population based on pre-defined inclusion and exclusion criteria, collecting clinical information and immunosuppressant drug concentration monitoring results, conducting baseline analysis, LASSO regression, correlation analysis, and machine learning algorithm evaluation, a predictive model is constructed to accurately predict infection in kidney transplant recipients.

Benefits of technology

This provides an accurate and efficient method for predicting infection in kidney transplant recipients, improving the effectiveness of infection prevention and treatment, reducing the incidence of infectious diseases, and minimizing adverse consequences.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119049735A_ABST
    Figure CN119049735A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of medical treatment, and particularly discloses a method and equipment for predicting infection of kidney transplantation recipients, and the method comprises the steps: determining a research population based on a preset containing and discharging standard; collecting clinical information of research people, immunosuppressor drug concentration monitoring results and conventional laboratory inspection results; performing baseline analysis, LASSO regression, correlation analysis, single factor analysis and subject working characteristic curve analysis processing on the basic variable data, and screening and determining a final variable for constructing a prediction model; evaluating a to-be-selected machine learning algorithm based on the ROC curve area, the sensitivity, the specificity, the positive predicted value and the negative predicted value, and determining a final algorithm for constructing a prediction model; and constructing a prediction model based on the data of the final variable and the final algorithm, and performing infection prediction based on the to-be-detected data and the prediction model. Therefore, infection prediction of the kidney transplantation recipient can be accurately and efficiently realized through the constructed prediction model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of medical technology, and in particular to a method and device for predicting infection in a kidney transplant recipient. Background Art

[0002] Immunosuppressive therapy is an important part of the anti-rejection process after kidney transplantation, and optimal immunosuppressive management is the basis for successful kidney transplantation. Due to the continuous use of immunosuppressants, kidney transplant patients, or recipients, have a higher risk of infection, and post-transplant infection is one of the most important causes of death in organ transplant recipients. Therefore, active and effective prevention and timely and reasonable treatment of infection are important components of long-term management of patients after organ transplantation.

[0003] Since the risk of infection in renal transplant recipients is affected by multiple factors such as the degree of immunosuppression, time after transplantation, type of infection, and the patient's basic conditions, accurate treatment of post-transplant infection becomes very difficult. At present, there are still controversies about when preventive anti-infection treatment should be started and whether transplant recipients can benefit most from this treatment. Inappropriate use of antibiotics is not uncommon, which may lead to many adverse consequences (such as antibiotic resistance and adverse drug reactions). Therefore, effective measures or methods are needed in clinical practice to assess the infection risk of renal transplant recipients in order to design reasonable and effective infection prevention management strategies. However, the current infection prevention management of solid organ recipients is largely based on clinical experience, and there is a lack of simple and easy-to-access effective methods for predicting infection in renal transplant recipients. Summary of the invention

[0004] In view of this, the object of the present invention is to provide a method and device for predicting infection in renal transplant recipients, so as to overcome the problem that the infection in renal transplant recipients cannot be accurately and efficiently predicted at present.

[0005] To achieve the above objectives, the present invention adopts the following technical solutions: In a first aspect, an embodiment of the present application provides a method for predicting infection in a renal transplant recipient, comprising: Based on the preset inclusion and exclusion criteria, the target population is screened to determine the research population; Collect clinical information, immunosuppressant drug concentration monitoring results and routine laboratory test results of the study population to generate basic variable data for building a prediction model; wherein the clinical information includes age, gender and transplantation time; the routine laboratory test results include liver function test results, renal function test results and blood routine test results; Performing baseline analysis, LASSO regression, correlation analysis, univariate analysis, and receiver operating characteristic curve analysis on the basic variable data to screen and determine the final variables for building a prediction model; Evaluate the selected machine learning algorithms based on ROC curve area, sensitivity, specificity, positive predictive value, and negative predictive value to determine the final algorithm for constructing the prediction model; Based on the data of the final variable and the final algorithm, a prediction model is constructed, and infection prediction is performed based on the data to be tested and the prediction model.

[0006] Furthermore, in some embodiments of the present application, the target population is screened based on the preset inclusion and exclusion criteria to determine the research population, including: The infected kidney transplant recipients whose transplant time was greater than 3 months, whose complete immunosuppressant drug concentration monitoring was performed within 1 month before admission or consultation, and whose routine laboratory test results were available were included in the infected group; Kidney transplant recipients who were followed up in the same hospital during the same period and did not develop infection were randomly selected as the control group; Wherein, the immunosuppressant drug concentration monitoring includes tacrolimus trough concentration and MPA concentration monitoring at each time point.

[0007] Further, in some embodiments of the present application, the basic variable data are subjected to baseline analysis, LASSO regression, correlation analysis, univariate analysis and receiver operating characteristic curve analysis to screen and determine the final variables for constructing a prediction model, including: Performing a baseline analysis on the basic variable data to determine a first basic variable with a significant difference; Performing LASSO regression identification on the basic variable data to determine the optimal combination including the second basic variable; Taking the intersection of the first basic variable and the second basic variable to obtain a third basic variable; Performing a correlation analysis on the third basic variables to delete some variables with strong correlation, thereby obtaining a fourth basic variable; wherein a correlation coefficient greater than 0.7 or less than -0.7 is defined as having strong correlation; Performing univariate analysis on the fourth basic variable to screen and identify outcome variables associated with infection in kidney transplant recipients, thereby obtaining a fifth basic variable; A receiver operating characteristic curve analysis is performed on the fifth basic variable to determine the final variable.

[0008] Furthermore, in some embodiments of the present application, the final variables include: MPA plasma concentration before medication, MPA plasma concentration 4 hours after medication, cystatin C level, neutrophil-lymphocyte ratio and hemoglobin concentration.

[0009] Furthermore, in some embodiments of the present application, For the infection group, the immunosuppressant drug concentration monitoring results included mycophenolic acid drug concentration and MPA-AUC0-4 and tacrolimus trough concentration before medication, 0.5h, 2h, and 4h after medication 1 month before admission; For the control group, the immunosuppressant drug concentration monitoring results included mycophenolic acid drug concentration and MPA-AUC0-4 as well as tacrolimus trough concentration before medication, 0.5h after medication, 2h after medication, and 4h after medication 1 month before and 1 month after enrollment.

[0010] Furthermore, in some embodiments of the present application, the selected machine learning algorithms include XGBoost, Logistic, LightGBM, RandomForest and AdaBoost.

[0011] Furthermore, in some embodiments of the present application, it also includes: determining an external validation study population for the target time period based on the inclusion and exclusion criteria, and externally validating the prediction model based on the data of the external validation study population.

[0012] In a second aspect, an embodiment of the present application provides a device for predicting infection in a renal transplant recipient, comprising a processor and a memory, wherein the processor is connected to the memory: Wherein, the processor is used to call and execute the program stored in the memory; The memory is used to store the program, and the program is used to execute the above-mentioned method for predicting infection in kidney transplant recipients.

[0013] The present invention relates to the field of medical technology, and specifically discloses a method and device for predicting infection in kidney transplant recipients, the method comprising: screening a target population based on a preset inclusion and exclusion standard to determine a research population; collecting clinical information, immunosuppressant drug concentration monitoring results and routine laboratory test results of the research population to generate basic variable data for building a prediction model; performing baseline analysis, LASSO regression, correlation analysis, univariate analysis and receiver operating characteristic curve analysis on the basic variable data to screen and determine the final variable for building a prediction model; evaluating a machine learning algorithm to be selected based on ROC curve area, sensitivity, specificity, positive predictive value and negative predictive value to determine the final algorithm for building a prediction model; building a prediction model based on the data of the final variable and the final algorithm, and performing infection prediction based on the data to be tested and the prediction model. In this way, infection prediction of kidney transplant recipients can be accurately and efficiently achieved through the constructed prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 is a schematic flow chart of a method for predicting infection in a renal transplant recipient provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the principle of determining the research population in the method for predicting infection in kidney transplant recipients provided in an embodiment of the present invention; Figure 3 It is a schematic diagram of the principle of determining the final variable in the method for predicting infection in kidney transplant recipients provided by an embodiment of the present invention; Figure 4 It is a schematic diagram of the principle of determining the final algorithm in the method for predicting infection in kidney transplant recipients provided by an embodiment of the present invention; Figure 5 It is a result diagram of characteristic factors screened based on LASSO regression in the method for predicting infection in renal transplant recipients provided by an embodiment of the present invention; Figure 6 is a heat map of correlation analysis between variables in the method for predicting infection in renal transplant recipients provided by an embodiment of the present invention; Figure 7 is a ROC analysis diagram of variables used to predict the occurrence of infection in KTRs in the method for predicting infection in renal transplant recipients provided in an embodiment of the present invention; Figure 8 It is a calibration curve and decision curve analysis diagram of five candidate machine learning algorithm models in the method for predicting infection in renal transplant recipients provided by an embodiment of the present invention; Fig. 9 It is a result diagram of some evaluation indicators of the model in the method for predicting infection in kidney transplant recipients provided in an embodiment of the present invention; Fig.10 It is another part of the evaluation index result diagram of the model in the method for predicting infection in renal transplant recipients provided in an embodiment of the present invention; Fig.11 is an analysis diagram of the importance of the final variables in the method for predicting infection in renal transplant recipients provided by an embodiment of the present invention; Fig.12 It is an analysis diagram of the externally validated ROC curve in the method for predicting infection in renal transplant recipients provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0017] Figure 1 is a flow chart of a method for predicting infection in kidney transplant recipients provided by an embodiment of the present invention. Figure 1 , this embodiment may include the following steps: S101. Based on the preset inclusion and exclusion criteria, screen the target population and determine the research population.

[0018] S102. Collect clinical information, immunosuppressant drug concentration monitoring results and routine laboratory test results of the study population to generate basic variable data for building a predictive model; clinical information includes age, gender and transplantation time; routine laboratory test results include liver function test results, renal function test results and blood routine test results.

[0019] S103. Perform baseline analysis, LASSO regression, correlation analysis, univariate analysis, and receiver operating characteristic curve analysis on the basic variable data to screen and determine the final variables for building a prediction model.

[0020] Specifically, the receiver operating characteristic curve analysis is ROC curve analysis, and the basic variable data can be analyzed and screened in turn through the above analysis method to determine the relationship between each basic variable and the relationship with the results, so as to determine the optimal final variable for building a prediction model.

[0021] S104. Evaluate the candidate machine learning algorithms based on the ROC curve area, sensitivity, specificity, positive predictive value, and negative predictive value, and determine the final algorithm for constructing the prediction model.

[0022] Among them, the area of ​​the receiver operating characteristic curve is the ROC curve area. The prediction performance of multiple candidate machine learning algorithms is tested and evaluated through the above information, so that the algorithm with the best performance is determined as the final algorithm for building the prediction model.

[0023] S105. Based on the data of the final variables and the final algorithm, a prediction model is constructed, and infection prediction is performed based on the data to be tested and the prediction model.

[0024] Specifically, after obtaining the prediction model, the corresponding data of the actual kidney transplant recipient to be predicted is input into the prediction model, and the prediction model outputs the prediction result to obtain the prediction result of whether the kidney transplant recipient is likely to be infected or the probability of infection.

[0025] The method for predicting infection in renal transplant recipients provided in the embodiment of the present application is to determine the study population based on the preset inclusion and exclusion criteria; collect clinical information, immunosuppressant drug concentration monitoring results and routine laboratory test results of the study population; perform baseline analysis, LASSO regression, correlation analysis, univariate analysis and receiver operating characteristic curve analysis on the basic variable data, screen and determine the final variable for building a prediction model; evaluate the machine learning algorithm to be selected based on the ROC curve area, sensitivity, specificity, positive predictive value and negative predictive value, and determine the final algorithm for building a prediction model; build a prediction model based on the data of the final variable and the final algorithm, and perform infection prediction based on the data to be tested and the prediction model. In this way, the prediction model constructed can accurately and efficiently realize the infection prediction of renal transplant recipients.

[0026] In some embodiments of the present application, the target population is screened based on the preset inclusion and exclusion criteria to determine the research population, which may specifically include: first, the infected kidney transplant recipients whose transplant time is greater than 3 months, who have undergone complete immunosuppressant drug concentration monitoring within 1 month before admission or consultation, and whose routine laboratory test results are available are selected as the infection group; wherein infection is defined as: clear pathogenic microorganism test results; or considered to be infected from imaging examinations; or having fever and cough symptoms and signs; and all patients are diagnosed with infection upon discharge; then, kidney transplant recipients who are followed up in the same hospital during the same period and have no infection within 1 month before and after the follow-up are randomly selected as the control group; wherein immunosuppressant drug concentration monitoring includes tacrolimus trough concentration monitoring, MPA concentration monitoring at each time point and simplified MPA concentration-time curve area under curve calculation (Area Under The Concentration-Time Curve, MPA-AUC0-4); routine laboratory tests include liver function tests, renal function tests and blood routine tests, and the routine laboratory test results are the corresponding test results.

[0027] For example, Figure 2 FIG. 1 is a schematic diagram showing the principle of determining a research population in the method for predicting infection in kidney transplant recipients provided in an embodiment of the present invention. Figure 2 In the study, 4685 kidney transplant recipients who were hospitalized in a target medical institution such as a certain hospital during the target period from January 2016 to November 2019 were taken as an example for screening.

[0028] For the infected patients, the inclusion criteria were: transplant time was greater than 3 months, the kidney transplant recipients had complete immunosuppressant concentration monitoring (tacrolimus trough concentration (TAC-C0), MPA concentration at each time point, and MPA-AUC0-4) within 1 month before admission or treatment, and routine laboratory test results such as liver function, kidney function, and blood routine were available. Finally, 180 kidney transplant recipients who were hospitalized due to infection were included as the infection group.

[0029] Among them, the definition of infection can be: clear test results of pathogenic microorganisms; or considered to be infected from imaging examinations; or symptoms and signs such as fever and cough; and all patients are diagnosed with infection upon discharge. Then, on this basis, 91 renal transplant recipients who were followed up in this hospital during the same period and whose general conditions were stable and who had no infection within 1 month before and after the follow-up were randomly selected as the control group for machine learning and model construction.

[0030] In addition, in the embodiment of the present application, the external validation study population of the target period can also be determined based on the same inclusion and exclusion criteria, and the prediction model can be externally validated based on the data of the external validation study population. For example, 100 kidney transplant recipients (70 in the infection group and 30 in the control group) treated in the same hospital from January 2021 to December 2023 can be collected with the same inclusion and exclusion criteria for external validation of the prediction model.

[0031] Figure 3 FIG. 1 is a schematic diagram of the principle of determining the final variable in the method for predicting infection in kidney transplant recipients provided in an embodiment of the present invention, such as Figure 3 As shown, the process specifically includes: First, a baseline analysis was performed on the basic variable data to determine the first basic variable with significant differences.

[0032] Then, LASSO regression is performed on the basic variable data to determine the optimal combination including the second basic variable.

[0033] Specifically, in the embodiment of the present application, the clinical information, immunosuppressant drug concentration monitoring results and routine laboratory test results of the above-mentioned researchers who meet the above-mentioned inclusion criteria are first collected, including: the age, gender and transplantation time of the renal transplant recipients; the mycophenolic acid drug concentration and MPA-AUC0-4 and tacrolimus trough concentration (TAC-C0) before medication (MPA-C0.5), 0.5h after medication (MPA-C2), 2h after medication (MPA-C4) and 4h after medication (MPA-C0) of the renal transplant recipients within 1 month before admission (this is for the infection group, and for the control group because they are not admitted to the hospital, it can be 1 month before and one month after admission); the blood routine, liver function, renal function, pathogenic microorganisms and other routine laboratory test results of the renal transplant recipients (the same as the above-mentioned immunosuppressant drug concentration monitoring results, for the infection group, it can be the data one month before admission, and for the control group, it can be 1 month before and one month after admission); and it can also include auxiliary examination materials such as imaging examinations and related information such as clinical diagnosis.

[0034] After obtaining the above data, the above data are processed, where categorical variables are expressed as counts, and continuous variables are expressed as mean ± standard deviation or median (depending on the situation). Welch's T-Test or Mann-Whitney U test can be used to compare continuous variables between two groups, and chi-square test can be used to compare categorical variables.

[0035] Then take the intersection of the first basic variable and the second basic variable to obtain the third basic variable.

[0036] Then, correlation analysis is performed on the third basic variables to delete variables with strong correlation and obtain the fourth basic variables.

[0037] The fourth basic variable was then subjected to univariate analysis to screen and identify outcome variables associated with infection in renal transplant recipients, thereby obtaining the fifth basic variable.

[0038] Finally, the fifth basic variable is subjected to ROC curve analysis to determine the final variable.

[0039] In the present application embodiment, the "Glmnet" package in R version 4.1.8 can be used to perform LASSO regression for filtering. Wherein, in LASSO regression, the beta coefficient of the variable that is not associated with the result is reduced to zero, thereby removing these variables. The correlation between the variables is analyzed using Spearman correlation. The relationship between the univariate analysis observation variable and the outcome is generated to generate a receiver operating characteristic curve (ROC), and the Youden index and the optimal threshold of each variable are automatically calculated at the same time, and the final analysis and screening are used to obtain the final variable for building a prediction model.

[0040] The final variables specifically include: MPA plasma concentration before medication (MPA-C0), MPA plasma concentration 4 hours after medication (MPA-C4), cystatin C level (Cystatin c, Cyc), neutrophil-to-lymphocyte ratio (NLR) and hemoglobin concentration (Hb).

[0041] Figure 4 FIG. 1 is a schematic diagram of the principle of determining the final algorithm in the method for predicting infection in kidney transplant recipients provided in an embodiment of the present invention, such as Figure 4 As shown in FIG, the process specifically includes: evaluating the selected machine learning algorithms based on the ROC curve area (also known as AUC), sensitivity, specificity, positive predictive value, and negative predictive value, and determining the final algorithm for building the prediction model.

[0042] Specifically, Figure 4 As shown, in some embodiments of the present application, in order to establish a prediction model, the candidate machine learning algorithms XGBoost, Logistic, LightGBM, RandomForest and AdaBoost can be tested separately through the Deepwise DxAI platform. By evaluating the area under the receiver operating characteristic (ROC) curve (Area Under The Curve, AUC), sensitivity (Sensitivity), specificity (Specificity), positive predictive value (Positive Predictive Value, PPV) and negative predictive value (Negative Predictive Value, NPV), and in some embodiments, the accuracy and F1 score can also be added to evaluate the performance of the five algorithms to build the model, and determine the algorithm with the best performance as the final algorithm.

[0043] On this basis, the final variables screened out are used with the final algorithm to construct an infection prediction model, and the accuracy of the model is evaluated by ROC-AUC, learning curve, calibration curve, and clinical decision curve analysis (DCA) of the training set, validation set, and test set. The variables used to construct the model are weighted and analyzed by SHAP values, and a model list is generated through a preset platform such as the Zhiyan platform to further generate an online prediction model. In addition, data from 100 kidney transplant recipients from January to December 2023 are collected for external validation of the model.

[0044] The following will introduce in detail the method for predicting infection in kidney transplant recipients provided by this application in combination with specific data information, including: In determining the research population, the above Figure 2 , specifically including: First, a preliminary screening was conducted on 4685 hospitalized renal transplant recipients, of which 2001 were hospitalized for infection and had a transplant time of more than 3 months. Only 180 infected renal transplant recipients had complete immunosuppressant concentration monitoring (MPA concentration at each time point, MPA-AUC0-4, TAC-C0) and routine laboratory test results (liver function, renal function and blood routine) within 1 month before admission.

[0045] Among them, there were no significant differences in gender composition, age and time after transplantation between the infected group of kidney transplant recipients (KTRs) (180) and the stable group (i.e., control group) of KTRs (91) (p>0.05).

[0046] Through analysis, the MPA concentration and MPA-AUC0-4 at each time point in the infected group of renal transplant recipients were significantly higher than those in the stable group (p < 0.001), while there was no significant difference in TAC-C0 between the two groups (p > 0.05).

[0047] In routine blood tests, the red blood cell count (RBC), hemoglobin concentration (Hb), hematokrit (Hct), lymphocyte percentage (Lym), and monocyte percentage (Mon) of the KTRs in the infection group were significantly lower than those in the stable group, while the neutrophil percentage (Neu), neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR) were significantly higher than those in the stable group (p < 0.001).

[0048] In the liver function test, the total bilirubin (TBi), indirect bilirubin (DBi), and glutamyltransferase (GGT) of KTRs in the infection group were significantly higher than those in the stable group (p < 0.05); while the total protein (TP) and albumin (Alb) were significantly lower than those in the stable group (p < 0.001). In the renal function test, the urea (Ure), serum creatinine (Cr) and cystatin C (Cyc) of KTRs in the infection group were significantly higher than those in the stable group (p < 0.001).

[0049] The basic variable data and related information of the specific research population are shown in Table 1 below: Infection group (n=180) Stable group (n=91) P-value Gender (female / male) 129 / 51 62 / 29 0.547 Age, years old 34.000[28.000,44.000] 34.000[24.000,44.000] 0.971 Time after transplantation, months 13.960[3.104-84.021] 15.320[3.454-75.041] 0.594 <![CDATA[TAC-C0,ng / mL]]> 5.400[4.200,6.900] 5.900[4.700,6.800] 0.060 <![CDATA[MPA-C0,ug / mL]]> 3.890[2.370,5.490] 2.460[1.820,3.000] <0.001 <![CDATA[MPA-C 0.5 ,ug / mL]]> 16.550[11.550,24.300] 11.650[6.300,22.650] 0.026 <![CDATA[MPA-C2,ug / mL]]> 8.350[5.530,11.350] 5.720[4.430,7.040] <0.001 <![CDATA[MPA-C4,ug / mL]]> 5.020[3.640,6.790] 2.890[2.420,3.540] <0.001 <![CDATA[MPA-AUC 0-4 ,mg.h / L]]> 65.200[54.100,78.700] 52.800[47.300,59.900] <0.001 RBC,10^12 / L 3.843±0.770 4.745±0.588 <0.001 Hb,g / L 115.000[100.000,128.000] 141.000[129.000,152.000] <0.001 Hct 0.360[0.320,0.400] 0.440[0.400,0.470] <0.001 PLT, 10^9 / L 192.000[145.000,251.000] 190.000[140.000,262.000] 0.652 WBC, 10^12 / L 7.190[5.590,9.200] 6.930[5.880,8.150] 0.204 Neu,% 72.800[65.000,78.100] 63.600[58.900,67.000] <0.001 Lym,% 17.200[11.900,22.600] 25.900[22.100,31.000] <0.001 Mon,% 7.300[5.700,8.900] 8.100[7.000,9.300] 0.002 NLR 4.079[2.986,6.673] 2.426[1.928,3.005] <0.001 PLR 149.111[103.520,233.854] 109.622[80.042,152.508] <0.001 TBi,umol / L 10.500[8.200,14.400] 7.800[5.700,10.300] <0.001 DBi,umol / L 3.000[2.300,4.300] 2.600[2.000,3.500] 0.009 ALT,U / L 15.000[11.000,25.000] 13.000[9.000,25.000] 0.223 AST,U / L 17.000[13.000,23.000] 18.000[15.000,22.000] 0.279 TP,g / L 68.700[62.100,72.500] 71.700[68.600,75.000] <0.001 Alb,g / L 44.100[39.800,46.600] 48.000[45.400,50.100] <0.001 ALP,U / L 83.000[62.000,109.000] 80.000[52.000,100.000] 0.119 GGT, U / L 23.000[17.000,36.000] 20.000[15.000,23.000] 0.001 Ure,mmol / L 8.700[6.300,12.900] 6.600[5.100,8.400] <0.001 Cr,mmol / L 128.000[103.000,169.000] 101.000[84.000,113.000] <0.001 Cyc,mg / L 1.780[1.410,2.350] 1.210[1.100,1.340] <0.001 UA,umol / L 363.000[300.000,438.000] 362.000[324.000,405.000] 0.337 Among them, in the above table, RBC represents red blood cell count; Hb represents hemoglobin; Hct represents hematocrit; PLT represents platelet count; WBC represents white blood cell count; Neu represents the percentage of neutrophils; Lym represents the percentage of lymphocytes; Mon represents the percentage of monocytes; NLR represents the percentage of neutrophils and lymphocytes; PLR represents the percentage of platelets and lymphocytes; TBi represents total bilirubin; DBi represents direct bilirubin; ALT represents alanine aminotransferase; AST represents aspartate aminotransferase; TP represents total protein; Alb represents albumin; ALP represents alkaline phosphatase; GGT represents glutamyl transpeptidase; Ure represents urea; Cr represents creatinine; Cyc represents cystatin C; UA represents uric acid.

[0050] On this basis, the above basic variable data are screened to determine the final variables used to build the model. The specific process is as follows: The Least Absolute Shrinkage And Selection Operator (LASSO) with the "Lambda.Min" criterion can be used to select variables for the prediction model. Specifically, when the standard error of the minimum distance λ is 0.014, the 17 best variables (i.e., the second basic variables) of the corresponding model are selected as: MPA-AUC 0-4 , MPA-C4, MPA-C0, Cyc, Alb, AST, ALT, DBi, TBi, NLR, Mon, Lym, WBC, Hct, Hb, ALP and GGT, as shown in the following table: Figure 5 shown.

[0051] Figure 5-Figure 7In the table, MPAC0 is MPA-C0; MPAC4 is MPA-C4; MPAAUC is MPA-AUC 0-4 (Because the analysis software cannot add relevant special characters, the symbol form in this part of the analysis diagram cannot be exactly the same as the symbol form in the text. Please note that) Figure 5 A in the figure is the cross-validation curve corresponding to the LASSO regression screening characteristic factor, that is, the basic variable. The upper horizontal axis represents the number of variables corresponding to different λ, the lower horizontal axis represents the logarithm of the Lambda penalty coefficient (logλ), and the vertical axis represents the cross-validation error. The smaller the vertical axis, the better the LASSO fitting effect, and the cross-validation error corresponding to the dotted line on the left side of the figure is the smallest, which determines how many variables can be used for analysis. Figure 5 B in the figure is the coefficient path diagram corresponding to the LASSO regression screening characteristic factors, where the upper horizontal axis represents the number of non-zero coefficients in the model, the vertical axis represents the value of the coefficient, and the lower horizontal axis represents the logarithm of the lambda penalty coefficient (logλ) (the end of each gene points to its own coefficient). Lines of different colors represent different variables, and each curve represents the change trajectory of each independent variable coefficient. The coefficient (Coefficients) corresponding to the dotted line on the right side of the figure has the smallest value, which determines how many variables can be used for analysis.

[0052] And the above 17 second basic variables are intersected with the 21 first basic variables with significant differences determined by baseline analysis to obtain 13 third basic variables (specifically Figure 6 ), and then conduct correlation analysis.

[0053] Specifically, through Spearman correlation analysis, it was found that there was a strong correlation between multiple groups of variables (specifically Figure 6 ). Among them, correlation coefficients greater than 0.7 or less than -0.7 are considered strong correlations, between 0.4 and 0.7 or between -0.4 and -0.7 are considered moderate correlations, and between 0 and 0.4 or between -0 and -0.4 are considered weak correlations. 0-4 The correlation coefficient between Hb and Hct is 0.757, the correlation coefficient between NLR and Lym is -0.981, and the correlation coefficient between TBi and DBi is 0.786. In this way, 13 third basic variables were screened by correlation analysis, and the variables with strong correlation were deleted to obtain 11 fourth basic variables.

[0054] On this basis, univariate analysis was used to preliminarily explore the relationship between these variables with significant differences (i.e., 11 fourth basic variables) and KTRs infection. The odd ratios (OR) values, 95% confidence intervals (95% CI) and p values ​​of each variable are shown in Table 2. Cyc has a higher OR value of 122.647, 95% CI (30.784, 488.639); followed by MPA-C4 with an OR value of 2.374, 95% CI (1.845, 3.055); MPA-C0 with an OR value of 1.850, 95% CI (1.490-2.296); and NLR with an OR value of 1.968, 95% CI (1.579, 2.453). Hb is a protective factor, with an OR value of 0.927, 95% CI (0.908, 0.945). MPA-C 0.5 The two indicators, DBi, were not related to the outcome of KTRs infection (P>0.05).Through analysis, the 11 fourth basic variables were further screened and 10 fifth basic variables were determined.

[0055] Table 2. ORs, 95% confidence intervals (95%CI), and p-values ​​of variables associated with infection outcomes in renal transplant recipients variable OR 95% CI P-value <![CDATA[MPA-C0]]> 1.850 [1.490,2.296] 0.000 <![CDATA[MPA-C4]]> 2.374 [1.845,3.055] 0.000 <![CDATA[MPA-AUC 0-4 ]]> 1.077 [1.050,1.105] 0.000 Hb 0.927 [0.908,0.945] 0.000 Mon 0.872 [0.791,0.961] 0.006 NLR 1.968 [1.579,2.453] 0.000 TBi 0.810 [0.750,0.874] 0.000 DbI 0.898 [0.768,1.049] 0.176 Alb 0.776 [0.717,0.841] 0.000 Cyc 122.647 [30.784,488.639] 0.000 GGT 1.025 [1.007,1.043] 0.005 Furthermore, on this basis, ROC curve analysis was performed, and the 10 fifth basic variables were further screened based on the analysis results.

[0056] Specifically, the ROC curve analysis of the variables used to predict the incidence of KTRs infection is shown in Figure 7 As shown in Table 3, the characteristic description of the ROC curve analysis results is shown in Table 3. In this application, the ROC-AUC, sensitivity, and specificity of these 10 variables (i.e., the fifth basic variable) are arranged from large to small, and finally 5 variables are comprehensively selected as the final variables for model construction, which are MPA-C0, MPA-C4, Cyc, NLR, and Hb. The ROC-AUC for Cyc to predict the occurrence of infection is 0.869, and the ROC-AUC for Hb, MPA-C4, NLR, and MPA-AUC are 0.869. 0-4 , MPA-C0 were 0.837, 0.824, 0.786, 0.749, and 0.727, respectively. The specificity of Cyc for predicting KTRs infection was 0.934, while the sensitivity was only 0.725; MPA-AUC 0-4The specificity of MPA-C0 was 0.978, and the sensitivity was 0.528; the specificity of MPA-C4 was 0.769, and the sensitivity was 0.758; the specificity of Hb was 0.860, and the sensitivity was 0.703; the specificity of NLR was 0.758, and the sensitivity was 0.753 (Table 3).

[0057] Table 3 Characteristic description of ROC curve analysis results Variables ROC-AUC Specificity Sensitivity Cut-off value Cyc 0.869 0.934 0.725 1.470 <![CDATA[MPA-C0]]> 0.727 0.978 0.528 3.630 <![CDATA[MPA-C4]]> 0.824 0.769 0.758 3.580 <![CDATA[MPA-AUC 0-4 ]]> 0.749 1.000 0.421 69.700 Hb 0.837 0.86 0.703 134.000 NLR 0.786 0.758 0.753 3.051 Mon 0.614 0.506 0.725 7.300 TBi 0.711 0.736 0.582 10.100 Alb 0.769 0.758 0.670 46.700 GGT 0.619 0.857 0.410 27.000 After obtaining the above-mentioned final variables, the prediction model is constructed using the final algorithm. The following is a detailed description of the process of determining the final algorithm and constructing and verifying the model in this application in combination with specific information data: First, in the training data set, XGBoost, Logistic, LightGBM, RandomForest and AdaBoost models were established. The ROC-AUC of the test data set were 0.926, 0.953, 0.936, 0.937 and 0.940, respectively (as shown in Table 4). The prediction performance of the five models is high, but in comparison, the prediction performance of Logistic is the highest among the five models (ROC-AUC=0.953), with a sensitivity of 0.900 and a specificity of 0.923. Calibration curves were drawn for the five machine learning models in the test data set, and the graph shows that all five models have good prediction accuracy. Among them, the Brier Score of Logistic and RandomForest is the lowest, which are 0.085 and 0.079, respectively (as shown in Table 4). Figure 8 A DCA analysis was performed on the five machine learning models in the test dataset to compare the net benefits of the best model and alternative approaches to clinical decision making. The treatment strategy provided by any of the five machine learning-based models was superior to the default strategy of treating all patients or not treating any patients. The net benefit of the logistic model exceeded that of the other machine learning models (such as Figure 8 (as shown in B in the figure).

[0058] Table 4 Prediction performance of the model constructed by each algorithm Model AUC Accuracy Sensitivity Specificity PPV NPV F1 Score XGBoost 0.926 0.834 0.883 0.880 0.915 0.716) 0.898 Logistic 0.953 0.886 0.900 0.923 0.948 0.790 0.922 LightGBM 0.936 0.845 0.828 0.945 0.924 0.732 0.873 RandomForest 0.937 0.882 0.883 0.902 0.931 0.798 0.906 AdaBoost 0.940 0.864 0.828 0.956 0.910 0.781 0.866 Among them, PPV is Positive Predictive Value; NPV is Negative Predictive Value. Figure 8DCA plots for calibration and decision curve analysis of the five models, where the straight line with a slope of 1 represents the ideal curve, the actual curve represents the calibration curve of the model, and the imaginary curve represents the uncalibrated curve. The closer the calibration curve is to the ideal curve, the better the calibration of the model. The black line represents the assumption that all patients received the intervention, while the red line represents that no patients received the intervention. If the DCA curve drawn is higher than these two curves, then the model is valuable.

[0059] Fig. 9 and Fig.10 The evaluation indicators of the model constructed with 5 final variables are shown. Specifically, Fig. 9 ABC in the figure are ROC-AUC, also known as ROC curve. Fig.10 ABC in it are learning curve, calibration plot, i.e. reliability curve, and DCA, i.e. decision curve. It should be noted that in the present application, the Logistic regression machine learning method is used to construct the infection prediction model of KTRs, and the variables used for modeling include MPA-C0, MPA-C4, Hb, NLR, and Cyc. 80% of the sample size was randomly selected from the total sample as the training set and validation set (n=217 cases) for cross-validation, and the remaining 20% ​​was used as the test set (n=54 cases). The ROC-AUC of the training set was 0.956±0.003, and that of the validation set was 0.917±0.027. Finally, the ROC-AUC of the model in the test set was 0.995, and the accuracy was 0.963 (such as Fig. 9 The logistic regression learning curve shows that the model has no underfitting or overfitting in the training set and the validation set. The model fits well and the ROC-AUC is close to 1 (as shown in Figure 1). Fig.10 A calibration plot was drawn for the test data set, which showed that the Brier Score of the infection prediction model constructed using logistic regression for the selected 5 indicators was 0.031, which had good prediction accuracy (as shown in Figure 2A). Fig.10 The DCA analysis was performed on the test data set to compare the net benefits of the model and alternative methods of clinical decision making. The treatment strategies provided by the prediction model built based on machine learning logistic regression were superior to the default strategy of treating all patients or not treating patients (such as Fig.10 as shown in C in the figure).

[0060] In addition, the SHapley Additive exPlanation (SHAP) algorithm is used to obtain the importance of each predictor variable to the prediction result. The variable importance graph is arranged in order from high to low (such as Fig.11In all prediction ranges, Cyc had the highest predictive value, followed by MPA-C4, MPA-C0, Hb, and NLR. In addition, in order to detect the positive and negative relationship between the predictor and the target outcome, SHAP Value was also used to explain the risk factors of infection. Fig.11 As shown in B, the horizontal axis is the SHAP value of the sample feature, and the vertical axis is different features. Red indicates that the feature has a positive impact on the prediction, and blue indicates that the feature has a negative impact on the prediction. 4, The red parts of the SHAP values ​​of MPA-C0 and NLR are almost all on the right, which means that these four variables are positive factors affecting the infection of kidney transplant recipients. The red parts of the SHAP values ​​of Hb are mostly on the left, which means that this variable is a negative factor.

[0061] It should be noted that in the model feature weight ranking, the importance of the evaluated variable is evaluated. The larger the SHAP mean, the more important the feature. In the SHAP value, the model feature order is ranked from the most important feature according to the SHAP absolute value. The blue value indicates the smaller the feature value, and the red value indicates the larger the feature value. The red value on the right indicates that the feature has a positive impact on the model prediction result.

[0062] Furthermore, the present application also includes external verification of the prediction model constructed based on the above-mentioned final variables and final algorithm.

[0063] Specifically, in practical applications, based on the inclusion and exclusion criteria in the above embodiment, 100 renal transplant recipients (70 in the infection group and 30 in the stable group) can be included as an external test data set to further evaluate the performance of the KTRs infection prediction model. The basic information is shown in Table 5. Fig.12 To externally validate the ROC curve analysis, the prediction performance of the prediction model constructed based on the final variables and the final algorithm in this application is as follows Fig.12 The AUC, sensitivity, and specificity were 0.963, 0.900, and 0.967, respectively. These results indicate that the machine learning prediction model established in this study has excellent predictive ability for infection in renal transplant recipients.

[0064] Table 5 Clinical characteristics of the external validation cohort Stable group (n=30) Infected group (n=70) P-value MPA-C0,ug / mL 2.440[2.140,3.150] 4.000[2.330,7.120] <0.001 MPA-C4,ug / mL 2.450[1.960,3.410] 4.570[2.690,6.800] <0.001 Hb, g / L 138.500[130.250,143.250] 106.000[97.250,121.500] <0.001 NLR 2.139[1.552,2.463] 3.634[2.506,6.739] <0.001 Cyc, mg / L 1.110[1.050,1.230] 1.890[1.600,2.680] <0.001 In addition, it should be noted that because the prediction model constructed in this application can be predicted online, in practical applications, relevant personnel such as patients can enter the Cyc, MPA-C4, MPA-C0, Hb and NLR test results through the web page, and then automatically obtain the prediction results, such as automatically calculating the risk of infection in the patient within the next month. The specific form can be to predict the probability of disease occurrence, and compare the predicted probability of disease occurrence with the threshold of disease occurrence. When the probability of disease occurrence is lower than the threshold of disease occurrence, it is considered that the feasibility of disease occurrence is low, and targeted suggestions can be given, such as prompting to continue to maintain the current living status for situations where the possibility of disease occurrence is low. If the incidence rate is higher than the threshold, it is considered that the possibility of disease occurrence is high, and it is recommended that patients follow up on time, take medication and receive treatment according to the doctor's instructions, etc.

[0065] The method for predicting infection in renal transplant recipients provided in this application and the constructed prediction model for infection in renal transplant recipients have high sensitivity and specificity. At the same time, the prediction model is based on conventional laboratory indicators and is easy to use. It can effectively help clinicians to make individualized predictions of infection in renal transplant recipients in a timely and accurate manner, and help improve the prevention and treatment of infection after renal transplantation. During the follow-up period of adult renal transplant recipients, standardized monitoring of mycophenolic acid therapeutic drug concentrations (MPA-C4, MPA-C0) and laboratory tests (Cyc, Hb and NLR) combined with the use of this prediction model can improve the prediction level of infectious diseases in renal transplant recipients, reduce the occurrence of infectious diseases by timely adjusting the dosage of immunosuppressants, and reduce the adverse effects on renal transplant outcomes caused by infection.

[0066] Based on the same inventive concept, the embodiment of the present application also provides a device for predicting infection in a renal transplant recipient, comprising a processor and a memory, wherein the processor is connected to the memory: wherein the processor is used to call and execute a program stored in the memory; the memory is used to store a program, and the program is at least used to execute the method for predicting infection in a renal transplant recipient in the above method embodiment. The specific implementation scheme of the device for predicting infection in a renal transplant recipient provided in the embodiment of the present application can refer to the implementation scheme of the method for predicting infection in a renal transplant recipient in any of the above embodiments, and will not be repeated here.

[0067] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0068] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0069] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0070] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0071] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0072] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0073] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0074] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0075] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for predicting infection in a renal transplant recipient, characterized in that: include: Based on the preset inclusion and exclusion criteria, the target population is screened to determine the research population; Collect clinical information, immunosuppressant drug concentration monitoring results and routine laboratory test results of the study population to generate basic variable data for building a prediction model; wherein the clinical information includes age, gender and transplantation time; the routine laboratory test results include liver function test results, renal function test results and blood routine test results; Performing baseline analysis, LASSO regression, correlation analysis, univariate analysis, and receiver operating characteristic curve analysis on the basic variable data to screen and determine the final variables for building a prediction model; Evaluate the selected machine learning algorithms based on ROC curve area, sensitivity, specificity, positive predictive value, and negative predictive value to determine the final algorithm for constructing the prediction model; Based on the data of the final variable and the final algorithm, a prediction model is constructed, and infection prediction is performed based on the data to be tested and the prediction model.

2. The method for predicting infection in renal transplant recipients according to claim 1, characterized in that: Based on the preset inclusion and exclusion criteria, the target population is screened to determine the research population, including: The infected kidney transplant recipients whose transplant time was greater than 3 months, whose complete immunosuppressant drug concentration monitoring was performed within 1 month before admission or consultation, and whose routine laboratory test results were available were included in the infected group; Kidney transplant recipients who were followed up in the same hospital during the same period and did not develop infection were randomly selected as the control group; Wherein, the immunosuppressant drug concentration monitoring includes tacrolimus trough concentration and MPA concentration monitoring at each time point.

3. The method for predicting infection in renal transplant recipients according to claim 1, characterized in that: The basic variable data are subjected to baseline analysis, LASSO regression, correlation analysis, univariate analysis and receiver operating characteristic curve analysis to screen and determine the final variables for constructing a prediction model, including: Performing a baseline analysis on the basic variable data to determine a first basic variable with a significant difference; Performing LASSO regression identification on the basic variable data to determine the optimal combination including the second basic variable; Taking the intersection of the first basic variable and the second basic variable to obtain a third basic variable; Performing a correlation analysis on the third basic variables to delete some variables with strong correlation, thereby obtaining a fourth basic variable; wherein a correlation coefficient greater than 0.7 or less than -0.7 is defined as having strong correlation; Performing univariate analysis on the fourth basic variable to screen and identify outcome variables associated with infection in kidney transplant recipients, thereby obtaining a fifth basic variable; A receiver operating characteristic curve analysis is performed on the fifth basic variable to determine the final variable.

4. The method for predicting infection in renal transplant recipients according to claim 1, characterized in that: The final variables included: MPA plasma concentration before medication, MPA plasma concentration 4 hours after medication, cystatin C level, neutrophil-lymphocyte ratio and hemoglobin concentration.

5. The method for predicting infection in renal transplant recipients according to claim 2, characterized in that: For the infection group, the immunosuppressant drug concentration monitoring results included mycophenolic acid drug concentration and MPA-AUC0-4 and tacrolimus trough concentration before medication, 0.5h, 2h, and 4h after medication 1 month before admission; For the control group, the immunosuppressant drug concentration monitoring results included mycophenolic acid drug concentration and MPA-AUC0-4 as well as tacrolimus trough concentration before medication, 0.5h after medication, 2h after medication, and 4h after medication 1 month before and 1 month after enrollment.

6. The method for predicting infection in renal transplant recipients according to claim 1, characterized in that: The candidate machine learning algorithms include XGBoost, Logistic, LightGBM, RandomForest and AdaBoost.

7. The method for predicting infection in renal transplant recipients according to claim 1, characterized in that: Also includes: Based on the inclusion and exclusion criteria, an external validation study population for the target period is determined, and the prediction model is externally validated based on the data of the external validation study population.

8. A device for predicting infection in kidney transplant recipients, characterized in that: The invention comprises a processor and a memory, wherein the processor is connected to the memory: Wherein, the processor is used to call and execute the program stored in the memory; The memory is used to store the program, and the program is at least used to execute the method for predicting infection in renal transplant recipients as described in any one of claims 1-7.

Citation Information

Patent Citations

  • Construction method of kidney transplantation anti-infection drug dosage prediction model

    CN113035369A

  • Method for detecting levels of tacrolimus phenolic acid ester and creatinine in body of kidney transplantation patient through VAMS

    CN114113570A

  • Method and system for constructing pulmonary infection risk prediction model based on machine learning

    CN118039177A

  • Machine learning-based early warning system for occurrence of acute kidney injury of critically ill patient

    CN118299054A

  • Use of machine learning models for prediction of clinical outcomes

    US20210327540A1