Prediction model of glomerular filtration rate after cardio-pulmonary resuscitation, construction method and application thereof

Abstract

The invention discloses a prediction model of a glomerular filtration rate after cardio-pulmonary resuscitation, a construction method and application thereof. The method comprises the following steps: collecting and preprocessing sample data; determining an independent variable, a dependent variable and an end variable; screening independent influence factors, and evaluating a relationship between the independent influence factors based on a variance expansion factor; constructing a column graph prediction model based on the independent influence factors; performing accuracy evaluation on the prediction model based on the accuracy and the mean absolute error; performing calibration degree analysis on the prediction model through a calibration curve, Bland-Altman consistency evaluation and pairing T test; evaluating the prediction efficiency of the prediction model through an ROC curve and an AUC; and verifying the universality and extrapolation of the prediction model. The early prediction model of the glomerular filtration rate after cardio-pulmonary resuscitation is constructed, so that the probability of chronic kidney diseases can be predicted within 24 hours, early discovery and early intervention are realized, and the disease progress is delayed.

Description

Technical Field

[0001] The present invention relates to the technical field of prediction model construction, and in particular to a prediction model for glomerular filtration rate after cardiopulmonary resuscitation, a construction method and an application thereof. Background Art

[0002] Cardiopulmonary resuscitation (CPR) is recognized worldwide as the most effective treatment for cardiac arrest (CA). After resuscitation and achieving return of spontaneous circulation (ROSC), the body suffers systemic ischemia-reperfusion injury, leading to multi-organ dysfunction and a significant contributing factor to high mortality rates after resuscitation. Renal tissue is particularly sensitive to ischemia. Once damaged, the probability of developing chronic kidney disease (CKD) or even end-stage renal disease (ESRD) increases significantly, along with the risk of death. Therefore, early dynamic assessment of renal function and identification of high-risk patients for CKD provide a "window of opportunity" for early intervention and are crucial for reducing the high mortality rate after resuscitation.

[0003] The current Kidney Disease: Improving Global Outcomes (KDIGO) guidelines set a diagnostic window for CKD at 3 months. This makes it impossible to assess the risk of CKD early after resuscitation, thus missing the optimal treatment window and leading to a high mortality rate.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the Invention

[0005] The technical task of this application is to address the above deficiencies and provide a prediction model, construction method and application of glomerular filtration rate after cardiopulmonary resuscitation. By constructing an early prediction model of glomerular filtration rate after cardiopulmonary resuscitation, this application can predict the probability of chronic kidney disease within 24 hours, achieve early detection and early intervention, thereby delaying disease progression and ultimately improving the prognosis of cardiac arrest patients.

[0006] To achieve the above objectives, this application provides the following technical solutions: According to one aspect of the present application, a method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation is provided, comprising: sample data collection and preprocessing; determining independent variables, dependent variables and outcome variables; screening independent influencing factors, and evaluating the relationship between the independent influencing factors based on the variance inflation factor; constructing a nomogram prediction model based on the independent influencing factors; evaluating the accuracy of the prediction model based on the accuracy rate and the mean absolute error; performing calibration analysis on the prediction model through a calibration curve, a Bland-Altman consistency evaluation and a paired T test; evaluating the predictive efficacy of the prediction model through an ROC curve and an AUC; and verifying the universality and extrapolation of the prediction model.

[0007] In some embodiments, the preprocessing includes: based on the exclusion criteria, screening out sample data that meets the exclusion criteria, wherein the inclusion criteria are age ≥ 18 years and time to ROSC after CA resuscitation > 24 hours, and the inclusion criteria are baseline eGFR < 60 mL / min / 1.73m 2 or a history of kidney-related disease or missing clinical data >15%.

[0008] In some embodiments, the independent variables include age, sex, body mass index, history of cardiovascular disease, history of diabetes, history of hypertension, no-flow time, time to ROSC, epinephrine dosage, white blood cells, hemoglobin, platelets, total cholesterol, triglycerides, albumin, globulin, alanine aminotransferase, aspartate aminotransferase, baseline eGFR, urea nitrogen, uric acid, blood glucose, glycosylated hemoglobin, blood potassium, blood sodium, blood calcium, lactate, brain natriuretic peptide, troponin, D-dimer, proteinuria, 24-hour urine volume and urine albumin / creatinine ratio; the dependent variable is the eGFR value measured on day 7; the outcome variable is the number of chronic kidney disease cases diagnosed according to the standard on day 90.

[0009] In some embodiments, the independent influencing factors include age, diabetes, no-flow time, baseline eGFR, ACR, and lactate.

[0010] According to another aspect of the present application, a prediction model constructed by a method for constructing a prediction model of glomerular filtration rate after cardiopulmonary resuscitation is also provided, and the linear regression equation of the prediction model is Y=89.439-0.344×age-3.610×hypertension-2.992×diabetes-0.577×no blood flow time+0.0349×baseline eGFR-0.042×ACR-0.650×lactic acid, wherein Y is the eGFR value on the 7th day after cardiac arrest and resuscitation to achieve spontaneous circulation recovery.

[0011] According to another aspect of the present application, there is also provided a prediction model constructed by a method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation, or the application of the prediction model in predicting the level of renal function after resuscitation and recovery of spontaneous circulation after cardiac arrest.

[0012] According to another aspect of the present application, a method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation is provided, and the prediction model constructed or the prediction model is used in predicting the risk of chronic kidney disease after cardiac arrest and recovery of spontaneous circulation through resuscitation.

[0013] In some embodiments, the application predicts the risk of developing chronic kidney disease within 24 hours.

[0014] Compared with the existing technology, the advantages and positive effects of this application are: by constructing an early prediction model of glomerular filtration rate after cardiopulmonary resuscitation, this application can predict the probability of chronic kidney disease within 24 hours, achieve early detection and early intervention, thereby delaying disease progression and ultimately improving the prognosis of cardiac arrest patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 A flow chart of the method in an embodiment of the present application is shown.

[0017] Figure 2 The nomogram prediction model in the embodiment of the present application is shown.

[0018] Figure 3 A scatter plot of the calibration curve of the prediction model in the embodiment of the present application is shown.

[0019] Figure 4 The Bland-Altman consistency evaluation of the prediction model in the examples of this application is shown.

[0020] Figure 5 The paired-T test analysis of the prediction model in the examples of this application is shown.

[0021] Figure 6 The ROC curve of the prediction model in the embodiment of the present application is shown.

[0022] Figure 7 A calibration scatter plot of the external validation prediction model in the examples of the present application is shown.

[0023] Figure 8 The ROC curve of the external validation prediction model in the embodiment of the present application is shown. DETAILED DESCRIPTION

[0024] In order to more clearly understand the above-mentioned objects, features and advantages of the present application, the present application is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments of the present application and the features therein can be combined with each other without conflict.

[0025] The present application will be further described below with reference to the accompanying drawings and specific embodiments.

[0026] S100, sample data collection and preprocessing Methods: Clinical data of patients hospitalized after cardiac arrest (CA) resuscitation who were admitted to Tianjin Medical University General Hospital Airport Hospital from January 2017 to December 2019 and Tianjin Medical University General Hospital from January 2017 to June 2024 were retrospectively collected. Chronic kidney disease (CKD) was diagnosed according to the Kidney Disease Improving Global Outcomes (KDIGO) diagnostic criteria: estimated glomerular filtration rate (eGFR) <60 mL / min / 1.73 m 2 Persistence for more than 90 days or the occurrence of renal damage. Renal damage includes: (1) proteinuria (urinary protein content exceeding 0.15 g in 24 hours); (2) abnormal urine sediment; (3) renal tubular lesions; (4) abnormal renal histopathological changes; (5) structural abnormalities seen on urinary system imaging; and (6) history of renal transplantation.

[0027] Glomerular filtration rate (GFR) is the gold standard for evaluating renal function and plays an important role in the diagnosis and monitoring of kidney disease.

[0028] Inclusion criteria for this study: (1) age ≥ 18 years; (2) time to ROSC after CA resuscitation > 24 hours. Exclusion criteria: (1) baseline eGFR < 60 mL / min / 1.73 m 2 ; Past history of kidney-related diseases such as CKD, renal failure, dialysis, etc.; (2) Clinical data missing >15%.

[0029] In this study, the patients were randomly divided into a training group and a validation group in a ratio of 7:3.

[0030] S200. Screen relevant variables and determine independent variables, dependent variables, and outcome variables.

[0031] Because direct measurement of glomerular filtration rate (GFR) is complex, an estimated formula (eGFR) based on serum markers is widely used in clinical practice. This formula represents the amount of plasma filtered through the glomerulus per unit time (per minute). Guidelines recommend calculating eGFR using the Chronic Kidney Disease-Epidemiology Collaboration (CKD-EPI) formula, which takes into account age, sex, and serum creatinine level.

[0032] The eGFR formula is: (1) Female: When Scr≤0.7 mg / dL, ; When Scr > 0.7 mg / dL: ; (2) Male: When Scr≤0.9 mg / dL, ; When Scr>0.9 mg / dL, ; Where Scr is the serum creatinine level (mg / dL).

[0033] Based on guideline recommendations and clinical significance, 62 variables related to GFR and CKD were initially screened. During data preprocessing, variables with strong correlations with other independent variables were removed, as were variables with missing data >15%. Data with missing data ≤15% were supplemented, with the mean imputed for normal distributions and the median imputed for skewed distributions.

[0034] Finally, 33 independent variables were determined: (1) demographic information: age, sex, body mass index; (2) underlying diseases and personal family history: history of cardiovascular disease, diabetes, and hypertension; (3) cardiopulmonary resuscitation-related indicators: no blood flow time (i.e., the time interval from the onset of CA to the start of CPR), time to ROSC, and epinephrine dosage; (4) first laboratory test results within 24 hours after resuscitation: white blood cells, hemoglobin, platelets, total cholesterol, triglycerides, albumin, globulin, alanine aminotransferase, aspartate aminotransferase, baseline eGFR (Scr is the first test value after resuscitation), urea nitrogen, uric acid, blood glucose, glycosylated hemoglobin, blood potassium, blood sodium, blood calcium, lactate, brain natriuretic peptide, troponin, D-dimer, proteinuria, 24-hour urine volume, and urine albumin-to-creatinine ratio (ACR). The dependent variable was the eGFR value measured on the 7th day (7d-eGFR). Outcome variable: number of confirmed CKD cases according to standard diagnosis on day 90. Table 1 shows the comparison of clinical baseline characteristics of the training group and the validation group in the examples of this application.

[0035] Table 1 Comparison of clinical baseline characteristics between the training group and the validation group

[0036] Note: BMI is body mass index, ROSC is return of spontaneous circulation, ALT is alanine aminotransferase, AST is aspartate aminotransferase, BNP is B-type brain natriuretic peptide, ACR is urine albumin / creatinine ratio, eGFR is estimated glomerular filtration rate, and CKD is chronic kidney disease.

[0037] S300: Screen independent influencing factors, and evaluate the relationship between the independent influencing factors based on the variance inflation factor.

[0038] SPSS 25.0 and R software (version 4.3.2) were used for data analysis and plotting. Normally distributed quantitative data were expressed as x ± s The two groups were compared using two independent sample t-tests; non-normal distribution measurement data were expressed as medians (quartiles) [ M ( Q1 , Q3 )] indicates that the two groups were compared using the Mann-Whitney U Enumeration data were expressed as number of cases and / or percentage, and the χ2 test was used for comparison between the two groups.

[0039] This study first used univariate linear analysis to screen out variables with P < 0.05 and incorporate them into the multivariate linear regression model (backward method) to obtain independent influencing factors. The variance inflation factor (VIF) was used to evaluate the relationship between influencing factors (VIF>5 was considered to have multicollinearity).

[0040] Table 2 shows the univariate and multivariate linear regression analysis in the examples of the present application. The multivariate linear regression analysis showed that age (β = -0.344, 95% CI: -0.528 to -0.160, P < 0.001), hypertension (β = -3.610, 95% CI: -5.968 to -1.252, P < 0.01), diabetes (β = -2.992, 95% CI: -5.295 to -0.689, P < 0.05), no blood flow time ( The independent influencing factors included baseline eGFR (β=-0.577, 95%CI: -0.996~-0.158, P<0.01), baseline eGFR (β=0.349, 95%CI: 0.269~0.429, P<0.001), ACR (β=-0.042, 95%CI: -0.073~-0.011, P<0.05), and lactate (β=-0.650, 95%CI: -1.214~-0.086, P<0.05).

[0041] Table 2 Univariate and multivariate linear regression analysis

[0042]

[0043] In addition, there was no multicollinearity between the variables (VIF<5): age VIF=1.105; hypertension VIF=1.157; diabetes VIF=1.116; no-flow time VIF=1.078; baseline eGFR VIF=1.043; ACR VIF=1.091; lactate VIF=1.184.

[0044] S400. Construct a nomogram prediction model based on independent influencing factors.

[0045] A nomogram prediction model was constructed based on independent influencing factors and verified in the validation group.

[0046] Figure 2 The nomogram prediction model in the embodiment of the present application is shown. For each indicator, a vertical line is drawn on the scoring axis to obtain a score, and finally the scores are added together to obtain a total score. The vertical line pointing downward from the total score corresponds to the predicted 7-day eGFR value of the patient after resuscitation.

[0047] like Figure 2As shown, the corresponding scores for age (55, 60, 65, 70, 75, 80, 85, 90) are (49, 42, 35, 28, 21, 14, 7, 0); The corresponding score for hypertension (yes, no) is (0, 15); The corresponding score for diabetes (yes, no) is (0, 12); The corresponding scores for the no-flow time (0, 2, 4, 6, 8, 10, 12, 14, 16) are (38, 33, 28, 24, 19, 14, 9, 5, 0); The corresponding scores for lactic acid (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16) are (37, 35, 32, 29, 27, 24, 24, 21, 19, 16, 13, 11, 8, 5, 3, 0); The corresponding scores for ACR (20, 40, 60, 80, 100, 120, 140, 160, 180, 200, 220, 240, 260, 280) are (45, 41, 38, 34, 31, 27, 24, 21, 17, 14, 10, 7, 3, 0); The corresponding scores for baseline eGFR (45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 105, 110, 115) are (0, 7, 14, 21, 29, 36, 43, 50, 57, 64, 71, 79, 86, 93, 100).

[0048] The total score corresponds to the predicted 7d-eGFR value: The predicted value for a score of 33 is 45 mL / min / 1.73 m 2 ; The predicted value of 54 points is 50mL / min / 1.73m 2 ; 77 points predicted value is 55mL / min / 1.73m 2 ; 97 points predicted value is 60mL / min / 1.73m 2 ; 118 points predicted value is 65mL / min / 1.73m 2 ; 138 points predicted value is 70mL / min / 1.73m 2 ; 159 points predicted value is 75mL / min / 1.73m 2 ; 179 points predicted value is 80mL / min / 1.73m 2 ; The predicted value of 200 points is 85mL / min / 1.73m 2 ; The predicted value of 220 points is 90mL / min / 1.73m 2 ; 241 points predicted value is 95mL / min / 1.73m 2; The predicted value for 261 points is 100 mL / min / 1.73 m 2 ; 282 points predicted value is 105mL / min / 1.73m 2 .

[0049] The linear regression equation of the nomogram prediction model was: Y = 89.439-0.344×age-3.610×hypertension-2.992×diabetes-0.577×no-flow time+0.0349×baseline eGFR-0.042×ACR-0.650×lactate.

[0050] S500, evaluating the accuracy of the prediction model based on the accuracy rate and the mean absolute error; and analyzing the calibration of the prediction model through a calibration curve, a Bland-Altman consistency evaluation, and a paired T test.

[0051] The accuracy of the training and validation groups was evaluated by accuracy analysis and mean absolute error (MAE). The results showed that (1) the accuracy of the training group was 97.07%; the MAE of the training group was 8.546; (2) the accuracy of the training group was 95.45%; the MAE of the validation group was 8.465.

[0052] Calibration analysis was performed using calibration curves, Bland-Altman consistency evaluation, and Paired-T test.

[0053] Figure 3 The calibration curve scatter plot of the prediction model in the embodiment of the present application is shown, wherein: Figure 3 A is the training group, Figure 3 B is the verification group. Figure 3 As shown in the calibration curve scatter plot for the training group: R = 0.928 (95% CI: 0.902-0.946), Accuracy = 97.07%, P < 0.001; this indicates that the predicted values ​​of the prediction model are highly correlated with the observed values, suggesting that the model has a good fit. The calibration curve scatter plot for the validation group: R = 0.893 (95% CI: 0.876-0.924), Accuracy = 95.45%, P < 0.001; this indicates that the predicted values ​​of the model are highly correlated with the observed values, suggesting that the model has a good fit.

[0054] Figure 4 The Bland-Altman consistency evaluation of the prediction model in the embodiment of the present application is shown, wherein Figure 4 A is the training group, Figure 4 B is the verification group. Figure 4As shown in the Bland-Altman diagram, the horizontal axis represents the mean of the results of the two methods for measuring each sample, and the vertical axis represents the difference in the measurement results of the two methods. The two blue horizontal dotted lines in the figure represent the upper and lower limits of the 95% consistency limit, which is 1.96 times the standard deviation; the red horizontal solid line in the middle represents the mean of the difference; the red horizontal dotted line indicates the position where the mean of the difference is 0. The higher the degree of consistency of the measurement results of the two methods, the closer the line representing the mean of the difference (red solid line) is to the line representing the mean of the difference of 0 (red dotted line). Among them Figure 4 The mean value of group A is 4.5. Figure 4 The mean value of group B was 3.6, and most of the differences between the two groups were within the 95% consistency limit, indicating that the prediction results of the two groups were well consistent.

[0055] Figure 5 The Paired-T test analysis of the prediction model in the embodiment of the present application is shown, wherein, Figure 5 A is the training group, Figure 5 B is the verification group. Figure 5 As shown in the figure, the Paired-T test in the training group showed P=0.89, indicating that there was no statistically significant difference between the model prediction value and the actual value (P>0.05), suggesting that the model has a good prediction fit. The Paired-T test in the validation group showed P=0.63, indicating that there was no statistically significant difference between the model prediction value and the actual value (P>0.05), suggesting that the model has a good prediction fit.

[0056] S600. The predictive efficacy of the linear regression model for CKD was evaluated by ROC curve and AUC.

[0057] The predictive efficacy of the linear regression model for CKD was evaluated by ROC (receiver operating characteristic) curve and AUC (area under the ROC curve). P The difference was statistically significant when the value was <0.05.

[0058] Figure 6 The ROC curve of the prediction model in the embodiment of the present application is shown. Figure 6 As shown in the figure, the prediction model predicted CKD after resuscitation in the training group with AUC = 0.882 (95% CI :0.857~0.925); validation group AUC=0.859 (95% CI :0.843~0.891), the results of the two groups showed that the model had good predictive efficiency; the AUC of the two groups was tested by Delong non-parametric test, and there was no statistically significant difference ( P =0.492), indicating that the grouped data are highly homogeneous.

[0059] S700. Verify the universality and extrapolation of the model through external unit data.

[0060] The same inclusion and exclusion criteria as step S100 were used to collect clinical data of 105 patients who underwent successful CPR at Shenzhen University General Hospital from January 2021 to June 2023.

[0061] Figure 7 Figure 2 shows a calibration scatter plot of the external validation prediction model in the embodiment of the present application. Figure 7 As shown, accuracy analysis and MAE were used to assess the accuracy of the external validation group: Accuracy = 91.08%; MAE = 6.215. The calibration curve scatter plot showed R = 0.499 (95% CI: 0.386-0.577), P < 0.001, indicating a high correlation between the model's predicted and observed values, suggesting a good fit for the model.

[0062] Figure 8 The ROC curve of the external validation prediction model in the embodiment of the present application is shown. Figure 8 As shown in the external validation group, the eGFR prediction model for CKD risk prediction AUC = 0.784 (95% CI :0.720~0.822), the results showed that the model had good CKD prediction efficiency.

[0063] This study, based on multivariate linear regression analysis, constructed an early prediction model for GFR, which not only helps to accurately quantify the level of renal function and guide key treatment decisions such as dialysis timing and drug dosage adjustment, but more importantly, predicts the probability of CKD within 24 hours, achieving early detection and early intervention, thereby delaying disease progression and ultimately improving the prognosis of CA patients.

[0064] Through the above specific embodiments, those skilled in the art can easily implement the present application. However, it should be understood that the present application is not limited to the above specific embodiments. Based on the disclosed embodiments, those skilled in the art can arbitrarily combine different technical features to implement different technical solutions.

Claims

1. A method for constructing a model for predicting glomerular filtration rate after cardiopulmonary resuscitation, characterized in that: include: Sample data collection and preprocessing; Identify independent, dependent, and outcome variables; Screening independent influencing factors and evaluating the relationships between the independent influencing factors based on variance inflation factors; Construct a nomogram prediction model based on independent influencing factors; The accuracy of the prediction model was evaluated based on the accuracy rate and mean absolute error; the calibration of the prediction model was analyzed by calibration curve, Bland-Altman consistency evaluation and paired T test; The predictive performance of the prediction model was evaluated by ROC curve and AUC; Verify the universality and extrapolation of the prediction model.

2. The method for constructing a post-cardiopulmonary resuscitation glomerular filtration rate prediction model according to claim 1, wherein: The preprocessing includes: based on the exclusion criteria, screening out sample data that meets the exclusion criteria, wherein the inclusion criteria are age ≥18 years and time to ROSC after CA resuscitation >24 hours, and the inclusion criteria are baseline eGFR <60 mL / min / 1.73m 2 or a history of kidney-related disease or missing clinical data >15%.

3. The method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation according to claim 2, wherein: The independent variables included age, sex, body mass index, history of cardiovascular disease, history of diabetes, history of hypertension, no-flow time, time to ROSC, epinephrine dosage, white blood cell count, hemoglobin, platelets, total cholesterol, triglycerides, albumin, globulin, alanine aminotransferase, aspartate aminotransferase, baseline eGFR, urea nitrogen, uric acid, blood glucose, glycated hemoglobin, serum potassium, serum sodium, serum calcium, lactate, brain natriuretic peptide, troponin, D-dimer, proteinuria, 24-hour urine volume, and urine albumin / creatinine ratio; the dependent variable was the eGFR value measured on day 7; and the outcome variable was the number of chronic kidney disease cases confirmed according to the standard diagnosis on day 90.

4. The method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation according to claim 3, wherein: The independent influencing factors included age, hypertension, diabetes, no-flow time, baseline eGFR, ACR and lactate.

5. The prediction model constructed by the method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation according to any one of claims 1 to 4, characterized in that: The linear regression equation of the prediction model is Y=89.439-0.344×age-3.610×hypertension-2.992×diabetes-0.577×no-flow time+0.0349×baseline eGFR-0.042×ACR-0.650×lactic acid, where Y is the eGFR value on the 7th day after cardiac arrest and resuscitation to achieve spontaneous circulation.

6. Use of the prediction model constructed by the method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation according to any one of claims 1 to 4 or the prediction model according to claim 5 in predicting the level of renal function after cardiac arrest and resuscitation to achieve spontaneous circulation.

7. Use of the prediction model constructed by the method for constructing a prediction model for glomerular filtration rate after cardiopulmonary resuscitation according to any one of claims 1 to 4 or the prediction model according to claim 5 in predicting the risk of chronic kidney disease after cardiac arrest and recovery of spontaneous circulation.

8. The use according to claim 7, characterized in that The application predicts the risk of developing chronic kidney disease within 24 hours.

Images