Construction and Evaluation of a Risk Prediction Model for Septic Cardiomyopathy Based on MIMIC-IV Database
10.3881/j.issn.1000-503X.16031
- VernacularTitle:基于MIMIC-Ⅳ数据库的脓毒性心肌病风险预测模型构建与评价
- Author:
Bin XIONG
1
;
Yinzhou LIU
;
An ZHANG
Author Information
1. 重庆医科大学附属第二医院重症医学科,重庆 400010
- Keywords:
MIMIC-Ⅳ database;
sepsis;
septic cardiomyopathy;
prediction model
- From:
Acta Academiae Medicinae Sinicae
2024;46(5):671-677
- CountryChina
- Language:Chinese
-
Abstract:
Objective To analyze the risk factors of septic cardiomyopathy(SC),and to construct and evaluate the risk prediction model of SC.Methods The clinical data of patients with sepsis were extracted from MIMIC-Ⅳ database and randomized into a training set and a validation set at a ratio of 7 to 3.According to the presence or absence of SC,the patients were assigned into SC and non-SC groups.The independent risk factors of SC were determined by univariate and multivariate Logistic regression analysis,and a risk prediction model and a nomogram were established.The area under the receiver operating characteristic curve(AUC),calibration curve,and decision curve analysis(DCA)were employed to evaluate the distinguishing degree,calibration,and clinical applicability of the model,respectively.Results A total of 2628 sepsis patients were enrolled in this study,including 1865 patients in the training set and 763 patients in the validation set.There was no signifi-cant difference in the incidence of SC between the training set and the validation set(58.98％vs.62.25％,P=0.120).Except chronic obstructive pulmonary disease(P=0.015)and length of stay in intensive care unit(P=0.016),there was no significant difference in other clinical indicators between the two groups(all P＞0.05).Logistic regression analysis showed that coronary heart disease(P=0.028),heart failure(P＜0.001),in-creased neutrophil count(P=0.001),decreased lymphocyte count(P=0.036),increased creatine kinase isoenzyme(P＜0.001),and increased blood urea nitrogen(P=0.042)were independent risk factors for SC.The AUC of the nomogram prediction model in the training set and validation set was 0.759(95％CI=0.732-0.785)and 0.765(95％CI=0.723-0.807),respectively.The established model showcased good fitting degrees in both data sets(training set:P=0.075;validation set:P=0.067).The DC A results showed that the nomogram prediction model had good clinical applicability.Conclusion The nomogram prediction model based on basic diseases and clinical biochemical indicators can effectively predict the risk of SC occurrence.
