Establishment of a prediction model for 28-day mortality rate of sepsis patients based on MIMIC database and its validation
10.11816/cn.ni.2025-246952
- VernacularTitle:基于MIMIC数据库的脓毒症患者28天死亡率预测模型的构建与验证
- Author:
Wanting ZENG
1
;
Jichuan CHEN
;
Li DENG
;
Junyu ZHU
;
Huaping LIANG
Author Information
1. 陆军军医大学大坪医院耳鼻咽喉科,重庆 400042
- Publication Type:Journal Article
- Keywords:
Sepsis;
Medical information mart for intensive care database;
28-day mortality rate;
Prediction model;
Nomogram
- From:
Chinese Journal of Nosocomiology
2025;35(17):2624-2629
- CountryChina
- Language:Chinese
-
Abstract:
OBJECTIVE To explore the indexes affecting the prognosis of the ICU patients with sepsis,establish the nomogram model for prediction of 28-day mortality rate,and validate it.METHODS On the basis of criteria for diagnosis of sepsis(3.0 version),the related data of the sepsis patients were extracted from Medical Information Mart for Intensive Care-Ⅳ(MIMIC-Ⅳ)database for retrospective study and were randomly divided into the train-ing set and the validation set in a 7∶3 ratio.Univariate analysis and multivariate Cox proportional risk regression analysis were performed for the screening of influencing factors for prediction of 28-day mortality of the sepsis pa-tients.The nomogram prediction model was established based on the screened factors and was validated by the val-idation set,and the effect was evaluated.RESULTS A total of 7 955 sepsis patients were included.Seven variables were selected to establish the nomogram model for prediction of the 28-day mortality rate.The nomogram showed favorable performance in identification between the two cohorts,the area under receiver operating characteristic curves(AUROCs)were 0.748 and 0.721,respectively.Calibration curves and Hosmer-Lemeshow test(x2=8.689,10.614;P=0.369,0.225)indicated that the model had remarkable effect on correction.With the per-formance of decision curve analysis(DCA)and clinical impact curve(CIC),the model was considered to have high clinical application value.CONCLUSIONS The nomogram model shows favorable performance in prediction of the 28-day mortality rate and calibration capability.It is worthy to be further promoted and applied.
