Construction and evaluation of a medium-and long-term prognosis model for severe community-acquired pneumonia based on MIMIC-Ⅳ database
10.11855/j.issn.0577-7402.1759.2025.0305
- VernacularTitle:基于MIMIC-Ⅳ数据库的重症社区获得性肺炎中长期预后模型的构建与评估
- Author:
Nan-Li DENG
1
;
Ren-Huai LIU
;
Xin CHAI
;
Xi-Jing ZHANG
;
Bin-Xiao SU
Author Information
1. 空军军医大学西京医院重症医学科,陕西 西安 710032;空军军医大学西京医院麻醉与围术期医学科,陕西 西安 710032
- Keywords:
severe community-acquired pneumonia;
MIMIC-Ⅳ database;
prognostic model;
LASSO regression
- From:
Medical Journal of Chinese People's Liberation Army
2025;50(4):400-408
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the risk factors for medium-and long-term mortality in patients with severe community-acquired pneumonia(SCAP)based on the Medical Information Mart for Intensive Care Ⅳ(MIMIC-Ⅳ),construct a prognostic model and evaluate its predictive efficacy.Methods In this retrospective cohort study,1943 SCAP patients from the U.S.MIMIC-Ⅳdatabase(2008-2019)were randomly divided into training(n=1363)and validation(n=580)sets(7:3 ratio).Primary and secondary endpoints were 1-year and 30-/90-day all-cause mortality,respectively.Prognostic factors were selected using LASSO regression and multivariable Cox proportional hazards modeling,and a visual nomogram model was built.Model performance was assessed via C-index,receiver operator characteristic(ROC)curves,and calibration curves,and compared with the CURB-65 score.Risk stratification was validated using Kaplan-Meier analysis.Results The 30-day,90-day,and 1-year all-cause mortality rates for SCAP patients were 25.9％,34.5％,and 42.6％,respectively.Seven independent risk factors were identified:age(HR=1.037),heart rate(HR=1.007),red blood cell distribution width(RDW,HR=1.092),Acute Physiology Score Ⅲ(APS-Ⅲ,HR=1.013),cerebrovascular disease(HR=1.453),liver disease(HR=1.272),and malignancy(HR=2.007).Based on these factors,Cox regression model was constructed and nomogram was drawn,C-indices of training set and validation set were 0.710 and 0.688,respectively.For 1-year mortality prediction,the model achieved superior area under the ROC curve(AUC)values(training set:0.768;validation set:0.738)compared with CURB-65 score(training set:0.648;validation set:0.616).Kaplan-Meier survival analysis revealed significantly worse survival in high-risk group than low-risk group(P<0.0001).Conclusions Age,heart rate,RDW,APS-Ⅲ,cerebrovascular disease,liver disease,and malignant tumor were medium-and long-term mortality risk factors in SCAP patients.The prognostic model constructed based on these factors has high predictive power and provides an important clinical diagnosis and treatment reference.
