Construction of a prediction model for severe pneumonia complicate with respiratory failure
10.12025/j.issn.1008-6358.2025.20250102
- VernacularTitle:重症肺炎并发呼吸衰竭预测模型的构建
- Author:
Siyu GAO
1
;
Sheng ZHANG
2
;
Xi CHEN
3
,
4
;
Zhixia ZHANG
1
;
Yumei YANG
1
Author Information
1. Department of Respiratory and Critical Care Medicine, Tianyou Hospital Affiliated to Wuhan University of Science and Technology, Wuhan 430064, Hubei, China.
2. Cancer Center, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430022, Hubei, China.
3. Brooks College (Sunnyvale), California 94089, the United States
4. School of Public Health, Zhejiang University, Hangzhou 310000, Zhejiang, China.
- Publication Type:Shortarticle
- Keywords:
severe pneumonia;
respiratory failure;
clinical predictive model;
internal validation;
LASSO regression
- From:
Chinese Journal of Clinical Medicine
2025;32(3):449-457
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore predictive factors of severe community-acquired pneumonia (CAP) complicated with respiratory failure (RF) and to develop and internally validate a clinical prediction model. Methods A retrospective study was conducted on 350 patients with severe CAP admitted to Tianyou Hospital Affiliated to Wuhan University of Science and Technology from September 2022 to December 2024. Patients were randomly divided into a training set (n＝245) and a validation set (n＝105) in a 7∶3 ratio, and further categorized into RF and non-RF groups. LASSO regression was applied to optimize variable selection. Multivariate logistic analysis was used to construct the prediction model, followed by internal validation. Results Univariate regression analysis identified male, hypertension, diabetes, coronary heart disease, age, CURB-65 score, white blood cell count, neutrophil count, C-reactive protein (CRP), serum amyloid A, procalcitonin, and hospital stay as risk factors for RF in severe CAP, while albumin level was a protective factor. LASSO regression selected CURB-65 score, albumin level, and CRP for inclusion in the final model. The area under the receiver operating characteristic curve was 0.903 in the training set and 0.919 in the validation set. Calibration curve analysis demonstrated excellent agreement between predicted and observed probabilities in both sets, and Hosmer-Lemeshow goodness-of-fit tests indicated no significant deviations. Threshold probabilities ranged from 0.01 to 0.99 in both training and validation sets. Conclusions CURB-65 score, albumin level, and CRP are independent predictors of RF in severe CAP. The clinical prediction model based on these factors exhibits strong discrimination, calibration, goodness-of-fit, and clinical utility.
