Related Experiment Videos
Admission-Based Machine-Learning Models for Predicting Mechanical Ventilation and Mortality in Fibrotic Interstitial
Jibo Sun1,2, Xirui Chen3, Xiangpeng Wang4
1Department of Pulmonary and Critical Care Medicine, West China Hospital, Sichuan University, Chengdu, China.
Balkan Medical Journal
|July 2, 2026
Summary
Machine learning models accurately predict mechanical ventilation and mortality in hospitalized fibrotic interstitial lung disease (F-ILD) patients. Key predictors include lactate dehydrogenase (LDH), neutrophil-to-lymphocyte ratio (NLR), and prognostic nutritional index (PNI).
Area of Science:
- Pulmonology and Respiratory Medicine
- Medical Informatics and Machine Learning
- Critical Care Medicine
Background:
- Existing risk stratification tools for hospitalized fibrotic interstitial lung disease (F-ILD) patients are insufficient.
- Accurate prediction of outcomes like mechanical ventilation and mortality is crucial for managing F-ILD.
- Need for validated, admission-based predictive models in F-ILD.
Purpose of the Study:
- To develop and externally validate machine-learning models for predicting mechanical ventilation (MV) and mortality in hospitalized F-ILD patients.
- To assess the models' ability to predict short-term (30-day, 3-month) and long-term (all-cause, cause-specific) mortality.
- To identify key clinical and laboratory predictors for risk stratification.
Main Methods:
- Multicenter retrospective cohort study involving 1,272 hospitalized adult F-ILD patients.
- Development of machine-learning models using clinical characteristics and lab results within 24 hours of admission.
- Internal testing and independent external validation of models for predicting MV, 30-day, 3-month, and long-term mortality.
Main Results:
- Models showed robust discrimination for MV (AUC, 0.913), 30-day mortality (AUC, 0.926), and 3-month mortality (AUC, 0.813) in external validation.
- Long-term all-cause mortality prediction achieved a C-index of 0.768 (24 months).
- Key predictors included lactate dehydrogenase (LDH), neutrophil-to-lymphocyte ratio (NLR), prognostic nutritional index (PNI), age, sex, smoking history, IPF phenotype, and specific ratios.
Conclusions:
- Admission-based machine-learning models demonstrate strong predictive performance for MV and mortality in hospitalized F-ILD patients.
- Elevated LDH, NLR, and low PNI identify high-risk patients prone to acute deterioration and death.
- Models require local recalibration for optimal implementation in new clinical settings.
