Machine Learning Prediction of In-Hospital Mortality in Elderly ICU Patients with COPD and Heart Failure: A Dual-Database Study

Abstract Background. Elderly patients with coexisting chronic obstructive pulmonary disease (COPD) and heart failure (HF) represent a high-risk ICU population with substantial in-hospital mortality. Traditional scoring systems rely on static physiologic thresholds and may not fully capture the complex, nonlinear interactions among demographic, laboratory, inflammatory, nutritional, and comorbidity factors. We aimed to develop and externally validate a machine learning (ML) model for predicting in-hospital mortality in this population. Methods. This retrospective cohort study used the Medical Information Mart for Intensive Care IV (MIMIC-IV, v3.1) as the training cohort and the eICU Collaborative Research Database (eICU v2.0) as the external validation cohort. Elderly patients (≥ 65 years) with concurrent COPD and HF admitted to the ICU were included (MIMIC-IV: n = 2,569, 18.4% mortality; eICU: n = 2,132, 13.8% mortality). Thirty-three candidate variables encompassing demographics, laboratory values, inflammatory and nutritional indices, comorbidities, and illness severity were harmonized across databases. Missing values were imputed using Multiple Imputation by Chained Equations (MICE). Feature selection comprised variance inflation factor (VIF) filtering (threshold = 5), followed by Least Absolute Shrinkage and Selection Operator (LASSO) regression and Boruta algorithm. Eight ML models were trained with hyperparameter optimization and compared. The best-performing model was evaluated on the internal test set and the external eICU cohort. Model performance was assessed by area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) values were used for model interpretability. Results. VIF filtering removed 8 collinear variables (retaining 25), and LASSO selected 11 features: sepsis, age, red cell distribution width, albumin, lactate, monocyte count, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, monocyte-to-lymphocyte ratio, neutrophil-to-albumin ratio, and illness severity score. The Random Forest model achieved the highest performance with an internal test AUC of 0.769 (95% confidence interval [CI]: 0.729–0.811) and an external validation AUC of 0.723 (95% CI: 0.694–0.754). Brier scores were 0.129 and 0.114, respectively. Calibration curves showed good agreement between predicted and observed probabilities. DCA demonstrated positive net benefit across clinically relevant threshold probabilities. SHAP analysis identified sepsis, illness severity, lactate, and neutrophil-to-lymphocyte ratio as the most influential predictors. Conclusions. A Random Forest model integrating routinely available clinical variables demonstrated good discrimination and calibration for predicting in-hospital mortality in elderly ICU patients with COPD and HF, with robust external validation. The combination of inflammatory, nutritional, and physiologic markers provides complementary prognostic information. This tool may assist clinicians in early risk stratification and clinical decision-making for this high-risk population.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11158956/v1
Primary Topic
Machine Learning in Healthcare
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Machine Learning Prediction of In-Hospital Mortality in Elderly ICU Patients with COPD and Heart Failure: A Dual-Database Study

Ruhao Wu, Shuaixian Han, Jianping Liu, Jiayi Lin et al.
7 links
Research Square
Machine Learning in Healthcare
preprint

Machine Learning Prediction of In-Hospital Mortality in Elderly ICU Patients with COPD and Heart Failure: A Dual-Database Study

Ruhao Wu, Shuaixian Han, Jianping Liu, Jiayi Lin, Nan Xie
preprint en

Abstract

Abstract Background. Elderly patients with coexisting chronic obstructive pulmonary disease (COPD) and heart failure (HF) represent a high-risk ICU population with substantial in-hospital mortality. Traditional scoring systems rely on static physiologic thresholds and may not fully capture the complex, nonlinear interactions among demographic, laboratory, inflammatory, nutritional, and comorbidity factors. We aimed to develop and externally validate a machine learning (ML) model for predicting in-hospital mortality in this population. Methods. This retrospective cohort study used the Medical Information Mart for Intensive Care IV (MIMIC-IV, v3.1) as the training cohort and the eICU Collaborative Research Database (eICU v2.0) as the external validation cohort. Elderly patients (≥ 65 years) with concurrent COPD and HF admitted to the ICU were included (MIMIC-IV: n = 2,569, 18.4% mortality; eICU: n = 2,132, 13.8% mortality). Thirty-three candidate variables encompassing demographics, laboratory values, inflammatory and nutritional indices, comorbidities, and illness severity were harmonized across databases. Missing values were imputed using Multiple Imputation by Chained Equations (MICE). Feature selection comprised variance inflation factor (VIF) filtering (threshold = 5), followed by Least Absolute Shrinkage and Selection Operator (LASSO) regression and Boruta algorithm. Eight ML models were trained with hyperparameter optimization and compared. The best-performing model was evaluated on the internal test set and the external eICU cohort. Model performance was assessed by area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) values were used for model interpretability. Results. VIF filtering removed 8 collinear variables (retaining 25), and LASSO selected 11 features: sepsis, age, red cell distribution width, albumin, lactate, monocyte count, neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, monocyte-to-lymphocyte ratio, neutrophil-to-albumin ratio, and illness severity score. The Random Forest model achieved the highest performance with an internal test AUC of 0.769 (95% confidence interval [CI]: 0.729–0.811) and an external validation AUC of 0.723 (95% CI: 0.694–0.754). Brier scores were 0.129 and 0.114, respectively. Calibration curves showed good agreement between predicted and observed probabilities. DCA demonstrated positive net benefit across clinically relevant threshold probabilities. SHAP analysis identified sepsis, illness severity, lactate, and neutrophil-to-lymphocyte ratio as the most influential predictors. Conclusions. A Random Forest model integrating routinely available clinical variables demonstrated good discrimination and calibration for predicting in-hospital mortality in elderly ICU patients with COPD and HF, with robust external validation. The combination of inflammatory, nutritional, and physiologic markers provides complementary prognostic information. This tool may assist clinicians in early risk stratification and clinical decision-making for this high-risk population.

Research Square
First Affiliated Hospital of Zhengzhou University (CN)
Good health and well-being
Machine Learning in Healthcare
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon