Machine Learning Identifies Variances Predicting PLOS
Researchers at Kyushu University Hospital used ePath data from 480 lung cancer patients undergoing video-assisted thoracoscopic surgery between 2019 and 2023 to train machine learning models predicting prolonged length of stay (PLOS >9 days). Ridge regression performed best (AUC 0.84 derivation, 0.82 test; Brier scores 0.16 and 0.17) and counterfactual analysis highlighted six key variances including postoperative fever, arrhythmia, impaired ambulation, and pulmonary air leaks.
Key Points
- 1Develops ridge-regression ML model on 480 VATS lung cancer patients (2019–2023), AUC 0.84/0.82.
- 2Highlights six clinical variances (e.g., fever, arrhythmia, air leaks) strongly associated with prolonged hospital stay.
- 3Enables early targeting of variance-driven interventions to potentially reduce resource use and improve discharge timing.
Scoring Rationale
Useful validated ML model with clinical counterfactuals; limited generalizability from single-center retrospective cohort of 480 patients.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
