JMIR Publications

ABSTRACT

Background:

Falls among hospitalized patients are a critical issue which often leads to prolonged hospital stays and increased healthcare costs. Traditional fall risk assessments typically rely on standardized scoring systems; however, these may fail to capture the complex and multifactorial nature of fall risk factors.

Objective:

this multicenter cohort study aimed to develop and validate a machine learning-based model to predict in-hospital falls and to evaluate its performance in terms of both discrimination and calibration.

Methods:

We analyzed the data of 83,917 inpatients aged ≥ 65 years with a hospital stay of at least three days. Using Diagnosis Procedure Combination data and laboratory results, we extracted demographic, clinical, functional, and pharmacological variables. Following the selection of 30 key features, four predictive models were constructed: logistic regression, extreme gradient boosting, light gradient boosting machine (LGBM), and categorical boosting (CatBoost). The synthetic minority oversampling technique and isotonic regression calibration were applied to improve the prediction quality and address class imbalance.

Results:

Among the models, CatBoost demonstrated the highest F1 score and PRAUC, whereas LGBM exhibited the best calibration, closely approximating ideal performance. In addition, SHAP analysis identified frailty-related factors (e.g., albumin level and mobility) as well as the use of sedatives, hypnotics, and diabetes medications as significant factors that contribute to fall risk. Detailed incident report analyses further indicated that toileting-related falls were particularly common, especially during late-night and early-morning near the bedside.

Conclusions:

These findings highlight the value of integrating machine-learning models with comprehensive patient assessments to enhance fall risk prediction. By focusing on individualized patient conditions and implementing targeted strategies for safe toileting and mobility, more precise and effective fall prevention measures can be developed for hospitalized patients. Clinical Trial: Not applicable.