Constructing a fall risk prediction model for hospitalized patients using machine learning

Area of Science:

• Healthcare Informatics
• Clinical Risk Management
• Machine Learning in Medicine

Background:

• Hospitalized patients face significant fall risks, leading to adverse outcomes.
• Accurate prediction of fall risk is crucial for effective prevention strategies.

Purpose of the Study:

• To identify risk factors for falls in hospitalized patients.
• To develop and validate a machine learning-based predictive model for fall risk.
• To evaluate the performance of various machine learning algorithms for fall prediction.

Main Methods:

• Utilized a cross-sectional design with data from the Fukushima Medical University Hospital Cohort Study (DRYAD database).
• Employed Synthetic Minority Oversampling Technique combined with Edited Nearest Neighbors (SMOTE-ENN) for data balancing.
• Applied univariate analysis, LASSO regression, and eight machine learning algorithms, including Random Forest, with SHAP for interpretability.

Main Results:

• The Random Forest model demonstrated strong predictive performance with an AUC of 0.795 in the test set.
• Key predictors identified include ADL (standing, evacuation), age group, planned surgery, wheelchair use, history of falls, hypnotic drugs, psychotropic drugs, and remote caring system.
• SHAP analysis provided insights into the importance of these risk factors.

Conclusions:

• Machine learning, specifically Random Forest, is effective for predicting fall risk in hospitalized patients.
• The developed model and identified risk factors can significantly enhance patient safety and inform fall prevention protocols in healthcare settings.