Machine Learning Prediction for Postdischarge Falls in Older Adults: A Multicenter Prospective Study.

Area of Science:

• Geriatric Medicine
• Artificial Intelligence in Healthcare
• Patient Safety

Background:

• Falls after hospital discharge are a significant concern for older adults.
• Existing fall prediction methods can be complex and burdensome.
• There is a need for efficient tools to identify high-risk individuals for targeted interventions.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting early postdischarge falls in older adults.
• To utilize easily collectable data from acute care settings.
• To reduce the burden of complex assessment tools.

Main Methods:

• Prospective multicenter study in Japanese geriatric wards (Oct 2019-Jul 2023).
• Included 706 participants aged ≥65 years.
• Extracted 19 variables; developed ML models (Extra Trees, Naive Bayes, AdaBoost, Random Forest); evaluated using 5-fold cross-validation and AUC.

Main Results:

• 16.1% of patients experienced a fall within 3 months postdischarge.
• The Extra Trees classifier achieved the highest predictive performance (AUC=0.73).
• Key predictors included Lawton IADL, Clinical Frailty Scale, urinary incontinence, Geriatric Depression Scale, and preadmission residence.

Conclusions:

• This is the first ML model to predict early postdischarge falls in older acute care patients.
• The model shows potential for assisting in fall prevention strategies.
• It supports a smoother transition of care from hospital to community settings.