Machine learning in geriatric care: a scoping review of models using multidimensional assessment data.

Area of Science:

• Geriatric Medicine
• Artificial Intelligence in Healthcare
• Data Science

Background:

• Geriatric assessments gather comprehensive data on older adults' physical, cognitive, psychological, and social health.
• Machine learning (ML) offers potential for improved clinical decision-making in aged care using this multidimensional data.
• A systematic synthesis of ML applications in geriatric assessment data is lacking.

Purpose of the Study:

• To systematically review and describe the data types, purposes, and performance of ML models applied to multidimensional geriatric assessment data.
• To identify common ML algorithms and their effectiveness in predicting health outcomes for older adults.

Main Methods:

• A scoping review was conducted following established frameworks (Arksey and O'Malley, PRISMA-ScR).
• Searched six databases for peer-reviewed studies from 2012-2024 on ML applied to geriatric assessment data.
• Assessed methodological quality using the PROBAST + AI tool.

Main Results:

• Forty studies were included, primarily from high-income countries, focusing on outcomes like falls, functional decline, frailty, mortality, and hospital readmission.
• XGBoost was the top-performing algorithm in several studies, but performance metrics varied widely.
• Common limitations included small sample sizes, lack of external validation, poor calibration, and data imbalance.

Conclusions:

• ML models utilizing geriatric assessment data demonstrate potential for predicting health outcomes in older adults.
• Methodological and reporting limitations currently hinder the clinical translation of these ML models.
• Future research must prioritize external validation, interpretability, and integration into clinical workflows for robust and ethical application in aged care.