Frailty Modeling Using Machine Learning Methodologies: A Systematic Review With Discussions on Outstanding Questions.

Area of Science:

• Gerontology
• Computational Health
• Biostatistics

Background:

• Frailty significantly impacts older adults' health and quality of life.
• Current frailty modeling often uses suboptimal, simple analytical techniques.
• There is a lack of large-scale systematic reviews on machine learning applications in frailty.

Purpose of the Study:

• To explore machine learning methods for predicting or classifying frailty in older persons.
• To systematically review existing literature on machine learning in frailty modeling.

Main Methods:

• Systematic review of 181 research articles.
• Categorization of analytical methods into generalized linear models, survival models, and non-linear models.
• Analysis of predictor variables and predicted outcomes.

Main Results:

• Reviewed methods showed moderate agreement with existing frailty scores and predictive validity for adverse outcomes.
• Non-linear methods showed potential to outperform generalized linear methods.
• Key predictors include diagnoses, functional performance, and cognition; outcomes include mortality and hospital admissions.

Conclusions:

• Machine learning offers potential for improved frailty prediction and classification.
• Classical methods and cross-sectional data are common, but longitudinal data and advanced ML methods are emerging.
• Future research should focus on advanced machine learning with high-dimensional longitudinal data for personalized frailty tools.