Predictive Risk Models for Frailty Onset in Older Adults: A Scoping Review of Methodological Trends, Model Performance, and Clinical Translation Gap

Purpose: Identifying older adults at risk of frailty is crucial for early intervention. Although numerous prediction models have emerged, no scoping review has systematically mapped their methodological trends and barriers to clinical implementation. This scoping review aimed to examine methodological characteristics, model performance, and translational gaps in frailty onset prediction models for older adults. Methods: A systematic search of six Chinese and English databases was conducted from inception to February 2026following Arksey and O'Malley's framework and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews guidelines. Results: Thirty studies, reporting 31 frailty prediction models published between 2018 and 2026, were included. Six overarching trends emerged: Most studies originated from China (73.33%), and community-dwelling older adults were the predominant study population (73.33%). After 2024, longitudinal designs using larger public databases became more common; Machine learning was increasingly adopted (38.7%) but showed no clear advantage over logistic regression (median AUC 0.813 vs 0.860); Conventional predictors, including age, multimorbidity, and depression, remained dominant; Calibration was under-reported, particularly in machine learning models (50%); Clinical translation was limited, with external validation and Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis plus Artificial Intelligence (TRIPOD+AI) adherence each reported in only 22.6% of models, static presentation formats predominating (83.9%), and no studies evaluating health economic outcomes or prospective clinical impact. Conclusion: Frailty prediction models are constrained by insufficient calibration reporting, limited external validation, and substantial translation barriers. Future research should prioritize rigorous external validation, TRIPOD+AI adherence, and clinically integrated digital tools supported by implementation science.