Machine Learning Models to Predict Future Frailty in Community-Dwelling Middle-Aged and Older Adults - ELSA Cohort Study

Aug 5, 2026

Upload by Finn

Machine learning models are evaluated for predicting future frailty in community-dwelling adults, addressing a common epidemiologic challenge: outcome categories are highly imbalanced, which can degrade predictive performance. A retrospective cohort uses participants aged 50+ from the English Longitudinal Study of Ageing who were nonfrail at baseline and reassessed at 4-year follow-up. Models including logistic regression, random forest, support vector machines, neural networks, k-nearest neighbors, and naive Bayes are trained after data balancing, improving performance. Random forest shows the strongest results, and key predictors include age, chair-rise test, household wealth, balance problems, and self-rated health.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Abstract

Background

Methods

Results

Conclusions

Introduction

Frailty and its epidemiology

Limitations of traditional methods

Role of machine learning

Methods overview

Study design and outcome assessment

Model set and balancing strategy

Key findings

The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 2023, XX(XX), 1–9 [https://doi.org/10.1093/gerona/glad127](https://doi.org/10.1093/gerona/glad127)[ ](https://doi.org/10.1093/gerona/glad127)Advance access publication 20 May 2023 Research Article Machine Learning Models to Predict Future Frailty in Community-Dwelling Middle-Aged and Older Adults:The ELSA Cohort Study Daniel Eduardo da Cunha Leme, PhD1,*, and Cesar de Oliveira, PhD2, 1Graduate Program in Gerontology, School of Medical Sciences, University of Campinas, Campinas, Brazil. 2Department of Epidemiology and Public Health, University College London, London, UK. *Address correspondence to: Cesar de Oliveira, PhD. E-mail: [c.oliveira@ucl.ac.uk](c.oliveira@ucl.ac.uk)[ ](c.oliveira@ucl.ac.uk)Decision Editor: Lewis A. Lipsitz, MD, FGSA Abstract Background: Machine learning (ML) models can be used to predict future frailty in the community setting. However, outcome variables for epidemiologic data sets such as frailty usually have an imbalance between categories, that is, there are far fewer individuals classified as frail than as nonfrail, adversely affecting the performance of ML models when predicting the syndrome. Methods: A retrospective cohort study with participants (50 years or older) from the English Longitudinal Study of Ageing who were nonfrail at baseline (2008–2009) and reassessed for the frailty phenotype at 4-year follow-up (2012–2013) . Social, clinical, and psychosocial baseline predictors were selected to predict frailty at follow-up in ML models (Logistic Regression, Random Forest [RF], Support Vector Machine, Neural Network, K-nearest neighbor, and Naive Bayes classifier) . Results: Of all the 4 378 nonfrail participants at baseline, 347 became frail at follow-up. The proposed combined oversampling and undersampling method to adjust imbalanced data improved the performance of the models, and RF had the best performance, with areas under the receiver-operating characteristic curve and the precision-recall curve of 0.92 and 0.97, respectively, specificity of 0.83, sensitivity of 0.88, and balanced accuracy of 85. 5% for balanced data. Age, chair-rise test, household wealth, balance problems, and self-rated health were the most important frailty predictors in most of the models trained with balanced data. Conclusions: ML proved useful in identifying individuals who became frail over time, and this result was made possible by balancing the data set. This study highlighted factors that may be useful in the early detection of frailty. Keywords: Artificial intelligence, Frailty, Outcome, Risk factors Frailty is a widely studied geriatric syndrome. It is characterized by social, clinical, and psychosocial stressors, and is associated with falls, hospitalization, institutionalization, disability, and mortality among older adults. It is also associated with high costs for health systems globally (1) . In a study of community-dwelling middle-aged and older adults in European countries, the prevalence of the frailty phenotype was 4.1% and 17.0%, respectively (2), and in a systematic review with meta-analysis, the global incidence of frailty was 13.6% in a mean of 3 years of follow-up (3) . Given its clinical importance, many studies around the world have investigated the determinants of this syndrome. However, they employed traditional statistical methods such as regression analysis (4) . Network analysis, which is a more sophisticated statistical method, has also been used recently to determine the complex relationships between different factors and the frailty phenotype (5). However, the aforementioned statistical methods were used exclusively to estimate the effect of a set of independent or random variables on the variable of interest, that is, frailty. Predictive models, such as those based on machine learning (ML), are intended primarily to make predictions that are as accurate as possible by lear

FAQs

Why is data imbalance a problem when predicting frailty with machine learning?

Frailty outcomes are usually far less common than nonfrail outcomes, creating an imbalanced category distribution. This imbalance can reduce model performance when predicting frailty.

What cohort and follow-up design are used to predict future frailty?

The study uses a retrospective cohort from the English Longitudinal Study of Ageing. Participants aged 50+ who were nonfrail at baseline (2008–2009) were reassessed at 4-year follow-up (2012–2013).

Which machine learning model performed best and what were the main results?

Random forest achieved the best performance. After applying combined oversampling and undersampling, it reported strong discrimination with AUC values of 0.92 (ROC) and 0.97 (precision-recall), and balanced accuracy of 85.5% on balanced data.

Aug 5, 2026

Upload by Finn