Predicting Risk of Dementia with Survival Machine Learning and Statistical Methods - Results on the English Longitudinal Study of Ageing Cohort

Sep 17, 2026

Uploaded by Miles

Machine learning approaches to predict dementia onset often use classification that ignores event timing. This study introduces a survival-analysis framework within machine learning, extending Random Forest and Elastic Net to model time-to-dementia risk using the English Longitudinal Study of Ageing (ELSA) cohort. The two survival machine learning models are compared with the conventional Cox proportional hazards model. Results indicate superior predictive capability and stability, supported by nested cross-validation and Monte Carlo validation, and discuss implications and future research directions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Literature review

Predicting Risk of Dementia with Survival Machine Learning and Statistical Methods: Results on the English Longitudinal Study of Ageing Cohort Daniel Stamate1,2, Henry Musto1,*, Olesya Ajnakina3,4, Daniel Stahl3 1. Data Science & Soft Computing Lab, Computing Department, Goldsmiths College, University of London, United Kingdom * Joint first-author 2. Division of Population Health, Health Services Research & Primary Care, School of Health Sciences, University of Manchester, United Kingdom 3. Institute of Psychiatry Psychology and Neuroscience, Biostatistics and Health Informatics Department, King’s College London, United Kingdom 4. Department of Behavioural Science and Health, Institute of Epidemiology and Health Care, University College London, United Kingdom Abstract. Machine learning models that aim to predict dementia onset usually follow the classification methodology ignoring the time until an event happens. This study presents an alternative, using survival analysis within the context of machine learning techniques. Two survival method extensions based on machine learning algorithms of Random Forest and Elastic Net are applied to train, optimise, and validate predictive models based on the English Longitudinal Study of Ageing – ELSA cohort. The two survival machine learning models are compared with the conventional statistical Cox proportional hazard model, proving their superior predictive capability and stability on the ELSA data, as demonstrated by computationally intensive procedures such as nested cross-validation and Monte Carlo validation. This study is the first to apply survival machine learning to the ELSA data, and demonstrates in this case the superiority of AI based predictive modelling approaches over the widely employed Cox statistical approach in survival analysis. Implications, methodological considerations, and future research directions are discussed. Keywords: Predicting risk of dementia, Survival machine learning, Survival random forests, Survival elastic net, Cox proportional hazard, Nested crossvalidation, Monte Carlo validation 1 Introduction Dementia, of which approximately two-thirds constitute Alzheimer's Disease (AD) cases [1], is associated with a progressive decline of brain functioning, leading to a significant loss of autonomy, reduced quality of life and a shortened life expectancy 2 [2] . Accumulated evidence indicates that individuals who have dementia have an excess mortality [3] and a shorter life expectancy [4] than individuals without this disease [5] . In England, dementia is now reported as being the leading cause of death for women, having overtaken cancer and cardiovascular disease [6] . The development of prognostic prediction models, built on combined effects of thoroughly validated predictors, using Machine Learning tools, can be used to forecast the probability of dementia developing within an individual. It is hoped that the availability of such prediction models will facilitate more rapid identification of individuals who are at a higher risk of dementia before the full illness onset [10] . This, in turn, would reduce time to treatment initiation, subsequently minimising the social and functional disability and thereby improving the quality of life for many people affected by these disorders. Identifying individuals at risk of developing dementia would allow the recruiting of patients at high risk for future clinical trials, thereby catalysing the assessment of new treatment or prevention programmes. Furthermore, identifying modifiable risk factors would allow the development of new prevention programmes. For example, there are already some indications that being physically active, staying mentally and socially active, and controlling high blood pressure can potentially deter onset of dementia in the general population [7] . 2 Literature review To date, several papers have been published which seek to predict, in a binomial or multinomia

FAQs

Why is survival analysis important for predicting dementia risk?

Most machine learning models predict dementia onset using classification and disregard the time until the event occurs. Survival analysis models the time-to-event component, enabling clinically relevant information about when diagnosis may occur.

Which survival machine learning methods are evaluated in the study?

The study extends survival-analysis techniques based on Random Forest and Elastic Net machine learning algorithms to train, optimize, and validate predictive models on the ELSA cohort.

How do the survival machine learning models compare with the Cox proportional hazards model?

The survival machine learning models demonstrate superior predictive capability and greater stability on ELSA data. The comparison is supported by computationally intensive validation procedures such as nested cross-validation and Monte Carlo validation.

Sep 17, 2026

Uploaded by Miles