Age-only versus multivariable models for dementia prediction: a comparative analysis

Purpose of review
Accurate dementia risk prediction is critical for prevention, yet it remains unclear which predictors add meaningful value beyond chronological age. This review evaluates the extent to which multivariable dementia risk models identify modifiable risk factors that enhance prediction value.

Recent findings
We systematically reviewed cohort studies reporting both age-only and multivariable dementia prediction models in the same populations. Six age-only models across five cohorts were included. Age-only models achieved poor to good discrimination (C-statistics 0.66–0.84). Adding modifiable cardiovascular and lifestyle factors provided consistent, modest improvements of 0.02–0.05 in the UK Biobank, Atherosclerosis Risk in Communities (ARIC), and Rotterdam cohorts. Larger improvements of 0.07–0.12 were observed in models including cognitive testing or genetic factors [e.g., UK Biobank Dementia Risk Score (UKBDRS-APOE)] with the Hanley–McNeil z-test confirming the improvements were significant, indicating genuine improvement rather than random variation.

Summary
While age is a significant risk factor for dementia, modifiable cardiovascular and lifestyle factors provide incremental predictive value beyond age and represent actionable targets for prevention. Despite modest statistical improvements, these factors offer the most clinically relevant targets for prevention strategies. Future efforts should prioritise interventions addressing these modifiable determinants to reduce dementia risk across populations.