Stratifying dementia risk factors: A prediction model and hypothesis-driven analysis

Most older adults present multimorbidity, but dementia risk factors are typically analyzed individually. Direct methodological comparisons evaluating simultaneous multiple risk factors are essential to provide the realistic effects of multimorbidity. We aimed to compare hypothesis- and data-driven approaches for dementia risk stratification in a real-world cohort.

We analyzed 9606 participants from the National Alzheimer's Coordinating Center (NACC) Uniform Data Set (2005-2023) using machine learning with interpretability analysis and survival models to simultaneously evaluate 13 risk factors for incident dementia conversion.

A total of 877 participants (9%) developed dementia over (mean ± SD, 6 ± 4.2) years of follow-up. Both approaches consistently identified four key predictors: age, depression, low education, and body mass index (protective). Convergent findings across methodologies demonstrated robust factor identification despite different analytical paradigms.

In this direct methodological comparison, age, depression, and low education emerge as major dementia risk factors regardless of analytical approach. Convergent interpretability of these approaches support simultaneous multifactorial risk assessment in clinical practice.

Data- and hypothesis-driven approaches identified convergent key risk factors Age, depression, and low education are major risk factors for dementia Higher body mass index was unexpectedly protective against dementia conversion Multimorbidity requires simultaneous evaluation of multiple risk factors Real-world analysis reveals complex interactions between dementia risks.