Development and validation of the LateDem-Risk score to predict dementia incidence in the InveCe.Ab and Trelong Italian cohorts

Identifying older adults at increased dementia risk is a public health priority. Existing scores often emphasize midlife factors or variables of limited utility in late life. We developed a late-life dementia risk score using data from the InveCe.Ab cohort, including 1100 dementia-free individuals aged 70–74 and followed for 12 years. Candidate risk and protective factors were selected based on systematic reviews and expert consensus. Stepwise Cox models were used to construct a weighted score with and without adjustment for non-modifiable factors. External validation was performed in the TRELONG cohort. The selected modifiable factors were diabetes, alcohol consumption, hypercholesterolemia, social participation, and cognitive activity, with strong predictive accuracy for dementia incidence (HR = 1.23; 95% CI 1.14–1.33; p < 0.026), also after adjustment for non-modifiable factors. In the TRELONG cohort, the score remained significant (HR = 1.19; 95% CI 1.05–1.10; p = 0.008), though its predictive value decreased when adjusting for non-modifiable factors. The LateDem-Risk score offers a practical tool for risk stratification in adults aged ≥ 70 years. Further studies are needed to evaluate its predictive performance in different settings and populations.