Background: The goal of the present study is to develop a classification model for use in primary care using markers which are relatively easy to determine to aid early identification of persons at risk for dementia. 

Methods: Data were used from the Origins of Variance in the Old-Old (OCTO-Twin) study. The baseline sample included 521 non-demented subjects aged 80 and older. Relevant predictors on dementia were collected two years prior to dementia diagnosis. Dementia diagnosis was based on DSM-III-R criteria. Data were analyzed using generalized estimating equations and Cox survival analyses. 

Results: Overall, the two-year incidence of dementia was 6.9%. Reporting memory complaints was the strongest predictor of dementia. Memory complaints and a MMSE score ≤ 25 resulted in a predictive value for dementia of 28.8%. No memory complaints, drinking no alcohol and a MMSE score ≤ 24 resulted in a predictive value of 18.0%. Reporting no memory complaints, drinking alcohol, having functional limitations and a MMSE score ≤ 24 resulted in a percentage of 24.7% identified with dementia after two years of follow-up. 

Conclusions: The developed classification tree could contribute to early identification of persons at risk for dementia in primary care in a feasible and cost-effective way.