Decision tree and random forest classification models were used to find associations between patient demographic and lifestyle risk factors with dementia status during more than 30 years of follow-up.