Predictors of dementia, the construction of classification trees

In order to identify persons who are at risk for dementia in an early phase, two classification trees were developed. Data were used from the Longitudinal Aging Study Amsterdam (LASA). The prevalence of dementia in the whole sample was 4.0%. In the first tree age seemed to be the strongest predictor, with an increased risk for persons older than 75. In this group the positive predictive value reached a maximum of 33.3% when the persons had memory complaints and a score on the Mini Mental State Examination (MMSE) <24. In a second classification tree, age was excluded as a predictor because of high association with the other potential predictors. In this tree functional limitations seemed the strongest predictor. In the group of persons with at least one functional limitations, the positive predictive value reached a maximum of 28.8% when the persons had memory complaints and a score <24 on the MMSE. In persons without memory complaints, persons with cardiovascular diseases or diabetes were at increased risk of dementia. Further research is necessary before these classification trees can be implemented in general health care.