RESUMO
OBJECTIVE: To characterize clusters of neuropsychiatric symptoms targeted for tracking the disease course in people with dementia, in relation to stage. METHODS: Baseline symptoms from 2922 subjects from two datasets (one clinic based, one online) were aggregated. Common neuropsychiatric symptoms identified by patients/carers as targets of treatment using a dementia SymptomGuide™ were selected. The Global Deterioration Scale was used for clinic staging, and an artificial neural network algorithm, for staging online subjects. Symptom clusters were detected using multiple correspondence analysis and connectivity graph analysis based on relative risk (RR). In a connectivity graph, each pair of nodes (representing symptoms) is connected if their co-occurrence is statistically significant; direction is indicated as positive if RR > 1 and negative otherwise. RESULTS: Neuropsychiatric symptoms were targeted for treatment in 1072 patients (37%). Agitation (37%) and sleep disturbances (28%) were most common symptoms. One cluster (in people with cognitive impairment, no dementia (CIND) or mild dementia) showed significant co-occurrence of anxiety and restlessness; decreased initiative was chiefly seen in isolation. A second cluster (in moderate/severe dementia) was defined by significant co-occurrence of delusions and hallucinations with sleep disturbances; in these subjects, decreased initiative was related to aggression. CONCLUSIONS: Two analytical methods identified neuropsychiatric symptom clusters targeted to track the disease course. In CIND/mild dementia, a profile of decreased initiative distinct from depression suggests possible executive dysfunction. In moderate/severe dementia, targets more reflected psychotic symptoms. Visual data displays allow the relationships between multiple symptoms to be considered simultaneously, which commonly is how they present in patients.