Background: The Neuropsychiatric Inventory (NPI) is widely used to assess psychopathology in dementia. The scoring involves ratings of frequency and severity, as well as the calculation of a composite score. It was suggested recently that, due to lower variance, the frequency score might be more sensitive to detect treatment-related change and to discriminate active treatment from placebo than the composite score, particularly in milder forms of the disease.
Methods: Based on data from three randomized controlled trials in patients with mild to moderate dementia, standardized changes were calculated for both frequency and composite scores for two strata of disease severity. The two strata were formed by dichotomizing the sample along the median score of the short cognitive performance test (SKT) battery.
Results: Across all studies and for both severity strata, standardized changes in frequency scores were not consistently larger than those in composite scores and both scores discriminated active treatment from placebo at similar probabilities for type-1 error.
Conclusion: Our findings do not support the notion that there is a difference between frequency score and composite score with respect to their sensitivity to treatment-related change.