Since its development, the Qualidem has had items that were considered unsuited for people with very severe dementia. This study attempted to investigate the applicability of all Qualidem items in people with all stages of dementia severity.

Four data sets that contained Qualidem observations on people with dementia were combined. Dementia severity was categorized based on the Global Deterioration Scale (GDS), with a dichotomization of very severe dementia (GDS 7) and others (GDS 1–6). Unidimensional latent-trait models (Mokken scaling) were estimated to fit the Qualidem responses in the overall sample and the dichotomized groups. Scalability was assessed using coefficients of homogeneity (Loevinger's H), while reliability was assessed with Cronbach's α and ρ.

Combining the four databases resulted in 4,354 Qualidem measurements. The scalability of all scales was considered acceptable in the overall sample, as well is in the subgroups (all H > 0.3). Additionally, the reliability was good–excellent in the scales: “positive affect,” “positive self-image,” “care relationship,” and “negative affect.” Reliability was questionable–acceptable for “feeling at home,” “social relations,” “social isolation,” and “restless tense behavior.” Reliability was poor for “having something to do.”

Statistical considerations allow using all Qualidem items in all dementia stages. Future research should determine balance of statistical- versus conceptual-based reasoning in this academic debate.