“…There are a number of limitations to this study: (1) the detected differences between SCUs could in theory be due to non-random sampling of the SCUs in the present study. However, most (51/56) SCUs included in this study were ordinary SCUs with a mixture of patients in various stages of the dementia and did not participate for having patients with remarkably low or high symptoms; (2) we did not categorize the dementia subtype, which is known to be a predictor of neuropsychiatric symptoms (Ballard et al, 1995;Lopez et al, 1996;Lyketsos et al, 2000;Bathgate et al, 2001;Lyketsos et al, 2002), although there is only limited evidence of this in nursing home patients (Drachman et al, 1992;Suh, 2004). Also other relevant patient correlates that we did not measure, such as cognitive or global functioning, pre-onset personality, co-morbidity, pain etc.…”