Objectives: This study examined the construct validity of two cognitive scales from the federally mandated Minimum Data Set (MDS) of the nursing home Resident Assessment Instrument.
Design: A cross-sectional comparisons of the MDS measures, with scales provided by the resident, a proxy person, and nursing staff.
Setting: Subjects residing in 59 nursing homes (NHs) in Maryland from 1992 to 1995.
Participants: Subjects were 1939 new admissions to NHs, aged 65 and older, with complete MDS information at admission.
Measurements: Two MDS scales, the Cognitive Performance Scale (CPS) and the MDS Cognition Scale (MDS-COGS), were compared with the Mini-Mental State Examination (MMSE) and the staff rating on the Psychogeriatric Dependency Rating Scale (PGDRS) Orientation scale, as well as measures of functioning and functional decline.
Results: The CPS and the MDS-COGS were highly correlated (r = 0.92). Both correlated moderately well with the MMSE (r = -0.65 and -0.68) and with staff's rating on the PGDRS Orientation scale (r = 0.63 and r = 0.66). Correlations with the MMSE (r < 0.70) are lower than previously reported (r > or = 0.80). The proportion of cognitively impaired residents in this NH admission cohort was higher using the MDS-COGS than the CPS (65% vs 57%), but both MDS scales produced lower proportions than the MMSE (70%) and higher proportions than the PGDRS (47%). The internal consistency of the CPS was better without the comatose item (alpha = 0.80 vs 0.70). The MDS-COGS had higher internal consistency (alpha = 0.85) and was simpler to compute.
Conclusions: This is the first study to examine the validity of the MDS in a large sample of residents and NHs in situations where the MDS was not completed by research-trained staff. Compared with other instruments, the MDS-COGS and the CPS had moderate and similar validity for assessing cognitive impairment. Differences in the scales could provide different estimates of impairment among persons admitted to nursing homes.