Objective: To compare the rates of hospitalization among cohorts of nursing home residents assembled before and after the implementation of the federally mandated Resident Assessment Instrument (RAI).
Sample: Subjects were nursing home residents chosen from 268 facilities in major Metropolitan Statistics Areas in 10 states and representing more than 1500 facilities and 60,000 residents. Two resident cohorts (1990 and 1993) were sampled (8 to 16 residents per facility, depending upon facility size) as part of an evaluation of the impact of implementing the RAI.
Methods: Research nurses reviewed records, interviewed staff, observed patients, and completed an RAI at baseline and 6 months later. All transitions during this interval (hospital admissions, nursing home transfers, returns home, death, etc.) were tracked. Using polytomous logistic regression, we tested the effect of cohort on the probability of being hospitalized in light of the competing risks of dying or remaining in the home, controlling for demographic and casemix variables, and having a DNR order in the chart.
Results: A total of 4196 residents were studied, 2118 in 1990 (age 81.3, female 77.7%, LOS 6+ months 49.8%) and 2078 in 1993 (age 81.7, females 75.5%, LOS 6+ months 50.2%). The unadjusted probability of hospitalization dropped from .205 to .151. Multivariate analyses revealed a significant adjusted odds of hospitalization of .74 (95% CI .60-.91) and no cohort effect on home discharge or death. Among severely cognitively impaired residents, the adjusted odds of hospitalization in 1993 compared with the 1990 cohort was 0.74 (.53-1.03). Finally, among survivors in both cohorts who had a follow-up MDS performed, and whose ADL remained stable, 15.9% were hospitalized in 1990, whereas only 10.9% were hospitalized in 1993. On the other hand, ADL decliners were more likely to have been hospitalized in 1993 than in 1990 (40.6% vs 25.2%).
Conclusions: Although other changes in the industry, clinical practice, and health care policy may have influenced hospitalization of nursing home residents, the substantial reductions observed among the cognitively impaired and those with stable ADL suggest superior and uniform assessment information in the form of the RAI contributed significantly to this decline.