Predicting discharge outcomes of VA nursing home residents.

This article's purpose was to identify predictors of discharge outcomes of VA nursing home stays. Using data tapes, diagnostic and assessment data were assembled on elderly individuals admitted to VA nursing homes nationwide during Fiscal Year 1987. Six-month outcomes for 3 groups were considered: all residents (n = 5,895), and those remaining in care after 6 (n = 2,815) and 12 months (n = 1,812), respectively. Logistic regression was used to evaluate predictors of death and community discharge. Limited activities of daily living (ADL) dependency, younger age, and receipt of rehabilitation services most consistently predicted community discharge. ADL dependency, older age, oxygen use, terminally ill prognosis, malignancy, and congestive heart failure most consistently predicted mortality. For both dependent variables, predictive ability declined as stay length increased. Predicting death and community discharge become increasingly problematic as stay lengthens. Comparing observed versus expected discharge outcomes has limited usefulness as a quality-improvement tool.