Meta-analysis of home-care effects on mortality and nursing-home placement

Abstract

This meta-analysis was designed to illustrate how the relatively new discipline of meta-analysis could be employed in health-services research. In pursuit of this goal, the results of 13 studies on the effect of home care on mortality and nursing-home placements were examined. The analysis demonstrated a small, beneficial effect of home care on mortality, which fell short of statistical significance. Study-to-study variation in the odds ratio was found that was not strongly associated with differences in the study samples, designs, or interventions as categorized here. The analysis produced stronger evidence of a reduction in nursing-home placements. In this case, differences in study design explained much of the heterogeneity in results, with the randomized controlled trials showing considerably weaker effects. The analysis also provides insight into the benefits and limitations of alternative meta-analytic methods. The "odd man out" method recently described by Walker et al. detected statistically significant heterogeneity of effects across studies for nursing-home placement, but not for mortality. Loglinear modeling made it easier to detect and explore study-to-study variation in home-care effects.