East Coast Ballroom
Using z-scores to measure within-hospital outcome improvement over time in a regional quality improvement collaborative (307845)
*Anne H. Cain-Nielsen, Department of Surgery, University of MichiganJill L. Jakubus, Department of Surgery, University of Michigan
Judy N. Mikhail, Department of Surgery, University of Michigan
Mark R. Hemmila, Department of Surgery, University of Michigan
Keywords: Hospital performance measures, outcome measures, quality improvement
Creating measures of hospital quality improvement that are credible and reliable is notoriously challenging. Clinicians often perceive quality outcome measures as being inadequately risk-adjusted. To promote consistent comparisons, the Michigan Trauma Quality Improvement Program implemented a set of within-hospital-only (i.e. 'competing with yourself') measures. We calculated changes over time in the outcomes of mortality and serious complications using separate logistic regression models for each center. Models included case-mix factors and time (quarters over three years). The z-statistic for the test of ßtime=0 (z=ßtime/se(ßtime)) was reported to each center in the collaborative as their 'z-score.' The median hospital z-score for serious complications was –0.7 (IQR –1.6-0.3) and –0.4 for mortality (IQR –1.0-0.1), with 18/35 centers showing some evidence of improvement (defined as z<-1) in one or both measures. The analysis accounted for high performers subject to floor effects. Reporting within-hospital outcome measures is feasible within a quality collaborative and may help to improve perceptions of inadequate risk adjustment when measuring longitudinal hospital performance.