•Background. The English Department of Health has introduced routine collection of patient-reported outcome data for selected surgical procedures to facilitate patient choice and increase hospital accountability. However, using aggregate health outcome scores, such as EQ-5D utilities, for performance assessment purposes causes information loss and raises statistical and normative concerns. Objectives. For hip replacement surgery, we explore a) the change in patient-reported outcomes between baseline and follow-up on 5 health dimensions (EQ-5D), b) the extent to which treatment impact varies across hospitals, and c) the extent to which hospital performance on EQ-5D dimensions is correlated with performance on the EQ-5D utility index. Methods. We combine information on pre- and postoperative EQ-5D outcomes with routine inpatient data for the financial year 2009–2010. The sample consists of 21,000 patients in 153 hospitals. We employ hierarchical ordered probit risk-adjustment models that recognize the multilevel nature of the data and the response distributions. The treatment impact is modeled as a random coefficient that varies at the hospital level. We obtain hospital-specific empirical Bayes (EB) estimates of this coefficient. We estimate separate models for each EQ-5D dimension and the EQ-5D utility index and analyze correlations of EB estimates across these. Results. Hospital treatment is associated with improvements in all EQ-5D dimensions. Variability in treatment impact is most pronounced on the mobility and usual activities dimensions. Conversely, only pain/discomfort and anxiety/depression correlate well with performance measures based on utilities. This leads to different assessments of hospital performance across metrics. Conclusions. Our results indicate which hospitals are better than others in improving health across particular EQ-5D dimensions. We demonstrate the importance of evaluating dimensions of the EQ-5D separately for the purposes of hospital performance assessment.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)