Case-Mix Adjustment of Multidimensional Patient-Reported Outcomes to Assess Hospital Performance

Multidimensional patient-reported outcome measures (PROMs) are increasingly used to compare the performance of health care providers. Meaningful comparisons require the data to be adjusted for differences in case-mix. The common approach in the literature is to aggregate PROM data into univariate summary scores prior to case-mix adjustment. In this paper, we extend the work by McCarthy (2015, 2016) to develop a dimension-specific adjustment methodology for performance assessment. This method permits to model the influence of patient characteristics separately for each dimension while retaining the maximum amount of information to inform different stakeholders. We conduct a simulation study to compare the performance of both methods in terms of their ability to identify provider performance, and to identify context-specific factors that should guide the choice of methodology in applied research. We follow this up with a real-world case study of knee replacement surgery in the English NHS. We find that the domain-specific approach outperforms the summary score approach in most cases but the differences are small. We conclude that the choice of method ought to consider not just performance but also ease of implementation and retention of decision-relevant granular information.