|
Abstract:
|
Monitoring health providers for mortality is important, and quality measures based on indirect and direct standardization have been proposed, with an ongoing debate as to which standardization method is more appropriate. It has been argued that provider-specific indirect standardized estimators should not be compared with one another, because each provider’s indirect standardized measure is adjusted to a different (provider-specific) covariate distribution. Based on a Cox model for covariate adjustment and provider effects, we propose a new approach to estimate a standardized mortality ratio based on indirect standardization (i.e. the observed number of deaths divided by an expected number under a national norm). We show that this measure is also a direct standardized measure provided the national norm is chose in a particular way. For this approach, the perceived disadvantage for ranking of indirect standardization disappears. We assess finite-sample properties by simulation and apply the methods to US kidney transplant data. We also consider generalizations that allow for a time varying provider risk.
|
