Appending limited clinical data to an administrative database for assessing institutional quality outliers for AMI...

The major objective of this study was to determine whether individual hospital performance would be assessed differently if clinical data were added to an administrative dataset.

Patients in the 2004 New York State AMI Registry (AMI registry) who could be matched to patients in the New York State Hospital Discharge Database (SPARCS model) comprised the study sample (n=3153). Stepwise logistic regression models were developed (SPARCS model, SPARCS/AMI registry model). Risk-adjusted mortality rates (RAMR) for each hospital in the matched dataset were determined and compared for the SPARCS and the SPARCS/AMI registry model. The RAMR for each hospital was determined by dividing its observed mortality rate by its expected mortality rate and multiplying by the overall mortality rate for the state of New York. Hospitals were considered outliers if they had a RAMR significantly higher or lower than the overall statewide mortality rate. Hierarchical Models were also used to identify hospital outliers.

The SPARCS logistic model identified two high hospital outliers; the SPARCS/AMI registry model identified one of those outliers and no others. When Hierarchical Models were used, the SPARCS model also identified two high outliers (one in common with the logistic model) and the SPARCS/AMI registry model identified one high outlier (the same as identified in the logistic model).

It is worth exploring the impact of the addition of a small number of clinical data elements to administrative datasets on hospital outlier status.