The Impact of Under Coding of Cardiac Severity and Comorbid Diseases on the Accuracy of Hospital Report Cards
Context: Hospital report cards are usually based on administrative discharge abstracts. However, cardiac severity and co-morbidities are generally under-reported in administrative data. Objective: To determine how under-coding of cardiac severity and co-morbidities affects the determination that some hospitals are high-mortality outliers. Design: Simulations using retrospective data on 18,795 patients admitted with an acute myocardial infarction (AMI) to 109 acute care hospitals in Ontario. Main Outcome Measure: Change in the number of hospitals that remained high-mortality outliers after adjusting for potentially increased prevalence of up to nine separate measures of cardiac severity and co-morbid conditions, individually or together. Results: For most measures of cardiac severity and co-morbidities, increasing the prevalence of each factor to the highest observed level hospital-specific prevalence seldom altered the status of high-mortality outlier hospitals. Increases in the prevalence of cardiogenic shock or acute renal failure to even the median level led to reclassification of up to 4 of the 12 high-mortality outlier hospitals to non-outlier status. The majority of high-mortality outlier hospitals were reclassified if the maximum prevalence was imputed for these two factors. Simultaneously increasing the prevalence of all co-morbidities to the median level typically converted the status of about half the outlier hospitals. Not until the prevalence of all measures of cardiac severity and co-morbidities were simultaneously increased to the maximum observed hospital-specific prevalence, did all hospitals initially classified as high-mortality outliers revert to non-outlier status. Conclusions: Under-coding of severity and co-morbidities in administrative data in itself is very unlikely to account for the outlier status of most hospitals. However, some potential for misclassification of individual institutions exists if influential factors are variably coded.
Peter C. Austin. "The Impact of Under Coding of Cardiac Severity and Comorbid Diseases on the Accuracy of Hospital Report Cards" Medical Care 43 (2005): 801-809.
This document is currently not available here.