Physician derived versus administrative data in identifying surgical complications. Fact versus Fiction

Administrative data are widely used as determinants of surgical quality. We compared surgical complications identified in a structured surgical review to coding and billing data of over a 19-month period.

A retrospective review of monthly morbidity and mortality conference reports was compared to a report over the same time period generated from hospital coding and billing data.

807 sequential operative procedures were included. Physician derived data compared to administrative data identified a complication of any severity in 205 (25.4%) versus 111 (13.8%) cases (r = 0.39), and major complications in 68 (8.4%) versus 46 (5.7%) cases (r = 0.36). Review of the administrative data regarding major complications identified 80 false negatives, 52 false positives, and 38 true positive designations. Overall sensitivity, specificity, positive and negative predictive values, and accuracy for administrative data in identifying major complications was 0.32, 0.99, 0.42, 0.99, and 0.99.

The correlation between physician determined and administrative data with regard to identifying surgical complications is poor. Administrative data are insensitive and lack positive predictive value.