Comparison of Anastomotic Leak Rate After Colorectal Surgery Using Different Databases

BACKGROUND: Anastomotic leaks are one of the most important clinical outcomes after colorectal anastomosis. Because of the lack of measurement of this outcome in databases, research has been limited by the need to perform chart review. OBJECTIVE: The aim of this study was to evaluate the ability of 2 sources, an administrative database and a clinical registry, to identify anastomotic leaks. DESIGN AND SETTING: A retrospective cohort study of patients undergoing colorectal procedures at an academic medical center over a 1-year period was performed. PATIENTS: International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes were used to identify patients, and chart review of all patient records was performed. Risk factors for anastomotic leak were recorded along with the presence or absence of anastomotic leak. MAIN OUTCOME MEASURES: Patients were identified as having a leak in the University HealthSystem Consortium (administrative database) by procedure codes and in the National Surgical Quality Improvement Program (clinical registry) if they had a postoperative organ space surgical site infection. The administrative and clinical data sources were compared with the use of sensitivity, specificity, accuracy, positive predictive value, and negative predictive value for anatomotic leak. RESULTS: We identified 424 patients; 66 that did not have an anastomosis and 6 that lacked outpatient follow-up were excluded. Anastomotic leak was identified by chart review in 24 patients (6.8%). The clinical registry had the highest specificity (97%) and sensitivity (8%). LIMITATIONS: Because of the lack of a definition for anastomotic leak in either the administrative database or clinical registry, logical proxies were used. CONCLUSIONS: Although the clinical registry had higher sensitivity and specificity for anastomotic leak, both databases had low sensitivity. Future research on anastomotic leaks would benefit greatly from a uniform definition and recording of this outcome in national databases.