Outlier Identification in Colorectal Surgery Should Separate Elective and Nonelective Service Components

BACKGROUND: The identification of health care institutions with outlying outcomes is of great importance for reporting health care results and for quality improvement. Historically, elective surgical outcomes have received greater attention than nonelective results, although some studies have examined both. Differences in outlier identification between these patient groups have not been adequately explored. OBJECTIVE: The aim of this study was to compare the identification of institutional outliers for mortality after elective and nonelective colorectal resection in England. DESIGN: This was a cohort study using routine administrative data. Ninety-day mortality was determined by using statutory records of death. Adjusted Trust-level mortality rates were calculated by using multiple logistic regression. High and low mortality outliers were identified and compared across funnel plots for elective and nonelective surgery. SETTINGS: All English National Health Service Trusts providing colorectal surgery to an unrestricted patient population were studied. PATIENTS: Adults admitted for colorectal surgery between April 2006 and March 2012 were included. INTERVENTION(S): Segmental colonic or rectal resection was performed. MAIN OUTCOME MEASURES: The primary outcome measured was 90-day mortality. RESULTS: Included were 195,118 patients, treated at 147 Trusts. Ninety-day mortality rates after elective and nonelective surgery were 4% and 18%. No unit with high outlying mortality for elective surgery was a high outlier for nonelective mortality and vice versa. Trust level, observed-to-expected mortality for elective and nonelective surgery, was moderately correlated (Spearman rho = 0.50, p < 0.001). LIMITATIONS: This study relied on administrative data and may be limited by potential flaws in the quality of coding of clinical information. CONCLUSIONS: Status as an institutional mortality outlier after elective and nonelective colorectal surgery was not closely related. Therefore, mortality rates should be reported for both patient cohorts separately. This would provide a broad picture of the state of colorectal services and help direct research and quality improvement activities.