Socio-economic predictors and between-hospital variation in permanent stoma rates after segmental resection of colorectal cancer. A population-based register study.

Abstract

Annual reports on colorectal cancer (CRC) in Denmark reveal marked variations between hospitals in the rates of permanent stomas. The objective of this study was to explore whether socio-economic factors and the treating hospital were independent predictors of creating a permanent stoma rather than an anastomosis in segmental bowel resection for CRC. This was a register-based national cohort study comprising patients with CRC stratified into four subpopulations according to area of segmental resection: right colon, sigmoid colon, upper third of rectum and mid third of rectum. Primary outcome was anastomosis or permanent stoma, as indicated by the surgeon. We employed multiple logistic regression analyses, adjusting for demographic, performance, lifestyle, comorbid and perioperative factors. The right colon had insufficient permanent stoma cases for regression analyses. The remaining three groups comprised a total of 14,113 patients. The rates of permanent stoma in both the upper rectum and the mid-rectum increased with lower household income quartiles. Odds ratios for the lowest income quartiles were 2.1 (95% CI: 1.36-3.13) for the upper rectum and 1.7 (95% CI: 1.25-2.18) for the mid-rectum compared with the highest household income quartile. Educational level and cohabitation were non-significant in the adjusted regressions. We found marked between-hospital variation in permanent stoma rates, compared to the largest hospital, in both sigmoid colon (up to 2.17), upper rectum (up to 4.0) and mid-rectum (up to 4.6) after adjustment (p < 0.01). A permanent stoma after elective resection of cancer in the upper two-thirds of the rectum was associated with lower household income. There were marked between-hospital variations in permanent stoma rates after resection of the sigmoid colon and the upper two-thirds of the rectum in CRC, even after adjustment for relevant covariates.