Predicting Postoperative Mortality from Comorbidity Indices in Administrative Databases Among Inflammatory Bowel Disease Patients

Purpose: Risk adjustment is employed in studies that use administrative data to evaluate outcomes such as mortality. The two most commonly applied risk adjustment measures are the Elixhauser and the Charlson/Deyo index. However, the predictive capabilities of comorbidity indices have not been studied in patients with inflammatory bowel disease (IBD). Thus, we evaluated whether comorbidity indices were predictive of postoperative mortality among IBD patients who underwent an IBD-related bowel operation. Methods: We used the Nationwide Inpatient Sample (NIS) database to identify 35,588 discharge abstracts of IBD patients who underwent an IBD-related operation from 1995–2005. We used coding algorithms in SAS to evaluate the frequency of comorbidities using the Elixhauser and Charlson/Deyo indices. Multiple logistic regression analyses were performed to compare the mortality prediction achieved for each comorbidity index after adjusting for age, gender, insurance, and emergency admission. We used the c-statistic, an approximation to the area under the receiver operating characteristic curve, as our measure of model discrimination. Bootstrapping was used to estimate the 95% confidence interval (CI) for each C-statistic. Results: The median age of IBD patients was 42 years (IQR 31–56), 48% were male, and the postoperative mortality rate was 1.9%. For the Elixhauser index, as the number of comorbidities increased from 0, 1, 2, or 3+, the unadjusted mortality following an IBD-related operation increased from 0.4%, 1.5%, 3.3%, and 7.9%, respectively; likewise, for the Charlson/Deyo index unadjusted mortality were 0.8%, 4.5%, 7.5%, and 16.3%, respectively. Among comorbidities with a prevalence > 0.4%, the most predictive comorbidities were coagulopathy (adjusted odds ratio (aOR) 12.1; 95% CI 9.2–15.9) and renal failure (aOR 8.7; 95% CI 5.4–14.1) using the Elixhauser method, and renal disease (aOR 5.3, 95% CI 3.7–7.6) and liver disease (aOR 4.7; 95% CI 2.1–10.3) for the Charlson/Deyo index. The C-statistic was higher with the Elixhauser (0.93; 95% CI 0.92–0.94) compared to the Charlson/Deyo index (0.91; 95% CI 0.90–0.92). The C-statistics for Crohn's disease and ulcerative colitis using the Elixhauser method were 0.91 (0.90–0.92) and 0.94 (0.93–0.95), respectively. Lower values were calculated for Crohn's disease (0.89, 0.87–0.91) and ulcerative colitis (0.92, 0.91–0.93) after applying the Charlson/Deyo index. Conclusion: Mortality from IBD surgery is rare. However, comorbidities increased the risk of postoperative mortality and thus should be considered when comparing the risk/benefit ratio of medical versus surgical therapy. Although the Elixhauser index outperformed the Charlson/Deyo index, both indices were predictive of postoperative mortality.