585 Background: For patients undergoing surgery for cancer, it has been suggested that risk-adjustment with cancer-specific variables is needed when evaluating short-term outcomes. Our objectives were to assess the influence of cancer-related variables on postoperative complications and hospital quality comparisons. Methods: Patients from ACS NSQIP and NCDB who underwent colorectal resection for cancer were linked (2006-2008) to create a dataset containing robust information on comorbidities, complications, and oncologic variables. Three hierarchical models were developed predicting the NSQIP outcome 30-day mortality or any serious morbidity using variables from (1) NSQIP only, (2) NCDB only, and (3) a combined model using NSQIP and NCDB. Models were compared with fit statistics and hospital outlier agreement. Results: From 146 NSQIP hospitals, 11401 patients underwent a colorectal resection for cancer, of which, 1954 (17%) experienced a mortality or serious morbidity event. The first five variables selected in the NCDB-only model were Charlson comorbidity score, neoadjuvant therapy use, T stage, primary payer, and M stage (c-statistic, 0.64; AIC, 9886). The first five variables selected in the NSQIP-only model were ASA class, preop sepsis, albumin, surgical procedure, and COPD (c-statistic, 0.66; AIC, 9787). In the combined model, neoadjuvant therapy use was the only cancer-specific variable selected in the top five. The remaining variables were ASA class, preop sepsis, albumin, and wound class (c-statistic, 0.67; AIC, 9455). At the hospital-level, the NCDB-only model identified three high outliers (worse than expected) and one low outlier (better than expected). Both the NSQIP-only and combined models identified the same four high and two low outlying hospitals (kappa: 1.0), which agreed marginally with the NCDB-only model (kappa: 0.59). Conclusions: Addition of cancer-specific variables to NSQIP models slightly improved model fit; however, hospital outcome comparisons were identical. For patients with colorectal cancer undergoing resection, cancer-related factors have limited predictive ability for short-term outcomes and did not influence hospital quality comparisons.