Background Surgical quality programs, such as the American College of Surgeons NSQIP, provide reports based on specialty or procedure, with patients aggregated together. It is unknown whether hospital performance differs by patient subgroup (eg cancer vs noncancer patients), masking opportunities for improvement. Our objectives were to determine whether performance differs within a given hospital for 6 contrasting patient subgroups and to identify the percentage of hospitals with greater than chance differences in performance. Study Design Using the American College of Surgeons NSQIP data, adults undergoing lung resection, esophagectomy, hepatectomy, pancreatectomy, colectomy, and proctectomy (2005 through 2012) were divided into 6 contrasting subgroups (elderly vs nonelderly, white vs nonwhite, obese vs nonobese, renal insufficiency vs normal renal function, cancer vs noncancer, emergency vs nonemergency). The main end point was serious morbidity or mortality. Observed to expected ratios were constructed using hierarchical models and compared using paired t-tests (eg observed to expected for cancer cases compared with noncancer). Variation in performance differences was assessed using a randomization test and z-tests for proportions. Results From 433 hospitals, 221,518 patients were included. Overall quality differed for elderly vs nonelderly, renal insufficiency vs normal renal function patients, cancer vs noncancer, and emergency vs nonemergency (p < 0.05). Variation in within-hospital performance differences exceeded chance expectations for renal insufficiency vs normal renal function in 31.1% of hospitals, cancer vs noncancer in 40.8%, and emergency vs nonemergency patients in 55.4% (p < 0.001). Conclusions Hospital performance within a given hospital varies by patient subgroup. Quality programs can consider separate reports for these subgroups to identify opportunities for quality improvement.