Abstract Background Administrative databases oversimplify the relationship of factors such as volume or training on surgical outcomes. Methods A prospective statewide surgeon-initiated database was queried to obtain incident cases of rectal cancer in Vermont from April 1999 to June 2001. Demographics, procedure performed, method of detection, American Society of Anesthesiologists classification, blood transfusions, length of stay, complications, stage, and use of adjuvant therapy were recorded by the operating surgeon. A post hoc analysis was performed on patients operated on for rectal cancer to define the specific impact of specialty training on care patterns. Results There was a marked difference in the distribution of surgical procedures, with colorectal surgeons using local excision and coloanal anastomosis in addition to anterior and abdominoperineal resection. Although the overall use of adjuvant therapy was similar, patients in the colorectal group were more likely to receive preoperative then postoperative radiation therapy (91% vs 17%, P <.0001) and more likely to receive radiation therapy when stage appropriate (98% vs 67%, P <.001). Conclusions Colorectal specialty training in this population was a surrogate for a wider array of surgical options and preoperative radiation. Failure to use radiation when stage-appropriate was related to patient comorbidities and/or refusal and not related to failure of the surgeon to offer adjuvant therapy. Prospective, surgeon-initiated databases provide an excellent opportunity to identify and understand practice variability.