To design a methodology to predict operative times for robot-assisted radical cystectomy (RARC) based on variation in institutional, patient, and disease characteristics to help in operating room scheduling and quality control. The model included preoperative variables and therefore can be used for prediction of surgical times: institutional volume, age, gender, body mass index, American Society of Anesthesiologists score, history of prior surgery and radiation, clinical stage, neoadjuvant chemotherapy, type, technique of diversion, and the extent of lymph node dissection. A conditional inference tree method was used to fit a binary decision tree predicting operative time. Permutation tests were performed to determine the variables having the strongest association with surgical time. The data were split at the value of this variable resulting in the largest difference in means for the surgical time across the split. This process was repeated recursively on the resultant data sets until the permutation tests showed no significant association with operative time. In all, 2 134 procedures were included. The variable most strongly associated with surgical time was type of diversion, with ileal conduits being 70 min shorter (P < 0.001). Amongst patients who received neobladders, the type of lymph node dissection was also strongly associated with surgical time. Amongst ileal conduit patients, institutional surgeon volume (>66 RARCs) was important, with those with a higher volume being 55 min shorter (P < 0.001). The regression tree output was in the form of box plots that show the median and ranges of surgical times according to the patient, disease, and institutional characteristics. We developed a method to estimate operative times for RARC based on patient, disease, and institutional metrics that can help operating room scheduling for RARC.
Read full abstract