| The operating suites of a busy hospital will see over 15,000 surgeries per year. Between 50 and 100 surgeries will be performed per day, involving over 300 staff: physicians, nurses, surgical technicians, and other support personnel. A hospital's surgery schedule for a day is constantly being revised for many reasons: difficult procedures may take longer to complete than expected, emergencies may require reshuffling of rooms and personnel, and ambulatory patients may not arrive when scheduled. While there is an inherent element of unpredictability in estimating in advance a surgery schedule, having and maintaining an accurate schedule is still important for a number of reasons: every change to the schedule requires many people to change their plans, expensive surgical resources can be wasted, and patient satisfaction is adversely effected.; This research makes two contributions: (1) a system for efficiently developing and executing a surgical schedule is described, and (2) an algorithm is described to estimate the surgery time of future surgeries by comparing the surgeon's descriptive coding of the proposed procedures against a historical database of similar procedures and it demonstrates that the implemented algorithm results in estimated surgery times that are statistically more accurate than the surgeons' estimates of the same surgeries.; The resulting algorithm uses a classification-tree approach modified to remove the computational intractability of the “classic” classification tree, and the modified approach provides an explanation of why the resulting surgery duration was chosen. As the size of the database containing the historical surgical information grows, the algorithm is well behaved—the size of the resulting classification tree, the computer time required to generate the tree, and the access time required to find the duration estimate from the tree all grow slightly faster than linearly. The algorithm can accommodate many years of surgical history with modest computer hardware resources. With a cumulative, absolute error rate of 28.3%, the algorithm proved to be more accurate in estimating the duration of 14,250 surgeries than the 33.0% error rate of the surgeons performing the surgery. |
