Operating Room Simulation
Operating Room Simulation
This experiment evaluates how different preoperative fasting (NPO — nil per os) policies affect the tradeoff between patient fasting burden and operating room (OR) efficiency, in a single-OR, single-day scheduling setting with 6 surgeries. The problem is formulated as a stochastic program and solved using sample-average approximation (SAA).
A single OR runs a single day of surgery with 6 total surgeries, distributed evenly across surgeons (i.e., each surgeon in a given scenario performs the same number of surgeries). The OR opens at 8:00 AM. The scheduled closing time is set to 8:00 AM plus the mean of the total scheduled surgery durations for that day; however, the OR may remain open indefinitely past this scheduled close until all 6 patients have been operated on (generating overtime when actual durations exceed the schedule). Between consecutive surgeries, the OR requires a turnaround time, modeled as a random variable with mean and standard deviation 30 minutes.
Each individual surgery duration and turnaround time are treated as a random draw from a log-normal distribution, capped at an upper bound of 3 standard deviations above the mean to avoid unrealistic outliers. These random 100 draws form the sample set used in the SAA procedure.
Four scenarios are tested, varying the number of surgeons and the mean surgery duration per surgeon (all with sd = 1 hour):
2 surgeons, homogeneous: both surgeons have mean surgery duration of 1 hour.
3 surgeons, homogeneous: all three surgeons have mean surgery duration of 1 hour.
2 surgeons, heterogeneous: surgeon A mean = 1 hour, surgeon B mean = 2 hours.
3 surgeons, heterogeneous: surgeon A mean = 1 hour, surgeon B mean = 2 hours, and surgeon C mean = 3 hours.
Within each scenario, three NPO (fasting start time) policies are tested:
Uniform: all patients begin fasting at the same scheduled time, regardless of surgeon or surgery order.
By surgeon: all patients assigned to the same surgeon begin fasting at the same scheduled time, but this time may differ across surgeons.
By patient: each patient has an individually optimized fasting start time.
Four cost terms are included in the objective, each with a per-minute cost coefficient varied independently between $0 and $1/minute to explore the resulting tradeoff space:
Excess fasting duration: average time each patient fasts beyond the 8-hour.
Surgeon wait time: average time each surgeon waits before their first surgery of the day begins.
OR idle time: average time the OR sits idle between cases, excluding scheduled turnaround time.
OR overtime: average time the OR must remain open past its scheduled closing time.
The experiment optimizes schedules across combinations of these four cost weights to characterize how the optimal NPO policy and schedule shift as the relative penalties on fasting burden versus OR efficiency change.
Every patient adheres exactly to their assigned NPO schedule (no early or late fasting starts).
Both patients and surgeons arrive precisely on time for their scheduled activities (no punctuality-related delays).