To address chronic complaints about “OR inefficiency” due to inaccurate case times and scheduling, a hospital utilizes their new surgical services computer information system to track case times. The plan is to then be able to compute a newly scheduled case’s expected duration based on analysis of previous cases’ surgical times.
However, 6 months after adoption of this new scheduling approach using historical data, the accuracy of the OR schedule has not improved. The hospital vice president notes that approximately the same number of complaints from surgeons and patients continue.
Why has accuracy not improved using analysis of historical data? It turns out that scheduling case durations correctly is a more complex undertaking than expected and the level of certainty desired by many stakeholders is often not possible. Best practice is to have OR management accept and manage some of the uncertainty in how long an individual case may last.
The take home message from this white paper is that averaging historical data for case duration predictions does not increase prediction accuracy as much as most people think it should. This is due to several key principles:
• The combination of a great variety of procedures and the large number of surgeons on most hospitals staffs makes it such that on average half of the cases scheduled in a hospital’s surgery suite will have less than 5 previous cases of the same primary procedure type & same surgeon during the preceding year. In other words, often there are not enough similar enough cases to make a prediction regardless of whether statistics are used or how long one goes back in the system to pull out similar cases.
• Also, if case durations for a surgeon performing a particular operative procedure vary significantly due to the nature of the surgery (cancer resection is one example as every tumor is different), then it is also intrinsically very difficult to make accurate predictions, no matter how many previous cases are examined.
• Yet another barrier to truth in scheduling is the statistical distribution of case times which most often are not bell shaped (normal) distributions. This variance, for example, complicates using the average of historical case durations because unusually long cases (outliers) have a disproportionately large effect
The surgeon and the surgical procedure are the two most important determinants of surgical time. Some case lengths are easier to predict than others. These include surgical specialties that operate on the body surface or extremities, where operations are often standardized. On the other hand, surgery duration for many cases is intrinsically difficult to predict especially if the procedure is complex, and the operative steps are not standardized such as for ENT cancer surgery and major intra-abdominal procedures.
Various methods to estimate case duration can be utilized. (Table 1)
Table 1. Models to predict case duration
• mean of historical case duration
• surgeon estimate
• use surgeon estimate in combination with historical data to create new estimate
• adjust for case complexity (e.g. simple, average, or complex)
• some combination of the above
The variance of statistical distributions of case times complicates just using the mean of historical case durations because unusually long cases (outliers) have disproportionately large effect. (Table 2)
Table 2. Possible values that can be computed from historical data
• Median – decreases the impact of unusually long cases
• Trimmed mean – delete lower & upper 10% of the durations and then take the average
• Geometric mean – At some hospitals surgeons consistently shorten their case duration estimates if they perceive they have too little OR time and need to make sure they “fit” their cases into the OR time they have. In contrast, other surgeons may purposely overestimate case durations to keep control/access to their OR time so that if a new case appears for them their OR time has not been given away.
Certainly, estimated case duration will only be helpful if the surgeon correctly schedules the operation they do. If this happens, how should the duration time be kept in database? We are recommending determining the mean case duration of previous cases using scheduled operations, not actual operations.
When surveying facilities many different other explanations are often heard for inaccurate scheduling besides that the surgeon estimates are incorrect. (Table 3)
Table 3. Common reasons provided for case duration inaccuracy
• Eroded “procedure” file
For example lap chole posted under a number of different names (laparoscopic chole, lap cholecystectomy) so system can’t aggregate
Facilities need standardized procedure dictionary
• Multiple procedures counted as one
For example lap chole with appy, lap chole with liver bx, and lap chole all counted as same even though complexity varies
• Accounting for setup & cleanup times that affect how late an OR runs
• Common approach: standard amount of time booked for turnover (setup & cleanup) times
Needs to vary according to case complexity
MANAGING THE UNCERTAINTY THAT IS INTRINSIC TO CASE DURATION PREDICTION
Scheduling accuracy decreases as estimated length of time for the surgical procedure goes up. It is more difficult to know when an 8 hour expected surgery will finish than a 30 minute case.
It would be ideal to have no uncertainty in case duration prediction. Obviously, that is unrealistic. When we ask, “How long will the case last?” we are expecting a defined answer. For example, “The case will last 2 hours.” This provides an “illusion of certainty” that feeds a human emotional need for certainty when none exists.
What research in the field of OR management science has discovered is that in fact, by analyzing historical case data for the same surgeon and procedure, one can assess the uncertainty surrounding the estimate. In other words, case durations have a probability distribution, such that the expected case duration is not a point value, but rather a probability estimate. Therefore, a more informative answer to the question, “How long will this case be” might be, for example, “There is a 67% probability that the case will be finished within 90 minutes.” This is similar to the approach used in reporting weather forecasts.
We are recommending that OR management accept and manage the uncertainty in how long an individual case may last. For example, for some decisions the OR manager needs to consider the shortest time possible that a case will last. This information may assist in deciding on whether to place an urgent case from the wait list in that OR. For other decisions, the OR manager needs to consider the longest possible duration of a case. There may be another OR waiting for equipment that is being utilized by the OR in question.
Best practice for case scheduling involves not having patients show up at a fixed time interval in advance of surgery. Rather, the time a patient is instructed to arrive at the hospital in advance of their surgery should vary based on the nature of the case(s) ahead of them.
For example, if patient Smith is scheduled to follow a case of predictable duration with patient Jones, then patient Smith can be told to arrive closer to her start time. On the other hand, if patient Smith is scheduled to follow a case of uncertain duration (e.g. a cancer resection around the liver), this patient’s instructions might be, “Please arrive early,” knowing that an “open and close” procedure is a possibility.
As far as the efficiency of the surgery suite is concerned, allocating the right amount of OR time to each service each day is paramount. This ensures that workload and staffing are optimally matched. Effectively managing the associated uncertainty in case duration can minimize unnecessary overtime expenditures.
1.Gigerenzer G, Gaissmaier W, Kurz-Milcke E, Schwart L, Woloshin S. Helping doctors and patients make sense of health statistics. Psychological Science in the Public Interest 2008;8: 53-96
2. Dexter F, Epstein RH, Traub RD, Xiao Y. Making management decisions on the day of surgery based on operating room efficiency and patient waiting times. Anesthesiology, 2004; 101:1444-53.