Why Your Scheduled Case Times Are Wrong (and How to Fix It)
Inaccurate scheduled case times wreck OR utilization. Why estimates drift, how surgeon-and-procedure-specific history beats guesses, and where prediction models actually help.
Inaccurate scheduled case times are one of the most underrated sources of OR inefficiency. When the booked duration is wrong, everything downstream inherits the error: utilization is miscalculated, blocks are mis-sized, start times slip, and the day either runs over or sits idle. Better surgical case duration prediction is therefore a foundational fix — it makes every other scheduling decision more accurate. This post explains why estimates drift, what actually improves them, and where prediction models genuinely help versus where they're oversold.
How bad case-time estimates sabotage the whole day
A schedule is only as good as the durations it is built on. Book a case at 90 minutes that reliably takes 120, and you have created a 30-minute deficit that propagates: the next case starts late, the block looks fuller than it is, and by afternoon the room is behind or staff are idle waiting on a case that finished early. The error doesn't stay contained in one case — it sets the baseline for the entire day. This is why case-duration accuracy belongs on the short list of metrics that actually matter.
Why surgeon-supplied estimates drift
Most scheduled times trace back to a surgeon's estimate or a generic procedure default, and both drift for predictable reasons: optimism (the case as it should go, not as it usually does), round numbers (everything becomes "an hour" or "90 minutes"), and case-mix blindness (the estimate doesn't distinguish a straightforward case from a complex one). None of this is dishonesty — it's the natural unreliability of memory and averages standing in for data.
Using historical, surgeon-and-procedure-specific data
The reliable fix is to replace estimates with the actual record: how long this surgeon takes for this procedure, drawn from real historical cases. Surgeon-and-procedure-specific history captures the individual variation that generic defaults erase, and it sharpens as cases accumulate. It's unglamorous and it works — and it's the necessary foundation before any fancier modeling is worth considering.
History first, models second
Before investing in predictive modeling, get the basics right: book from surgeon-and-procedure-specific historical durations, not estimates or facility-wide averages. That single change captures most of the available accuracy gain for none of the complexity.
Where prediction models help (and their limits)
Here the evidence is refreshingly grounded. A classic Dexter and Traub study found that having perfect information on case duration reduces overtime by only a few minutes per OR compared with simply using good historical durations. The implication: duration prediction by itself has a modest ceiling — history already gets you most of the way.
The bigger gains appear when prediction is coupled with optimization. A more recent predict-then-optimize study in orthopedics trained machine-learning models on hundreds of thousands of cases to predict duration, then fed those predictions into a schedule-optimization step — and reduced overtime by roughly 300 to 500 minutes per week versus mean-duration scheduling (with a modest tradeoff of slightly more underused time). The lesson is precise: the value is in using better durations to build better schedules, not in the prediction number on its own. We dig into that distinction in AI and predictive analytics in OR scheduling.
How ORbit learns your actual case durations
Accurate durations require continuously learning from your real cases — exactly what a static booking template can't do. ORbit builds surgeon-and-procedure-specific duration profiles from your own case history, median-based so a single marathon case doesn't distort the estimate, and facility-scoped. That makes scheduled times reflect how your ORs actually run, which in turn makes utilization and start-time metrics trustworthy. To see how close your booked times are to reality, take a look at your own data.
Frequently asked questions
Why are scheduled surgical case times so often wrong?
Because they usually rely on surgeon-supplied estimates or generic procedure defaults rather than that surgeon's actual history for that procedure. Estimates drift for understandable reasons — optimism, round numbers, and case-mix differences — and the booked time inherits the error, which then distorts utilization and start times all day.
How can I make case-time estimates more accurate?
Use historical, surgeon-and-procedure-specific durations instead of generic estimates. The actual record of how long a given surgeon takes for a given procedure is a far better predictor than a remembered estimate, and it improves as more cases accumulate.
Do AI case-duration prediction models help?
They can, but mostly when paired with schedule optimization rather than used alone. One study found that perfect duration information reduces overtime by only a few minutes per OR versus good historical data — while a more recent predict-then-optimize approach that fed ML duration predictions into schedule optimization cut overtime by 300 to 500 minutes per week. The optimization, not the prediction alone, is where most of the value is.