Forecasting and perception of average and latest hours worked by on-call anesthesiologists

We studied the value of providing information to anesthesia providers about the length of time typically worked during on-call shifts. The mean time at which a shift ends can be used for purposes of trades, payments, or reverse auctions, because the mean is proportional to the total time. The 80th percentile (with a suitable upper confidence limit for uncertainty due to limited sample sizes) can be used for judging the earliest time by which after-work activities reasonably can be planned.

(A) Three years of operating room (OR) information system data were analyzed. Dependent variables were the earliest times when the numbers of ORs running were always

(A) For the three thresholds and two end points (mean and 80th percentile), differences among days of the week were as large as 45 min. Differences between end points for the same weekdays were as large as 245 min. Comparatively, additional knowledge about the number or hours of cases provided in the late afternoon on the working day before surgery reduced the mean absolute error by only 4.1-6.0 min. Results were insensitive to a variety of analytic methods. Information available more days before the day of surgery (e.g., 1 wk) would have had even less incremental predictive value. (B) The mean absolute error of anesthesiologists' estimates for 80th percentiles was 60 min, principally because of underestimation of the 80th percentiles. More than half (69%, P = 0.0003) of anesthesiologists' estimates for 80th percentiles had error >30 min, whereas errors of this magnitude were less for the mean (44%, P = 0.0004).

Historical data from OR or anesthesia information management systems, or from anesthesia billing systems, can be used months before staff scheduling to provide insight to anesthesia providers on respective calls. The data are useful because experience provides limited intuition. Updates on scheduled workload available closer to the day of surgery provided only marginal increases in knowledge over the use of historical data.