BACKGROUND: Predictive variability of operating room (OR) times influences decision making on the day of surgery including when to start add-on cases, whether to move a case from one OR to another, and where to assign relief staff. One contributor to predictive variability is process variability, which arises among cases of the same procedure(s). Another contributor is parameter uncertainty, which is caused by small sample sizes of historical data. METHODS: Process variability was quantified using absolute percentage errors of surgeons' bias-corrected estimates of OR time. The influence of procedure classification on process variability was studied using a dataset of 61,353 cases, each with 1 to 5 scheduled and actual Current Procedural Terminology (CPT) codes (i.e., a standardized vocabulary). Parameter uncertainty's sensitivity to sample size was quantified by studying ratios of 90% prediction bounds to medians. That studied dataset of 65,661 cases was used previously to validate a Bayesian method to calculate 90% prediction bounds using combinations of surgeons' scheduled estimates and historical OR times. RESULTS: (1) Process variability differed significantly among 11 groups of surgical specialty and case urgency (P < 0.0001). For example, absolute percentage errors exceeded the overall median of 22% for 57% of urgent spine surgery cases versus 42% of elective spine surgery cases. (2) Process variability was not increased when scheduled and actual CPTs differed (P = 0.23 without and P = 0.47 with stratification based on the 11 groups), because most differences represented known (planned) options inherent to procedures. (3) Process variability was not associated with incidence of procedures (P = 0.79), after excluding cataract surgery, a procedure with high relative variability. (4) Parameter uncertainty from uncommon procedures (0-2 historical cases) accounted for essentially all of the uncertainty in decisions dependent on estimates of OR times. The Bayesian method moderated the effect of small sample sizes on uncertainty in estimates of OR times. In contrast, from prior work, the use of broad categories of procedures reduces parameter uncertainty but at the expense of increased process variability. CONCLUSIONS: For procedures with few historic data, the Bayesian method allows for effective case duration prediction, permitting use of detailed procedure descriptions. Although fine resolution of scheduling procedures increases the chance of performed procedure(s) differing from scheduled procedure(s), this does not increase process variability. Future studies need both to address differences in process variability among specialties and accept the limitation that findings from one may not apply to others.
BACKGROUND: Predictive variability of operating room (OR) times influences decision making on the day of surgery including when to start add-on cases, whether to move a case from one OR to another, and where to assign relief staff. One contributor to predictive variability is process variability, which arises among cases of the same procedure(s). Another contributor is parameter uncertainty, which is caused by small sample sizes of historical data. METHODS: Process variability was quantified using absolute percentage errors of surgeons' bias-corrected estimates of OR time. The influence of procedure classification on process variability was studied using a dataset of 61,353 cases, each with 1 to 5 scheduled and actual Current Procedural Terminology (CPT) codes (i.e., a standardized vocabulary). Parameter uncertainty's sensitivity to sample size was quantified by studying ratios of 90% prediction bounds to medians. That studied dataset of 65,661 cases was used previously to validate a Bayesian method to calculate 90% prediction bounds using combinations of surgeons' scheduled estimates and historical OR times. RESULTS: (1) Process variability differed significantly among 11 groups of surgical specialty and case urgency (P < 0.0001). For example, absolute percentage errors exceeded the overall median of 22% for 57% of urgent spine surgery cases versus 42% of elective spine surgery cases. (2) Process variability was not increased when scheduled and actual CPTs differed (P = 0.23 without and P = 0.47 with stratification based on the 11 groups), because most differences represented known (planned) options inherent to procedures. (3) Process variability was not associated with incidence of procedures (P = 0.79), after excluding cataract surgery, a procedure with high relative variability. (4) Parameter uncertainty from uncommon procedures (0-2 historical cases) accounted for essentially all of the uncertainty in decisions dependent on estimates of OR times. The Bayesian method moderated the effect of small sample sizes on uncertainty in estimates of OR times. In contrast, from prior work, the use of broad categories of procedures reduces parameter uncertainty but at the expense of increased process variability. CONCLUSIONS: For procedures with few historic data, the Bayesian method allows for effective case duration prediction, permitting use of detailed procedure descriptions. Although fine resolution of scheduling procedures increases the chance of performed procedure(s) differing from scheduled procedure(s), this does not increase process variability. Future studies need both to address differences in process variability among specialties and accept the limitation that findings from one may not apply to others.
Authors: Rodney A Gabriel; Bhavya Harjai; Sierra Simpson; Nicole Goldhaber; Brian P Curran; Ruth S Waterman Journal: Anesth Analg Date: 2022-04-07 Impact factor: 6.627