Case duration prediction and estimating time remaining in ongoing cases.

In this issue of the British Journal of Anaesthesia, Jiao and colleagues applied a neural network model for surgical case durations to predict the operating room times remaining for ongoing anaesthetics. We review estimation of case durations before each case starts, showing why their scientific focus is useful. We also describe managerial epidemiology studies of historical data by the scheduled procedure or distinct combinations of scheduled procedures included in each surgical case. Most cases have few or no historical data for the scheduled procedures. Generalizability of observational results such as theirs, and automatic computer assisted clinical and managerial decision-making, are both facilitated by using structured vocabularies when analysing surgical procedures.

Machine Learning Can Improve Estimation of Surgical Case Duration: A Pilot Study.

Operating room (OR) utilization is a significant determinant of hospital profitability. One aspect of this is surgical scheduling, which depends on accurate predictions of case duration. This has been done historically by either the surgeon based on personal experience, or by an electronic health record (EHR) based on averaged historical means for case duration. Here, we compare the predicted case duration (pCD) accuracy of a novel machine-learning algorithm over a 3-month period. A proprietary machine learning algorithm was applied utilizing operating room factors such as patient demographic data, pre-surgical milestones, and hospital logistics and compared to that of a conventional EHR. Actual case duration and pCD (Leap Rail vs EHR) was obtained at one institution over the span of 3 months. Actual case duration was defined as time between patient entry into an OR and time of exit. pCD was defined as case time allotted by either Leap Rail or EHR. Cases where Leap Rail was unable to generate a pCD were excluded. A total of 1059 surgical cases were performed during the study period, with 990 cases being eligible for the study. Over all sub-specialties, Leap Rail showed a 7 min improvement in absolute difference between pCD and actual case duration when compared to conventional EHR (p < 0.0001). In aggregate, the Leap Rail method resulted in a 70% reduction in overall scheduling inaccuracy. Machine-learning algorithms are a promising method of increasing pCD accuracy and represent one means of improving OR planning and efficiency.

A Machine Learning Approach to Predicting Case Duration for Robot-Assisted Surgery.

Robot-assisted surgery (RAS) requires a large capital investment by healthcare organizations. The cost of a robotic unit is fixed, so institutions must maximize use of each unit by utilizing all available operating room block time. One way to increase utilization is to accurately predict case durations. In this study, we sought to use machine learning to develop an accurate predictive model for RAS case duration. We analyzed a random sample of robotic cases at our institution from January 2014 to June 2017. We compared the machine learning models to the baseline model, which is the scheduled case duration (determined by previous case duration averages and surgeon adjustments). Specifically, we used: 1) multivariable linear regression, 2) ridge regression, 3) lasso regression, 4) random forest, 5) boosted regression tree, and 6) neural network. We found that all machine learning models decreased the average root-mean-squared error (RMSE) as compared to the baseline model. The average RMSE was lowest with the boosted regression tree (80.2 min, 95% CI 74.0-86.4), which was significantly lower than the baseline model (100.4 min, 95% CI 90.5-110.3). Using boosted regression tree, we can increase the number of accurately booked cases from 148 to 219 (34.9% to 51.7%, p < 0.001). This study shows that using various machine learning approaches can improve the accuracy of RAS case length predictions, which will increase utilization of this limited resource. Further work is needed to operationalize these findings.

Modeling daily veterinary anesthetist patient care hours and probabilities of exceeding critical thresholds.

OBJECTIVE: Use a referral dental clinic model to study how to calculate accurate 95% upper confidence limits for probabilities of workloads (total case duration, including turnover time) exceeding allocated times. ANIMALS: Dogs and cats undergoing dental treatments. METHODS: Managerial data (procedure date and duration) collected over 44 consecutive operative workdays were used to calculate the daily anesthetist workload. Workloads were compared with a normal distribution using the Shapiro-Wilk test, serial correlation was examined by runs test, and comparisons among weekdays were made using the Kruskal-Wallis test. The 95% confidence limits for normally distributed workloads exceeding allocated times were estimated with a generalized pivotal quantity. The impact of a number of procedures was assessed with scatterplots, Pearson linear correlation coefficients, and multivariable linear regression. RESULTS: Mean anesthetist's workload was normally distributed (Shapiro-Wilk P = .25), without serial correlation (P = .45), and without significant differences among weekdays (P = .52). Daily workload, mean 9.39 hours and SD 3.06 hours, had 95% upper confidence limit of 4.47% for the probability that exceeding 16 hours (ie, 8 hours per each of 2 tables). There was a strong positive correlation between daily workload and the end of the workday (r = .85), significantly larger than the correlation between the end of the workday and the number of procedures (r = .64, P < .0001). CLINICAL RELEVANCE: There are multiple managerial applications in veterinary anesthesia wherein the problem is to estimate risks of exceeding thresholds of workload, including the costs of hiring a locum, scheduling unplanned add-on cases, planning for late discharge of surgical patients to owners, and coordinating anesthetist breaks.

Preventing Prolonged Times to Awakening While Mitigating the Risk of Patient Awareness: Gas Man Computer Simulations of Sevoflurane Consumption From Brief, High Fresh Gas Flow Before the End of Surgery.

Prolonged times to tracheal extubation are associated with adverse patient and economic outcomes. We simulated awakening patients from sevoflurane after long-duration surgery at 2% end-tidal concentration, 1.0 minimum alveolar concentration (MAC) in a 40-year-old. Our end-of-surgery target was 0.5 MAC, the Michigan Awareness Control Study's threshold for intraoperative alerts. Consider an anesthetist who uses a 1 liter/minute gas flow until surgery ends. During surgical closure, the inspired sevoflurane concentration is reduced from 2.05% to 0.62% (i.e., MAC-awake). The estimated time to reach 0.5 MAC is 28 minutes. From a previous study, 28 minutes exceeded ≥95% of surgical closure times for all 244 distinct surgical procedures (N=23,343 cases). Alternatively, the anesthetist uses 8 liters/minute gas flow with the vaporizer at MAC-awake for 1.8 minutes, which reduces the end-tidal concentration to 0.5 MAC. The anesthetist then increases the vaporizer to keep end-tidal 0.5 MAC until the surgery ends. An additional simulation shows that, compared with simulated end-tidal agent feedback control, this approach consumed 0.45 mL extra agent. Simulation results are the same for an 80-year-old patient. The extra 0.45 mL has a global warming potential comparable to driving 26 seconds at 40 kilometers (25 miles) per hour, comparable to route modification to avoid potential roadway hazards.

Impact of body mass index on surgical case durations in an academic medical center.

STUDY OBJECTIVE: To investigate the association between patient body mass index (BMI) and operating room duration. DESIGN: Retrospective cohort analysis. SETTING: Demographic data and anesthesia/surgical times for adult surgical patients at University of Virginia Health between August 2017 and February 2019 were collected and analyzed. PATIENTS: A total of 31,548 cases were included in the final analysis. 55% of patients were female, and 51% were classified as ASA Physical Status 2. The mean operating room (OR) duration was 144.2 min ± 112.7 (median = 118, IQR = 121). Orthopedic surgery (32%) was the most common surgery. MEASUREMENTS: Linear mixed effects models were used to examine whether procedure intervals differed across three BMI categories (BMI < 30, 30 ≤ BMI < 40, BMI ≥ 40), considering within-surgeon correlations. Surgical times were log-transformed to correct for positive skewness. MAIN RESULTS: The average time in the operating room was longer for patients with higher BMI (mean ± SD [median, IQR] = 139.5 ± 111.2 [113.0, IQR = 114], 150.2 ± 115.4 [125, IQR = 127], and 153.1 ± 111.1 [130, IQR = 134] for BMI < 30, 30 ≤ BMI < 40, and BMI ≥ 40), respectively. We found a 2% [95% CI = 1-3%] and 3% [95% CI = 1-5%] increase in OR time for 30 ≤ BMI < 40 and BMI ≥ 40, respectively, compared to BMI < 30, after controlling for within-surgeon correlations and covariates. The excess time was primarily determined by anesthesia times. CONCLUSION: In an academic hospital, patients with BMI ≥ 30 required more time in the operating room than patients with BMI < 30, when controlling for confounders. This information can be incorporated into modern-day OR scheduling software, potentially resulting in more accurate case duration estimates that reduce waiting and improve OR efficiency.

An Ensemble Learning Approach to Improving Prediction of Case Duration for Spine Surgery: Algorithm Development and Validation.

BACKGROUND: Estimating surgical case duration accurately is an important operating room efficiency metric. Current predictive techniques in spine surgery include less sophisticated approaches such as classical multivariable statistical models. Machine learning approaches have been used to predict outcomes such as length of stay and time returning to normal work, but have not been focused on case duration. OBJECTIVE: The primary objective of this 4-year, single-academic-center, retrospective study was to use an ensemble learning approach that may improve the accuracy of scheduled case duration for spine surgery. The primary outcome measure was case duration. METHODS: We compared machine learning models using surgical and patient features to our institutional method, which used historic averages and surgeon adjustments as needed. We implemented multivariable linear regression, random forest, bagging, and XGBoost (Extreme Gradient Boosting) and calculated the average R2, root-mean-square error (RMSE), explained variance, and mean absolute error (MAE) using k-fold cross-validation. We then used the SHAP (Shapley Additive Explanations) explainer model to determine feature importance. RESULTS: A total of 3189 patients who underwent spine surgery were included. The institution's current method of predicting case times has a very poor coefficient of determination with actual times (R2=0.213). On k-fold cross-validation, the linear regression model had an explained variance score of 0.345, an R2 of 0.34, an RMSE of 162.84 minutes, and an MAE of 127.22 minutes. Among all models, the XGBoost regressor performed the best with an explained variance score of 0.778, an R2 of 0.770, an RMSE of 92.95 minutes, and an MAE of 44.31 minutes. Based on SHAP analysis of the XGBoost regression, body mass index, spinal fusions, surgical procedure, and number of spine levels involved were the features with the most impact on the model. CONCLUSIONS: Using ensemble learning-based predictive models, specifically XGBoost regression, can improve the accuracy of the estimation of spine surgery times.

A Machine Learning Solution to Predict Elective Orthopedic Surgery Case Duration.

We used surgery durations, patient demographic and personnel data taken from the East Kent Hospitals University NHS Foundation Trust (EKHUFT) over a period of 10 years (2010-2019) for a total of 25,352 patients that underwent 15 highest volume elective orthopedic surgeries, to predict future surgery durations for the subset of elective surgeries under consideration. As part of this study, we compared two different ensemble machine learning methods random forest regression (RF) and XGBoost (eXtreme Gradient Boosting) regression. The two models were approximately 5% superior to the existing model used by the hospital scheduling system.

Improving intraoperative handoffs for ambulatory anesthesia: challenges and solutions for the anesthesiologist.

Permanent transitions of care from one anesthesia provider to another are associated with adverse events and mortality. There are currently no available data on how to mitigate these poor patient outcomes other than to reduce the occurrence of such handoffs. We used data from an ambulatory surgery center to demonstrate the steps that can be taken to achieve this goal. First, perform statistical forecasting using many months of historical data to create optimal, as opposed to arbitrary shift durations. Second, consider assigning the anesthesia providers designated to work late, if necessary, to the ORs estimated to finish the earliest, rather than latest. We performed multiple analyses showing the quantitative advantage of this strategy for the ambulatory surgery center with multiple brief cases. Third, sequence the cases in the 1 or 2 ORs with the latest scheduled end times so that the briefest cases are finished last. If a supervising anesthesiologist needs to be relieved early for administrative duties (eg, head of the group to meet with administrators or surgeons), assign the anesthesiologist to an OR that finishes with several brief cases. The rationale for these recommendations is that such strategies provide multiple opportunities for a different anesthesia provider to assume responsibility for the patients between cases, thus avoiding a handoff altogether.

Timeliness of Operating Room Case Planning and Time Utilization: Influence of First and To-Follow Cases.

Resource and cost constraints in hospitals demand thorough planning of operating room schedules. Ideally, exact start times and durations are known in advance for each case. However, aside from the first case's start, most factors are hard to predict. While the role of the start of the first case for optimal room utilization has been shown before, data for to-follow cases are lacking. The present study therefore aimed to analyze all elective surgery cases of a university hospital within 1 year in search of visible patterns. A total of 14,014 cases scheduled on 254 regular working days at a university hospital between September 2015 and August 2016 underwent screening. After eliminating 112 emergencies during regular working hours, 13,547 elective daytime cases were analyzed, out of which 4,346 ranked first, 3,723 second, and 5,478 third or higher in the daily schedule. Also, 36% of cases changed start times from the day before to 7:00 a.m., with half of these (52%) resulting in a delay of more than 15 min. After 7:00 a.m., 87% of cases started more than 10 min off schedule, with 26% being early and 74% late. Timeliness was 15 ± 72 min (mean ± SD) for first, 21 ± 84 min for second, and 25 ± 93 min for all to-follow cases, compared to preoperative day planning, and 21 ± 45, 23 ± 61, and 19 ± 74 min compared to 7:00 a.m. status. Start time deviations were also related to procedure duration, with cases of 61-90 min duration being most reliable (deviation 9.8 ± 67 min compared to 7:00 a.m.), regardless of order. In consequence, cases following after 61-90 min long cases had the shortest deviations of incision time from schedule (16 ± 66 min). Taken together, start times for elective surgery cases deviate substantially from schedule, with first and second cases falling into the highest mean deviation category. Second cases had the largest deviations from scheduled times compared to first and all to-follow cases. While planned vs. actual start times differ among specialties, cases of 61-90 min duration had the most reliable start times, with neither shorter nor longer cases seeming to improve timeliness of start times.

An evaluation of operating room throughput in a stand-alone soft-tissue trauma operating theatre.

BACKGROUND: Operating room time is a limited, expensive commodity in acute hospitals. Strategies aimed at reduction of non-operative time improve operating room throughput and capacity. We conducted a prospective study to evaluate and augment operating room throughput and capacity using context-specific work practice changes. METHODS: Following institutional and ethical approval, an interdisciplinary group designed and introduced a series of work practice changes specific to a stand-alone soft tissue trauma theatre, comprising modifications to patient processing, staff behaviours and additional anaesthesiologist hours. Time intervals relating to each patient were measured during a 16 week period before and after implementing work practice changes. The primary outcome measure was non-operative time, with daily caseload and cancellations amongst secondary outcome measures. RESULTS: 251 procedures were included over 58 working days (8 to 17 Monday to Friday). Non-operative time [55.6 (31.1) vs 52.3 (9.8) minutes, p = 0.48], daily caseload [4 [1-9] vs 4 [2-7], p = 0.56], and the number of daily cancellations [3 [0-11] vs 5 [0-8], p = 0.38], did not differ between baseline and study phases. Regional anaesthesia for upper limb surgery increased during the study phase [26/59 (44.0%) vs 10/63 (15.9%), p = 0.014] with resultant decrease in mean duration of recovery room stay [20.7 (17.7) vs 30 (20.5) minutes, p = 0.0001] and increased recovery room bypass [26/116 (22.4%) vs 6/135 (4.4%), p = 0.0002]. Avoidable delays accounted for 124.8 (72.2) minutes of theatre time lost each day. CONCLUSION: In conclusion, additional attending anaesthesiologist hours combined with work practice changes did not impact on measures of theatre throughput and capacity. The study identified important variables that contribute to avoidable delays, and points the way for future research.