The use of a machine learning model may aid in accurately predicting surgical procedure time, according to the results from a randomized clinical trial.
Researchers found that improving planned surgery duration accuracy reduced the use of presurgical resources, patient wait times, and presurgical length of stays, all without increasing surgeon wait times between cases.
Although the study was limited to patients undergoing colorectal and gynecological surgery, the researchers, led by Christopher T. Strömblad, MS, Department of Strategy and Innovation, Memorial Sloan Kettering Cancer Center, New York City, concluded that their predictive model could be generalizable beyond specific procedures and services. The findings were published in JAMA Surgery.
“Planning the OR schedule for a mean of more than 100 cases per day is an endeavor filled with uncertainty,” Strömblad and colleagues pointed out. And these areas of uncertainty include whether the first case is able to start on time, turnover time between cases, cases booked close to the day of surgery, and planned versus actual surgical case duration.
They further noted that studies that have focused on efforts to improve reliable time estimates of surgical procedures show that this can lead to improved efficiency in the operating room. Here, the authors collected patient, procedure, surgeon, and systems data in order to train a machine learning model for colorectal and gynecology procedures at Memorial Sloan Kettering Cancer Center.
The study enrolled 683 patients undergoing colorectal or gynecology procedures, and their cases were randomized to either the machine learning-assisted surgical prediction group, or the control group.
The primary outcome measure was the accurate prediction of the duration of each scheduled surgery as measured by (arithmetic) mean (SD) error and mean absolute error. Strömblad and colleagues found that patients whose cases were assigned to the machine learning algorithm had significantly lower mean (SD) absolute error compared to the control group (49.5 [66] minutes versus 59.3 [72] minutes for a difference of −9.8 minutes).
According to the authors, this reduction in schedule duration error meant that patients spent less time in the presurgical area, with a mean start-time delay for following cases reduced by 62.4 minutes (from 70.2 minutes to 7.8 minutes) and 16.7 minutes (from 36.9 minutes to 20.2 minutes) for patients receiving colorectal and gynecology surgery, respectively. The overall mean reduction of wait time was 33.1 minutes per patient (from 49.4 minutes to 16.3 minutes per patient).
Strömblad and colleagues also determined that patients who were assigned to the machine learning algorithm spent less time in the facility than control patients (148.1 minutes versus 173.3 minutes), suggesting there is “a potential benefit vis-à-vis available resources for other patients before and after surgery.”
The authors further pointed out that these results (reduced error, and improved wait time and resource usage metrics) were accomplished without adversely affecting surgeons’ wait time, as there was a small improvement of 1.5 minutes in time between the end of cases and start of to-follow cases using the model, compared with the control group.
“With implementation of the predictive model, we were able to not only minimize the underpredicted skew of the estimates’ error distribution but also reduce the absolute prediction error; the result was a substantial reduction in patient wait time, without an increase in surgeon wait time,” Strömblad and colleagues observed.
In a commentary accompanying the study, Alik Farber, MD, MBA, Department of Surgery, Boston Medical Center, Boston University School of Medicine, noted that while associated patient experience scores were not reported in this study along with the finding that overall patient wait times were significantly reduced, “one wants to assume that waiting less would be welcome by most patients.”
Farber added that the increased emphasis on patient-centric care means that approaches that improve operational efficiency and patient experience — like the one followed by Strömblad and colleagues – is becoming more relevant.
“Surgeons need to continue using innovative management science techniques to actively improve the operating room environment,” wrote Farber. “Stakeholders that stand to benefit include the hospital, the surgeons, and, most importantly, the patients.”
-
Machine learning can be used to better predict the duration of surgical procedures, according to results from a randomized clinical trial.
-
This predictive model could improve case duration accuracy, presurgical resource use, and patient wait time, all without increasing surgeon wait time between cases.
Michael Bassett, Contributing Writer, BreakingMED™
Trömblad and Farber reported no disclosures.
Cat ID: 159
Topic ID: 97,159,728,791,730,192,925,159
