Development, Deployment, and Implementation of a Machine Learning Surgical Case Length Prediction Model and Prospective Evaluation.

OBJECTIVE: To implement a machine learning model using only the restricted data available at case creation time to predict surgical case length for multiple services at different locations. BACKGROUND: The operating room is one of the most expensive resources in a health system, estimated to cost $22 to $133 per minute and generate about 40% of hospital revenue. Accurate prediction of surgical case length is necessary for efficient scheduling and cost-effective utilization of the operating room and other resources. METHODS: We introduced a similarity cascade to capture the complexity of cases and surgeon influence on the case length and incorporated that into a gradient-boosting machine learning model. The model loss function was customized to improve the balance between over- and under-prediction of the case length. A production pipeline was created to seamlessly deploy and implement the model across our institution. RESULTS: The prospective analysis showed that the model output was gradually adopted by the schedulers and outperformed the scheduler-predicted case length from August to December 2022. In 33,815 surgical cases across outpatient and inpatient platforms, the operational implementation predicted 11.2% fewer underpredicted cases and 5.9% more cases within 20% of the actual case length compared with the schedulers and only overpredicted 5.3% more. The model assisted schedulers to predict 3.4% more cases within 20% of the actual case length and 4.3% fewer underpredicted cases. CONCLUSIONS: We created a unique framework that is being leveraged every day to predict surgical case length more accurately at case posting time and could be potentially utilized to deploy future machine learning models.

The found down patient: A Western Trauma Association multicenter study.

BACKGROUND: Unconscious patients who present after being "found down" represent a unique triage challenge. These patients are selected for either trauma or medical evaluation based on limited information and have been shown in a single-center study to have significant occult injuries and/or missed medical diagnoses. We sought to further characterize this population in a multicenter study and to identify predictors of mistriage. METHODS: The Western Trauma Association Multicenter Trials Committee conducted a retrospective study of patients categorized as found down by emergency department triage diagnosis at seven major trauma centers. Demographic, clinical, and outcome data were collected. Mistriage was defined as patients being admitted to a non-triage-activated service. Logistic regression was used to assess predictors of specified outcomes. RESULTS: Of 661 patients, 33% were triaged to trauma evaluations, and 67% were triaged to medical evaluations; 56% of all patients had traumatic injuries. Trauma-triaged patients had significantly higher rates of combined injury and a medical diagnosis and underwent more computed tomographic imaging; they had lower rates of intoxication and homelessness. Among the 432 admitted patients, 17% of them were initially mistriaged. Even among properly triaged patients, 23% required cross-consultation from the non-triage-activated service after admission. Age was an independent predictor of mistriage, with a doubling of the rate for groups older than 70 years. Combined medical diagnosis and injury was also predictive of mistriage. Mistriaged patients had a trend toward increased late-identified injuries, but mistriage was not associated with increased length of stay or mortality. CONCLUSION: Patients who are found down experience significant rates of mistriage and triage discordance requiring cross-consultation. Although the majority of found down patients are triaged to nontrauma evaluation, more than half have traumatic injuries. Characteristics associated with increased rates of mistriage, including advanced age, may be used to improve resource use and minimize missed injury in this vulnerable patient population. LEVEL OF EVIDENCE: Epidemiologic study, level III.