Machine learning algorithms can use preoperative clinical data to predict complications of abdominal wall reconstruction
MONDAY, April 11, 2022 (HealthDay News) — Machine learning (ML) models can help predict outcome following abdominal wall reconstruction (AWR), according to a study published online April 7 in the Journal of the American College of Surgeons.
Abbas M. Hassan, M.D., from The University of Texas MD Anderson Cancer Center in Houston, and colleagues developed, validated, and evaluated ML algorithms for predicting complications following AWR. Data were included from 725 patients, who were divided into training and testing sets (80 and 20 percent, respectively).
The researchers found that the rates of hernia recurrence (HR), surgical site occurrences (SSOs), and 30-day readmission were 12.8, 30, and 10.9 percent, respectively. ML models had good discriminatory performance for predicting HR, SSOs, and 30-day readmission, with areas under the receiver operating characteristic curve of 0.71, 0.75, and 0.73, respectively. The mean accuracy rates for the ML models were 85, 72, and 84 percent for predicting HR, SSOs, and 30-day readmission, respectively. ML identified and characterized four, 12, and three unique significant predictors of HR, SSOs, and 30-day readmission, respectively. ML models had a superior net benefit regardless of the probability threshold in a decision curve analysis.
“We believe the models can be improved and made to be more generalizable in subsequent iterations, and we’re currently embarking on a multicenter study to validate the models and develop a first-of-its-kind integrated tool that uses these models and clinical data and imaging data to provide a robust prediction tool,” Hassan said in a statement.
One author disclosed ties to Allergan; a second author disclosed research funding from Blue Cross Blue Shield.
Copyright © 2022 HealthDay. All rights reserved.
