American Association of Plastic Surgeons
AAPS Home AAPS Home Past & Future Meetings Past & Future Meetings

Back to 2022 Posters


Development and Assessment of Artificial Intelligence-Driven Models to Predict Hernia Recurrence, Surgical Complications, and 30-Day Readmission Following Abdominal Wall Reconstruction
Abbas M. Hassan, MD, Sheng-Chieh Lu, PhD, Malke Asaad, MD, Anaeze C. Offodile, MD, MPH, Christopher Sidey-Gibbons, PhD, Charles E. Butler, MD.
University of Texas MD Anderson Cancer Center, Houston, TX, USA.

PURPOSE: Despite advancements in AWR techniques, hernia recurrences (HR), surgical site occurrences (SSOs), and unplanned hospital readmissions persist. We sought to develop, validate, and evaluate machine learning (ML) algorithms for predicting complications following abdominal wall reconstruction (AWR).
METHODS: We conducted a comprehensive review of patients who underwent AWR from March 2005 to June 2019. Nine supervised ML algorithms were developed to predict HR, SSOs and 30-day readmission. Patient data was partitioned into training (80%) and testing (20%) sets. Test data was blinded from the ML models until training was completed
RESULTS: We identified 725 patients (52% women), with a mean age of 6011.5 years, mean body mass index of 317 kg/m2, and mean follow-up time of 4229 months. The HR rate was 12.8%, SSOs rate was 30%, and 30-day readmission rate was 10.9%. ML models demonstrated good discriminatory performance for predicting HR (area under the receiver operating characteristic curve [AUC]=0.71), SSOs (AUC=0.72), and 30-day readmission (AUC=0.70). ML models achieved mean accuracy rates of 88% (95% confidence interval [CI], 83-92%), 70% (95% CI, 62-78%), and 81% (95% CI, 73-87%) for predicting HR, SSOs, and 30-day readmission, respectively. ML identified and characterized 12 unique significant predictors of HR, 6 of SSOs, and 8 of 30-day readmission
CONCLUSION: ML algorithms trained on readily available perioperative clinical data accurately predicted complications of AWR. Our findings support incorporating ML models into the perioperative assessment of patients undergoing AWR to provide data-driven, patient-specific risk assessment.


Back to 2022 Posters