Grants Funded

Abstract

Project Summary: Complications occur commonly after breast reconstruction resulting in patient distress, suboptimal aesthetic outcomes, higher health care costs, and delays in oncologic treatment. Previous studies have developed nomograms to predict surgical risks based on large databases. Although these studies have been helpful and can help guide clinical care, current risk prediction models rely on the use of logistic regression that may over or underestimate risks for some patients. Machine learning (ML), a subset of artificial intelligence, involves developing and employing algorithms capable of identifying correlations and patterns of complex risk factors regardless of statistical significance. Using non-linear techniques, ML can "learn" patterns between complex variables and help develop more accurate models to predict outcomes after breast reconstruction. In preliminary studies, we have used ML to develop robust, comprehensive, integrated predictive models using our institutional database. These models demonstrated higher performance compared with previously reported logistic regression nomograms. In addition, these models recognized the non-linear multifactorial nature of adverse events. The purpose of this proposal is to externally validate our models using additional data and compare their performance to previously reported logistic regression models. ML models will identify patients who are at high risk of developing complications using non-linear predictors and help guide surgical planning and patient education. Our central hypothesis is that ML models provide superior outcomes prediction for breast reconstruction patients compared to available statistical models. The rationale is that use of logistic regression in current predictive models limits their effectiveness, predictive power, and generalizability. Our approach is distinguished by the use of AI-driven predictive models that overcome traditional risk prediction barriers by incorporating non-linear algorithms capable of accurately predicting clinical outcomes. We will test our central hypothesis and achieve our objective with the following specific aims: Aim 1: Externally validate and augment ML predictive models for breast reconstruction outcomes. Aim 2: Compare the performance of ML models to current risk assessment tools models in predicting breast reconstruction outcomes.

Impact Statement: This proposal will deliver artificial intelligence predictive models to identify patients who are at high risk of developing complications using non-linear predictors and help guide surgical planning and patient education. Thereby these ML models will significantly improve breast reconstruction outcomes; resulting in a significant advancement and positive impact for breast reconstruction. The product of this project will be used in a future proposal for federal funding to integrate these models into a clinical decision support-mobile user interface and assess their impact in a before-after, prospective observational study evaluating baseline practice patterns, patient and surgeon satisfaction with care, use of preoperative optimization, changes in surgical planning, and complication rates.

Biography
Jonas Nelson, MD, MPH, is Assistant Attending in the Plastic and Reconstructive Surgery Service in the Department of Surgery at Memorial Sloan Kettering Cancer Center (MSK). Dr. Nelson specializes in reconstruction following surgery for breast cancer and head and neck cancers. He has expertise in microsurgical breast reconstruction, head and neck reconstruction with free-tissue transfer, and complex abdominal wall reconstruction. His research focuses on patient-reported outcomes after reconstruction. Through a better understanding of what leads to long-term satisfaction following reconstructive procedures, Dr. Nelson aims to improve preoperative counseling to guide patient decision-making and manage expectations. After receiving his medical degree from the University of Maryland School of Medicine, Dr. Nelson completed residency at the University of Pennsylvania School of Medicine and a Breast Reconstruction and Microsurgery fellowship at MSK. Throughout his training, he distinguished himself with receipt of several honors and awards, including a Clinical Research Fellowship from the Doris Duke Charitable Foundation (2008) and induction into the Alpha Omega Alpha Medical Honor Society in 2010. To further refine and expand his skills in patient-reported outcomes research, Dr. Nelson recently obtained a Master of Public Health degree from Johns Hopkins Bloomberg School of Public Health. He has published extensively (135 publications; h index: 29, 41 as first author).