Predicting Patient Outcomes Using Machine Learning

Abstract

Surgical quality was historically measured based on objective endpoints like mortality, readmission to the hospital, and rates of postoperative complication. To produce predictions, the Random Forest approach mixes several decision trees that have been trained using bootstrapped samples.  The average prediction of every tree is the end result. These results can assist medical practitioners in making patient-centered choices and offering patients undergoing breast reconstruction individualized treatment. 1553 women who had mastectomy and breast reconstruction participated in a follow-up study to assess the machine learning models in more detail.  Two years later, 45.2% of patients were happy with the way their breasts looked, compared to 27.2% who weren't.  Following training, testing, and validation using the new data, the models demonstrated enhanced performance. Additionally, the authors contrasted their machine learning technique with NHS's hip prediction tool, which uses a linear regression model.