A Regression Analysis for Predicting Surgical Complications

Abstract

A surgical complication is any undesirable and unexpected result of an operation. Surgical complications could be fatal to a patient if they are not detected earlier. One of the factors that could affect the severity of the complication is the time between a patient's diagnosis and the surgery. The patient might be at risk if the doctor misdiagnoses them or concludes that the patient has no severe symptoms. This paper aims to study the correlation between post-surgical conditions & time duration with possible surgical complications. Using regression analysis, the research intends to evaluate predictive possibilities of early discovery of these complications. The results reveal that the Gradient Boosting Regressor performs with minimal error rate and predicts almost all complications in line with the original data, measured across MAE, RMSE and R2 with scores of 0.07, 0.11 and 0.98 respectively. In comparison to Random Forest Regressor and Decision Tree Regressor, Gradient Boosting Regressor performs 70-80% efficiently across the three major aforementioned metrics on average. Thus, presenting itself as a valuable tool for finding the correlations in surgical data and early intervention of possible surgical complications.