Evoluzione delle tecniche chirurgiche in proctologia: valutazione del rischio preoperatorio con modelli di machine learning

Machine learning is a component of artificial intelligence; it relies on computer algorithms and data analysis to learn patterns that exceeds the capacity of the human mind to comprehend. It uses statistical methods to infer relationships between predictors and outcomes in large datasets, and it has been successfully applied to predict adverse events in health care settings. In the preoperative phase, for risk stratification of patients with surgical conditions, various types of supervised learning have been used with large clinical databases. In our work, the aim was to investigate the potential role of machine learning (ML) versus classical statistical methods (SM) for the preoperative risk assessment in proctological surgery. We used clinical data from a nationwide audit: the database consisted of 1510 patients affected by Goligher's grade III hemorrhoidal disease who underwent elective surgery. We collected anthropometric, clinical, and surgical data and we considered ten predictors to evaluate model-predictive performance. The clinical target was the complication rate evaluated at 30-days follow-up. Logistic regression and three ML techniques were compared. ML models included a Decision Tree, a Support Vector Machine, and a classification Extreme Gradient Boosting (XGB). These methodologies could be used to develop a surgical risk calculator, which is already used and widespread for other diseases, that could help clinicians to estimate the chance of an unfavorable outcome after surgery.