ETD

• Healthcare
• machine learning
• management
• waiting lists

Healthcare waiting lists represent an increasingly critical management challenge for public health systems operating with limited resources, a problem that has intensified significantly in recent years. Traditional analytical approaches struggle to address this complexity: the convergence of supply constraints, organizational factors, demographic pressures and systemic shocks creates intricate dynamics that resist conventional single-discipline methods. This thesis applies machine learning techniques to analyze dermatology appointment waiting times in Tuscany's regional healthcare system, leveraging data access obtained through an internship at the Regional Health Directorate. Working with administrative data comprising 126,523 prescription-visit pairs from 2024, the research employs XGBoost algorithms combined with SHAP interpretability analysis within a rigorous methodological framework including nested cross-validation and stratified sampling. Results reveal temporal factors, particularly prescription timing and contact dates, as the strongest predictors of waiting times, followed by medical priority codes and geographic variables. The primary contribution is methodological rather than substantive, demonstrating how interdisciplinary approaches integrating economics, management science, computer science and medical knowledge can effectively address complex healthcare policy problems that traditional methods cannot adequately capture, establishing a replicable framework applicable beyond dermatology services.