Hospital scheduling of medical examinations constitutes a complex operational problem, characterized by high variability in procedure duration and the occurrence of no-shows, which compromises the efficient use of resources and the quality of the service provided. This work presents a case study based on real-world data from the Unidade Local de Saúde de Santo António (ULSSA), in Porto, where predictive models are developed to support the scheduling of medical examinations using exclusively ex-ante information, that is, information available at the time of scheduling. Examination duration is modeled as a supervised regression problem, while no-show analysis is addressed as a supervised Classification task. For both problems, different machine learning approaches were evaluated. The results demonstrate superior performance in predicting examination duration. In the case of no-shows, the findings reveal structural limitations associated with rigid binary Classification, reinforcing the usefulness of risk indicators to support planning decisions.
Touças et al. (Thu,) studied this question.