Using KDD process to predict the duration of surgery

Combes, Catherine Vandamme, Jean-Philippe [FUCAM] Rivat, Céline Levecq, Philippe [FUCAM] Meskens, Nadine [UCL]

This paper presents a methodological framework based on data warehousing and knowledge discovery in database approaches to plan surgery in operating theatre suites. We suggest a decisional tool which estimates the appropriate duration for a patient in the operating theatre. To achieve this, we first describe a data warehouse model used to extract data from various, possibly non-interacting, databases. Then we compare two data mining methods: rough sets and neural networks. The aim is to identify classes of surgery likely to take different lengths of time according to the patient's profile. These tools permit to identify patients' profiles from administrative data, medical previous history, etc., but the surgical environment (surgeon, type of anaesthesia, etc.) is also token into account to estimate the duration of the surgery.