Data mining techniques for analyzing stroke care processes

Controlled randomized clinical trials and meta-analyses show that stroke patients benefit from access to specialized Stroke Units, in terms of mortality, disability and dependency. However, many issues relating to stroke diagnosis and therapy and to the organization of stroke care remain to be solved and little is known about what interventions make Stroke Units more effective. It is also agreed that compliance with clinical practice guidelines improves health outcomes for these patients, but little is known about the relative weight of the different guideline recommendations. Over the last decade, many hospital- or population-based stroke registers have been set up with the aim of identifying specific key indicators able to monitor the quality and adequacy of acute stroke care. Registers seem to be adequate tools for collecting the data needed to analyze care processes, providing data useful for both national healthcare planning and scientific research. In this paper we applied data mining techniques to data collected within the stroke register of the Lombardia region in Italy. From our analyses both expected and unexpected results have been found: not always compliance to recommendations is related to a good patients' outcome.