Short-Term Cardiovascular Disease Risk Assessment: Comparison of Two Combination Approaches

Cardiovascular disease (CVD) is the main cause of death in the world. Therefore, in order to achieve an adequate intervention for a given patient, the prediction of a cardiovascular event is of fundamental importance. Although, there are many tools for cardiovascular risk assessment available in clinical practice, the selection of the most proper model is not a simple task. Moreover, there are some limitations in these models, such as the incapacity to cope with missing information, under specific conditions they might present some lack of performance, and they may not facilitate the clinical interpretability. In order to mitigate these problems, we propose two different approaches to estimate the risk of a cardiovascular event: i) fusion of individual risk models based on a common representation (Naive Bayes classifier) of such tools; ii) personalization through a grouping strategy, where the most suitable risk tool is considered for each particular patient. We use a real dataset collected in Santa Cruz Hospital (Lisbon, Portugal) to validate and compare the two approaches. Promising results were obtained in both approaches achieving a sensitivity, specificity and geometric mean of (78:79%, 73:07% and 75:87%) and (75:69%, 69:79% and 72:71%), respectively.