Integration of Different Risk Assessment Tools to Improve the Event Risk Assessment in Cardiovascular Disease Patients
Authors
Abstract
Each year cardiovascular disease (CVD) causes over 1.9 million deaths in the European Union (42% of all deaths), and contributes to health costs with a total estimated of €169 billion. These unaffordable social and health costs tend to increase as the European population ages. In this context, the correct prognosis of cardiovascular disease is a key factor to defeat the current statistics.Some useful tools have been developed to predict the risk of occurrence of a cardiovascular disease event (e.g. hospitalization or death). However, these tools present some major drawbacks as they: i) ignore the information provided by other risk assessment tools that were previously developed; ii) consider (each individual tool) a limited number of risk factors; iii) have difficulty in coping with missing risk factors; iv) do not allow the incorporation of additional clinical knowledge; v) do not assure the clinical interpretability of the respective parameters; vi) impose a selection of a standard tool to be applied in the clinical practice; vii) may present some lack of performance.
This work aims to minimize the identified weaknesses, through the development of two different methodologies: i) combination of individual risk assessment tools; ii) personalization based on grouping of patients.
The former creates a flexible framework that is able to combine a set of distinct current risk assessment tools. The methodology is based on two main hypotheses: i) it is possible to implement a common representation (naïve Bayes classifier) of the individual risk assessment tools. Actually, current tools are diversely represented which does not facilitate their integration/combination. Moreover, these different representations are not suitable to deal with missing risk factors nor they can incorporate additional clinical knowledge; ii) it is possible to combine individual models exploiting the particular features of Bayesian probabilistic reasoning. The combination of individual models permits the creation of a global model that avoids the selection of a standard model as well as it can be adjusted to a specific population (optimized) through genetic algorithms operation.
The personalization based on the grouping of patients is proposed as an approach to enhance the performance of the risk prediction when compared to the one obtained with current risk assessment tools. This methodology is based on the evidence that risk assessment tools perform differently among different populations. Therefore, the main hypothesis that supports this methodology can be stated as: if the patients are properly grouped (clustered) it would be possible to find the best classifier for each patient.
Validation was performed based on three real patient datasets: i) Santa Cruz Hospital, Lisbon/Portugal, ACS-NSTEMI patients; ii) Leiria-Pombal Hospital Centre, Portugal, ACS-NSTEMI patients; iii) Castle Hill Hospital, Hull/U.K., heart failure patients.
Considering the obtained results it is possible to state that the initial goals of this work were achieved, which makes it a valid contribution for the improvement of the risk assessment applied to cardiovascular diseases. However, other research directions should be pursued in order to improve the proposed methodologies and respective results.