Beyond the prediction of health care costs related to dyslipidemias and other cardiometabolic risk factors: explainable analysis through causal structure learning and inference algorithms

Abstract

Cardiovascular diseases (CVD) are major causes of mortality worldwide, leading to pre- mature deaths and loss of quality of life for populations in several countries. They also impose substantial socioeconomic toll for individuals, communities, and health systems. Dyslipidemias, characterized by alterations in lipid levels, represent risk factors significantly associated with CVD due to their involvement in the pathophysiology of atherosclerosis and result in substantial burden to national health systems. There is robust evidence on the role of metabolic and genetic markers as risk factors for CVDs and dyslipidemia. Nonetheless, the exact pathways and contributions that each of these markers might have on lipid traits and health costs, and how they relate to each other in terms of causal relationships is still to be unraveled. Developing and employing statistical and computer science methods that address causal relationships in this particular data is a challenging task with the potential of facilitating and accelerating research progress. In order to achieve this, Structure Learning and the Fast Causal Inference (FCI) algorithm will be used to facilitate the translation of this knowledge into the clinic. This project represents an opportunity to explore new data and causal inference approaches in the Brazilian context, and will enable simulation of future scenarios in the health system for disease prevention, treatment and control. (AU)