Coronary heart disease (CHD) is the leading cause of morbidity and mortality worldwide. In the United States, although mortality from CHD has declined over recent decades, the incidence of CHD has remained high. This year, nearly 600,000 Americans will experience their first myocardial infarction (MI) and a larger number will have undiagnosed ischemia. Both groups are at risk of developing ischemic cardiomyopathy (ICM), which compounds the burden of morbidity and mortality they face. Efforts to prevent CHD and its complications like ICM have been hampered by poor patient adherence, which is in large part due to misperceptions of risk. Patients are more likely to engage in preventive behaviors if they know their risk of disease, particularly if their risk is imminent. However, although several models estimate long-term risk of CHD over the next 10 years, there are no established models for predicting short-term risk of CHD within a one year time frame. Likewise, among patients with CHD, there are no existing models to assess the overall risk of coexisting ICM. To address these knowledge gaps in the field, I will build robust prediction models by applying the latest machine learning algorithms to the recently available large datasets comprising the UK Biobank and the Veterans Affairs (VA) cohorts.
In Aim 1, I will examine the 477,000 UK Biobank participants without preexisting CHD to develop a prediction model for one-year risk of new onset CHD. I will use several different machine learning algorithms, compare their respective predictive performances, and select the best-performing and most clinically applicable model. I will also compare the risk classification ability of this new model with existing long-term risk prediction models such as the Framingham and Atherosclerotic Cardiovascular Disease (ASCVD) risk scores.
In Aim 2, I will apply machine learning techniques to a large VA patient cohort consisting of over 300,000 patients with known coronary anatomy and left ventricular systolic function to develop a prediction model for the presence of ICM. In light of the two primary mechanisms for ischemic cardiomyopathy, irreversible myocyte loss from MI and hibernating myocardium from chronic ischemia, I will perform a stratified analysis by history of prior MI to determine if the risk factors for ICM vary depending on whether the pathological mechanism is infarction versus ischemia. An improved understanding of the predictors of CHD and ICM will lead to insights into the biological basis of ischemic heart disease. Moreover, accurate identification of individuals who are at imminent risk for CHD and its complication, ICM, is of central importance to efforts in preventing cardiovascular disease.
Coronary heart disease and its complications such as ischemic cardiomyopathy are the global leading cause of morbidity and mortality, and an improved ability to assess risk along this continuum of ischemic heart disease is urgently needed. I will apply the latest machine learning techniques to the recently available large datasets comprising the UK Biobank and VA cohorts to develop robust prediction models for short-term risk of coronary heart disease and overall risk of ischemic cardiomyopathy. My work is anticipated to yield important insights into the biological basis of ischemic heart disease and to improve preventive measures for public health.
