Leveraging Informatics to Optimize Cardiovascular Risk Prediction in Women

Cardiovascular diseases (CVD) are the number one cause of premature death in American women, but are highly preventable chronic illnesses. Compared with men, women are disproportionally at risk for developing CVD, which presents and progresses differently in women. However, female-specific risk factors (such as endometriosis and hypertensive disorders of pregnancy) for CVD are not included in clinical tools used to guide prevention and treatment of CVD. The overarching goal of this career development award proposal is to support the applicant in developing the necessary skills and training in informatics, epidemiology, and preventive cardiology to lead the advancement of risk assessment for CVD in women. Commonly used CVD risk-prediction tools are biased towards undertreatment for women. This may include the newly released CVD risk-prediction tool from American Heart Association called the PREVENT Equations, which do not include female-specific CVD risk factors. The overall research objectives of this application are to 1) test sex-differences in the predictive performance of the PREVENT Equations, 2) determine if female-specific risk factors are available in electronic health record data and if they accurately reflect participant-reported history, and 3) use best-subset statistical modeling to examine if adding female-specific CVD risks to traditional CVD risk factors improves our ability to predict CVD in women. The central hypotheses are that the PREVENT Equations are valid measures of CVD risk in men and women, female-specific data in the electronic health record is discordant from self-reported data, and that using robust statistical modeling will identify one or more female-specific CVD risk factors that improves model performance when added to traditional CVD risk factors. The first aim of the proposed research is to test the predictive performance of the PREVENT Equations using the NIH-supported All of Us Research Program data. Using sex-stratified data, discriminative statistics are expected to validate the PREVENT Equations predictive ability for time-to-first cardiovascular event among both women and men. The second aim is to assess the availability and quality of female-specific CVD risk factors in the All of Us electronic health record data using comparative statistics. The third aim is to measure the incremental value of female-specific CVD risk factors on the traditional CVD risk factors using the All of Us Research Program data. We will use best-subset logistic regression modeling to find the best fitting model for female-specific CVD risk prediction. The proposed research is significant in addressing gaps in clinical assessment of CVD in women, and innovative by engaging the relatively new NIH All of Us Research Program and leveraging informatics for accurate identification of CVD risks in the electronic medical record to improve women's CVD health. The proposed research and mentored training plan included in this award will support the applicant in their long-term career goal to become an independent investigator focused on reducing the burden of cardiovascular disease among women.