In the U.S., approximately 840,000 Americans die from cardiovascular disease (CVD) each year, and it is the leading cause of morbidity and mortality worldwide. The prevalence of CVD is on the rise and widespread disparities in CVD exist across economic, racial, and ethnic groups. In order to address the rising prevalence of CVD and persistent disparities, there has been a shift in focus to strategies promoting cardiovascular health (CVH) across the life course. CVH is a broader and more positive construct beyond the absence of CVD, and allows for clinical and public health strategies focused on disease prevention and health promotion, rather than solely on treatment once CVD develops. Despite this recent focus on improving CVH, widespread disparities still exist, and social determinants of health (SDOH) appear to be important contributors to these continued disparities. The World Health Organization (WHO) defines SDOH as the “structural determinants and conditions in which people are born, grow, live, work, and age.” There has been limited work studying how longitudinal exposures of SDOH change over time, perform in the prediction of CVH, and are associated with later-life CVH and CVD events. The primary objectives of this dissertation are to identify patterns of SDOH exposure over time by generating data-driven SDOH clusters using a novel machine learning method, and determine whether the addition of SDOH variables allowed for better prediction of an individual’s CVH status and were associated with mid-life CVH (a critical milestone in CVH maintenance) and CVD events. The primary hypothesis for this study is that a diverse set of SDOH from young adulthood through middle age will be predictive of mid-life CVH and will be associated with mid-life CVH and CVD events, independent of baseline CVH and other covariates. This dissertation begins with a general introduction highlighting the burden of CVD and the current evidence and methods used to link SDOH and CVD. I then present the three chapters included in this dissertation, all of which leveraged comprehensive longitudinal data from the Coronary Artery Risk Development in Young Adults (CARDIA) study cohort. Chapter 2 includes a detailed description of the methods used to identify frequent time-dependent SDOH patterns and generate the novel SDOH clusters in a well-phenotyped long-standing community-based study. This chapter demonstrates that the clusters generated improved the prediction of mid-life CVH. Chapter 3 presents a more detailed description of time-dependent individual- and neighborhood-level SDOH exposure patterns and shows an association between clusters and mid-life CVH, overall and by self-identified race groups. Chapter 4 evaluates whether the SDOH clusters are associated with mid-life CVD events before and after adjustment for mid-life CVH and subclinical CVD. The clusters are associated with mid-life CVD events, but not after adjustment for mid-life CVH and subclinical CVD. The three chapters support our original hypothesis. First, we can use a novel machine learning method to identify time-dependent SDOH patterns from young adulthood to middle age and create novel SDOH clusters of those patterns that provide insight into the complex inter-relationships of SDOH. Additionally, the clusters were predictive of mid-life CVH and associated with mid-life CVH and CVD events. Further refinement and validation of the clusters is necessary. The findings from this dissertation may be used to inform programs looking to develop targeted, timely, and multi-component interventions to address SDOH and improve CVH in young adults, with the potential to improve population CVH and reduce disparities.

Creator

• Zimmerman, Lindsay

DOI

• https://doi.org/10.21985/n2-8n5n-7g87

Subject

• Health sciences
• Public health

Language

• en

Alternate Identifier

• http://dissertations.umi.com/northwestern:15739
• etdadmin_upload_844529

Keyword

• data science
• informatics
• machine learning
• social determinants of health
• cardiovascular health

Date created

• 2021-01-01

Resource type

• Dissertation

Rights statement

• In Copyright