UCLA

In 2020, one in ten people had diabetes in the United States. Despite the recent advancement of medical therapies, the prevalence of diabetes is still increasing, and thus more research is needed about the causal impact of diabetes and its related factors such as exercise and mental health on cardiovascular disease (CVD) and mortality. Particularly, despite the recent substantial focus on the culturally tailored and targeted approaches to improve cardiovascular health, the evidence is still limited among older Mexican Americans, a large racial/ethnic group in the US with a high prevalence of diabetes. Moreover, although elevated glycated hemoglobin (HbA1c) levels are well known to be associated with worse health outcomes, it has been under debate whether relatively lower HbA1c levels are beneficial or harmful for the long-term health outcomes among people without diabetes. Therefore, in this dissertation, I conducted the following three studies:

First, using a longitudinal cohort of community-dwelling older Mexican Americans, along with causal mediation analysis, I found that diabetes mediated around 10% of the association of low physical activity with all-cause mortality and CVD events. Second, using the same cohort, I found that diabetes and subsequent depressive symptoms had a synergistic effect on the increased risk for cardiovascular mortality after adjusting for time-varying confounders with a marginal structural model. Third, using a nationally representative sample of US adults, along with ensemble machine learning algorithms within g-formula, I found that adults having low HbA1c levels without diabetes were at an increased risk of all-cause mortality. These findings highlight the importance of i) public health interventions targeting diabetes prevention and management among older Mexican Americans who have difficulties increasing physical activity levels, ii) mental health management for older Mexican Americans after a diagnosis of diabetes, and iii) careful monitoring of low HbA1c levels to prevent early death among US adults.

I hope that this dissertation will contribute to not only better diabetes care but a better understanding of the usefulness of causal modeling to answer clinically important questions, and will encourage dialogue and an appreciation for data sciences by clinicians and epidemiologists in the dawn of the computational era.