|Research in Population Health
|Year of Publication
|University of Pennsylvania
|Demographics, Health Conditions and Status, Healthcare, Methodology
This dissertation presents three chapters on population health and mortality risk using cause of death and underlying conditions data as well as longitudinal studies that include a wide spectrum of individual, household, and community variables. Although the three studies contained in this dissertation are not directly linked topically, they share a concern about the role of smaller components of a population to the mortality dynamics of the population as a whole. Each of the three papers builds on previous research to provide a level of detail or innovation in measurement to improve understanding of mortality dynamics and their effects on the population of interest. In the first chapter, I hold wealth rank in the poorest and richest global wealth quintiles constant over time to provide estimates of the deaths by primary cause group in 1970, 1990, and 2020. I also provide projections of deaths by cause for the richest and poorest countries in 2020 for twenty-one causes of death. The mortality structure among the poorest countries in the world continues to be dominated by infectious causes of death despite global estimates that suggest the epidemiologic transition has progressed to an advanced stage. The second chapter describes age-graded mortality risk for sub-groups of the U.S. population taking into account several demographic, economic, and health characteristics. Using data from different representative cohorts, I estimate age-specific relative risks of death associated with different health, socioeconomic, marital status, and behavioral characteristics to the mean risk of death. The third chapter describes a method for estimating variance in individual mortality risk based on observed characteristics using representative cohorts of the Health and Retirement Study (HRS). I compare distributions of predicted mortality risk to estimates of the degree of population-level heterogeneity in previous studies that have used parameterized frailty models for unobserved heterogeneity in mortality. Findings in this chapter show that the coefficient of variation for estimates of individual mortality risk decrease with age, thus suggesting that the inverse gaussian distribution may be more appropriate in frailty models than the gamma distribution, which implies a coefficient of variation that is constant with age.
|Endnote Author Address