A growing literature has explored the relationship between life course factors and health in later life. In particular, childhood circumstances have been shown to be associated with several health outcomes many decades later. Most of this literature has focused on high-income countries and on associations between single events or exposure variables in childhood and individual health outcomes among older adults. Less is known about the combined effect of harmful exposure in childhood on later-life health outcomes and, especially, in lower- and middle-income countries. Also, imperfect strategy identifications in observational studies make difficult to assess the robustness of these associations.

This dissertation aims to address this gap in the literature with three separate studies on the relationship between childhood disadvantage and several later-life health outcomes among Brazilians aged 50 and over using data from the baseline assessment of the Brazilian Longitudinal Study of Aging. The first study sheds light on the associations between several potentially harmful exposures in childhood (separately and combined) and the occurrence of chronic conditions, separately and together (multimorbidity). The second study focuses on the impact on distinct theoretically appropriate domains of childhood disadvantage on three measures of cognitive performance. These first two studies also investigate the potential mediation effect of adulthood socioeconomic status (SES) on these relationships. And the last study discusses the appropriateness of a novel sensitivity framework (SenseMakr) to analyze observational studies by analyzing the robustness of 65 hypothesized associations between individual exposure and outcome variables.

Results from the first study shows that a childhood disadvantage scale was associated with 8 different chronic conditions as well as the total count of chronic conditions. Mediation analyses suggest that part of the effect of childhood disadvantage (10%) on multimorbidity is mediated by higher SES in adulthood, while extensive sensitivity analyses suggest that omitted confounding is very unlikely. In the second study, we found that childhood disadvantage is associated with low performance in memory tests and semantic verbal fluency tests among older Brazilians. Adulthood SES fully mediated the association between all domains of childhood disadvantage and memory performance and only partially mediated its association with verbal fluency. The last study found that out of the 65 possible associations between single exposure and outcome variables, 24 were found to be statistically significant. Although the SenseMakr framework does not provide thresholds to support any mechanistic conclusion as it is not possible to establish universal cutoff values for its robustness measures, this approach can be highly useful in other observational studies of aging and the life course.