Uncovering Life Course Constellations of Exposures through Big Data on Place, Time, and Family Factors Project Abstract This project will trace the mortality of birth cohorts of the early 20th century in the US by place, time, and family factors. Combining “big data” with a large array of contextual exposures, we substantially deepen our understanding of the complexities of how childhood exposures to disease, economic change, and natural disasters shape old age mortality profiles of cohorts born ~1910-1930. We fuse together hypothesis driven tests, data driven discoveries, and omnibus measures from variance decompositions. Our proposal combines the massive CenSoc data, which contains >15 million death records between 1975-2005 to test specific hypotheses as well as generate new hypotheses around the main, interactive, and cumulative effects of exposures during sensitive periods of development that may shape mortality experiences of these cohorts. Our interdisciplinary group of sociologists, demographers, economists, epidemiologists and others combines our expertise with the CenSoc data as well as with testing hypotheses from the Developmental Origins of Adult Health and Disease theories to pursue an interconnected set of specific aims to push forward the frontier of understanding the complex links between early life exposures and later life mortality. Aim 1 begins with a set of variance decompositions stratified by time and place across the early 20th century in order to construct an “Atlas” of estimates of the importance of family background (sibling correlations) as well as shared environmental factors (childhood neighbor correlations) determining old age mortality experiences at the close of the 20th century. We then ask whether these estimates are shaped by major disease events and the extent to which the patterns are explained through socioeconomic status markers in mid life. Aim 2 pivots from the forest to the trees by leveraging “natural experiment” research designs to estimate causal main and interactive effects of specific early life exposures and how these effects vary by sex, geography, and family background. We then make use of machine learning tools to synthesize estimates that may vary by age of exposure, sequence of exposures, and domain of exposures during early life. These models explore impacts of cumulative exposures, dynamic complementarity of exposures and potential for reversibility of early insults using well powered analysis not available elsewhere. Aim 3 concludes our analysis by pushing the frontier of intergenerational analysis by using data in previous aims to link back to parental information on exposures and ask whether parental exposures affect the next generation’s old age mortality as well as whether the effects of exposures interact across generations.