An Expanded Dataset to Study Psychosocial Influences on Health Disparities

DESCRIPTION (provided by applicant): Some researchers believe that non-medical determinants of health are much more important than medical factors, such as access to health care. Key among these non-medical determinants are putative health risk factors such as low social status (whether objectively measured or perceived), material deprivations, neighborhood characteristics, and negative thoughts, feelings, or beliefs. Because disadvantaged populations by definition lack access to material and social resources, the non-medical determinants of health may also be major contributors to health disparities. However, until the advent of the 2008 General Social Survey-National Death Index (2008 GSS-NDI), there has been no nationally-representative dataset that could be used to study the non-medical determinants of health. In the short 20 months since its public release, this dataset has been downloaded thousands of times and has been productive, exploring a wide array of research questions. The 2008 GSS-NDI consists of GSS surveys from 1978 to 2002 linked to mortality data through 2008. We now propose to expand the 2008 GSS-NDI to include the 2004-2012 GSS survey data, and detailed cause of death through 2013 (2013 GSS-NDI). In addition, we propose to create a geographically-coded GSS-NDI dataset linked with state data that will be used to assess the effect of local policies on health. The 2013 GSS-NDI and 2013 GSS-NDI-GEO will provide a wealth of additional sociological research questions, expanded statistical power, and will contain additional health and psychosocial variables, such as intravenous drug use. These two new datasets will contain 35 years of prospective mortality data and geographic information. With the extended statistical power and spatiotemporal reach of the new data, it is possible to execute age-period-cohort analyses, run instrumental variable analyses, employ hierarchical models, and exploit spatiotemporal variation in policy implementation, while all but eliminating survival bias. Our datasets therefore allow users to devise much more sophisticated, causally-oriented, statistical models than is possible using traditional prospective datasets. Thus, our datasets not only offer a 400 additional variables pertaining to health, material circumstances, sociological phenomena, beliefs, attitudes, and ideas, they also open the door to innovative epidemiological analyses. The previous 2008 GSS-NDI required a large up front effort because paper records containing identifiers had to be retrieved from warehouses and manually entered into databases. The present effort will be much less intensive because this up front work on the earlier records has already been done (and more recent records are already in electronic format), requiring only a small amount of funding.