Sexually transmitted infections (STIs) have increased steeply in the United States (US) since the beginning of the 21st century. At the same time the US population has become less religious, with growing numbers of individuals especially in the younger age groups reporting no religious affiliation. As patterns of STIs and religious belief have changed several studies exploring the relationship between these two areas have been conducted with mixed results. This dissertation adds to the existing literature by exploring a suspected predictive relationship between religion and notifiable STI outcomes at the substate level in the US.In the first aim, predicted estimates of substate region-specific religiosity levels were derived from four religion items assessed by the 2002-2011 National Survey on Drug Use and Health (NSDUH). I then aggregated STI case numbers from 2012-2019 at the NSDUH-defined substate region-level and used the religiosity estimate derived in aim one to examine the role of religiosity in predicting STI outcomes for the three nationally notifiable sexually transmitted infections individually and in aggregate. In aim two I used this data to examine the effect of population size and age distribution of the substate regions on the ability of religiosity to predict STI outcomes. The final aim explores the effect of the addition of supplemental covariates to each of the four models first individually then together to build a final best fit model for each outcome. In the first aim I confirmed that the four religion items assessed by NSDUH can be used to predict a single latent dimension that I have called ‘religiosity’. In the second aim I assessed the ability of religiosity alone to predict STI outcomes. I then added covariates for population size and the age distributions of the substate regions in 2002-2011 into the crude model which showed that both variables improved the fit of the religiosity predictive model for the 2012-2019 STI outcomes. Finally, in my third aim I considered several additional covariates for inclusion in the model and evaluated whether they might improve the fit of the STI prediction model restricted to terms for religiosity level, population size, and age distributions. Following this I created predictive models with plausible and unlikely predictors to arrive at a best fit model for each outcome under study. While the best fit models for each outcome varied somewhat, several covariates qualified for inclusion in all models including the substate region population, the proportion of 26- to 35-year-olds, and the extra-medical use of a set of drug and medicine sub-types identified in NSDUH modules on this topic. Taken together, the results of this dissertation point towards three conclusions. First, religiosity may be an important predictor of STI outcomes in the US and warrants further consideration. Second, while this project does not draw causal conclusions about the relationship between religiosity and STI outcomes it suggests that a modeling approach like the one used here may be useful for future study and to better target public health programming to high-risk populations. Finally, several covariates assessed as part of this project might prove to have importance in future predictive models for STI occurrence in community and substate regions of the United States. Future studies should build upon these models by considering additional covariates as well as attempting to replicate these results with alternative or expanded data sources.
Read
