The effects of the social environment on health outcomes have been demonstrated in numerous research studies. However, the mechanisms or intermediate variables through which these effects occur are largely unknown. While methods for mediation analysis exist for the study of intermediate variables (or mediators), these methods are inadequate for handling the challenging features of social environment data. The proposed research will develop new mediation analysis methods using a potential outcomes and structural (causal) model framework. Our two-pronged approach, referred to as generalized causal mediation analysis, will provide new methods for both the fitting of causal models and the computation of mediation effects. In particular, we will extend previous methods by handling models with the following features: 1) mixed discrete and continuous variables, 2) multiple exposure variables, 3) heterogeneous mediation effects, 4) multi-level data, and 5) latent variables. Further extending the causal model framework, in which mediation has an intervention interpretation, we will develop a method to predicting the effect of a new intervention. This approach will involve user augmentation of a fitted causal model, thus synthesizing theory and empirical results. The methods will be applied to three motivating datasets: 1) family environment and dental caries data from 224 very low birth weight and normal birth weight adolescents and their parents;2) data from a study of an intervention to prevent STDs involving 1357 suburban high schools students;and 3) data from a study of oncologist-patient communication and its effect on patient decision satisfaction. The new methods will be evaluated and refined through the data applications and through simulation studies. Finally, the project will develop user-friendly computer programs to allow behavioral and other health researches to apply the new methods. The new methods will allow a more refined, valid, and thorough exploration of mediating pathways for social environment effects. They will also facilitate the formulation of effective behavioral and other interventions, utilizing knowledge about social environment mechanisms, to improve health outcomes.
This project will develop new statistical methods that will help researchers explain the relationship between social environments and health outcomes. The new methods, designed to model complex causal relationships, will be applied to data from dental caries, school STD prevention, and oncologist- patient communication studies. The improved understanding provided by the new methods will facilitate the design of effective new behavioral and other interventions for preventing and reducing the many diseases on which the social environment has an impact.
|Wang, Wei; Albert, Jeffrey M (2017) Causal Mediation Analysis for the Cox Proportional Hazards Model with a Smooth Baseline Hazard Estimator. J R Stat Soc Ser C Appl Stat 66:741-757|
|Albert, Jeffrey M; Geng, Cuiyu; Nelson, Suchitra (2016) Causal mediation analysis with a latent mediator. Biom J 58:535-48|
|Albert, Jeffrey M; Wang, Wei (2015) Sensitivity analyses for parametric causal mediation effect estimation. Biostatistics 16:339-51|
|Lee, Wonik; Kim, Seok-Joo; Albert, Jeffrey M et al. (2014) Community factors predicting dental care utilization among older adults. J Am Dent Assoc 145:150-8|
|Wang, Wei; Nelson, Suchitra; Albert, Jeffrey M (2013) Estimation of causal mediation effects for a dichotomous outcome in multiple-mediator models using the mediation formula. Stat Med :|
|Wang, Wei; Albert, Jeffrey M (2012) Estimation of mediation effects for zero-inflated regression models. Stat Med 31:3118-32|
|Nelson, S; Lee, W; Albert, J M et al. (2012) Early maternal psychosocial factors are predictors for adolescent caries. J Dent Res 91:859-64|
|Albert, Jeffrey M (2012) Distribution-free mediation analysis for nonlinear models with confounding. Epidemiology 23:879-88|