Mediators and moderators of a neighborhood experiment on alcohol use

Investigator: Maria Glymour, ScD, MS
Sponsor: University of Minnesota

Location(s): United States


Excessive drinking in adolescence disrupts healthy development and leads to adverse social, economic and health consequences later in life. Although existing research focuses primarily on individual-level causes, neighborhood context is increasingly recognized as an upstream determinant of alcohol use. Yet since most prior neighborhood research on health has not been experimental, it is potentially biased, and weakens the foundation for policy translation. Our study proposes a secondary analysis of newly-available data from a social experiment of voluntary neighborhood relocation using housing vouchers in 5 cities (the Moving to Opportunity, MTO, Study), to test whether, how, and why random assignment of an offer to move to a low- poverty neighborhood over a 15 year period influenced the alcohol use, excessive drinking, and alcohol dependence of adolescents and their mothers. We will complement existing data with additional neighborhood- level data to enrich the dataset and test multilevel hypotheses. MTO is the only available large-scale study that has randomly assigned families to receive different neighborhood housing contexts, to support strong causal inferences of how early-life neighborhood context shapes alcohol use. When launched, MTO was one of the most promising policies to reverse the damaging effects of neighborhood poverty for minority families. But health researchers have had limited access to MTO data, compromising the scientific payoff of the $70+ million investment. Since residential mobility experiments like MTO are rare, expensive, and since MTO profoundly affected substance use, it is crucial to probe its effects. We propose new secondary data analyses with newly available 15 year follow-up data from this RCT, using novel methods to test mediation and moderation, applied to alcohol outcome measures that have not been examined prior. We apply innovative methods for assessing mediation (i.e., weight-based causal methods) and effect modification (i.e., machine learning techniques). Aside from the RCT design for strong causal inference of how early-life neighborhood changes influence excessive drinking, we also leverage one of the few longitudinal intervention designs sampling both mothers and their children, to illuminate familial transmission of drinking as it emerges in adolescence. Our R21 project builds on a productive, interdisciplinary team, experienced with the MTO study, and proposes 3 aims: to test how moving to a low-poverty neighborhood in the MTO experiment influenced excessive drinking and alcohol dependence; to examine if the MTO experimental effect on excessive drinking and alcohol dependence was less beneficial for more vulnerable families; to test whether effects of moving to a low-poverty neighborhood in MTO for excessive drinking and alcohol dependence were mediated by individual-level stress; family-level context; or neighborhood-level stress, alcohol availability, or alcohol norms. The results will enrich our understanding of neighborhood and family context as a cause of excessive drinking and alcohol use in adolescents and adults, and inform the next generation of alcohol prevention policy.