Drinkable Neighborhoods? Identifying Latent Neighborhood Subtypes Related to Alcohol Misuse

Introduction: Person-centered statistical approaches such as latent class or profile mixture models have been increasingly used in prevention research to identify subgroups of individuals according to multiple characteristics. This study extends such models to classify neighborhoods into discrete categories based on five neighborhood characteristics. We also examined cross-sectional associations between derived neighborhood classes and frequency of past year heavy episodic drinking (HED) and prevalence of alcohol use disorder (AUD) among adults.

Method: The sample consisted of 536 adults, ages 32 to 34 years, living in King County, WA, assessed at a single wave as part of the Seattle Social Development Study. At the time of the survey, the participants were living in 303 different census block groups (the neighborhood units). For the neighborhood-centered mixture models, indicators included Census-based socio-demographic characteristics (percent living in poverty and percent White race) and a measure of perceived neighborhood social disorder as continuous variables, and tax parcel measures of density of liquor stores and bars within the block group as high vs. low categorical variables.

Results: Fit indices supported a 5-class solution as the best fitting model. The 5 classes could be described as: 1) high poverty/high disorder/mixed race/low alcohol outlet density (4%); 2) moderate poverty/moderate disorder/mixed race/low outlet density (12%); 3) low poverty/low disorder/high White racial composition/low outlet density (69%); 4) low poverty/high disorder/mixed race/low outlet density (5%), and 5) low poverty/low disorder/high White racial composition/high outlet density (10%). Adjusted for individual characteristics including race and annual household income, there were statistically significant differences in HED across the five neighborhood classes. Negative binomial models showed that compared to those living in Class 1 (high poverty/high disorder/mixed race/low outlet density) neighborhoods, those in Class 3,4 and 5 neighborhoods reported significantly less frequent past year HED. Further, log-binomial models showed that compared to those in Class 1, residents of Class 5 neighborhoods had a marginally significant lower likelihood of AUD.

Conclusions: Neighborhood subtypes derived from finite mixture models may represent meaningful categories that can help identify residential areas at elevated risk for alcohol misuse. This study suggests that neighborhoods more characterized by socioeconomic disadvantage and disorder may be more problematic contexts in regards to alcohol misuse than those characterized by availability of alcohol outlets.