Alcohol outlets and alcohol consumption in changing environments: prevalence and changes over time

Background To examine whether changes in density of neighborhood alcohol outlets affected changes in alcohol consumption 1-year after regulatory changes increased alcohol availability. Methods Person-level data came from a population-based cohort (aged 21–64) residing in/around the Philadelphia, Pennsylvania metropolitan area (2016–2018, N = 772). Fifty-eight percent lived in a state that began implementing new regulations (Pennsylvania) and the remainder lived in states without major regulatory changes (Delaware and New Jersey). Alcohol consumption was assessed as days per week (pw), drinks pw, high consumption (≥8 drinks pw), and binge drinking. Availability of off-premise alcohol outlets was assessed using 1-mile density and distance. Regression models adjusted for age, gender, race/ethnicity, income, education, health status, state and population density. Results Cross-sectional analyses found that higher outlet density was associated with more alcohol consumption (days, drinks, high consumption; all p < 0.03) and residing farther from an outlet was associated with less alcohol consumption (days and drinks; all p < 0.04). In longitudinal analyses, relative to no change in outlets, exposure to more outlets was associated with 64% higher odds of drinking on more days pw (p < 0.049) and 55% higher odds of consuming more drinks pw (p < 0.081). However, the longitudinal association between changes in outlets and changes in consumption did not differ for residents in Pennsylvania vs. nearby states. In cross-sectional and longitudinal analyses, outlets were not related to binge drinking. Conclusion Off-premise outlets were associated with alcohol consumption consistently in cross-sectional analysis and in some longitudinal analyses. Results can inform future studies that wish to evaluate longer-term changes in increased alcohol availability and effects on consumption. Supplementary Information The online version contains supplementary material available at 10.1186/s13011-021-00430-6.


Supplement to main text description of the analytical sample
There were 2555 survey participants aged ≥21 who were asked the alcohol questions at baseline, and 802 of these participants responded to the follow-up survey (31% retention). Losses to follow-up were primarily due to no response after at least 7 call attempts (50%) or refusal at follow-up (hard refusal, requests to be added to do not call list, or other reason). Among the 802 participants who responded to the follow-up survey, 30 participants were excluded due to missing outcome or exposure data, leaving 772 participants in the cross-sectional analyses.
Despite low cohort retention at follow-up, those included vs. lost-to-follow-up were mostly similar (Supplement Table 1) and the analytic sample was comparable to the age, gender, race distribution in the census population in the areas where the cohort resided (U.S. Census Bureau, 2018). The exception was that the lost-to-follow-up sample had more participants with other race, lower income and lower education, and higher proportion of non-drinking. The lost-to-follow-up pattern was similar for Pennsylvania and non-Pennsylvania except that a larger share of non-Whites were lost-to-follow-up in non-Pennsylvania.  Participants aged ≥21 were asked if they consumed any type of alcoholic beverage in the past 30 days. Only past-30 days was asked in order to keep the phone survey focused/brief (and has been used by others (CDC, 2019)). If the participant answered yes' to past 30 day alcohol consumption, then subsequent questions were asked regarding alcohol consumption.

Supplement
Subsequent questions were asked regarding the number of days alcohol was consumed ("In the past 30 days, how many days per week or per month did you drink any alcoholic beverages?"), and the number of drinks consumed per day ("Thinking about the most recent occasion when you drank any alcoholic beverage, how many drinks did you have?"). A question was also asked about binge drinking, using the definition from the U.S. Substance Abuse and Mental Health Services Administration (SAMHSA, 2017) as drinking ≥5 alcoholic drinks for males and ≥4 for females on a single occasion ("Many people are able to drink X or more alcoholic beverages on an occasion. In the past 30 days, on how many days do you think you had X or more drinks of any alcoholic beverage on an occasion? An example of an occasion would be one evening.") Because Pennsylvania's relaxation of alcohol control primarily increased access to wine and beer, the survey also asked what type of alcohol was consumed in the past 30 days. However, results were not shown by alcohol type because 90% reported consuming both beer and wine and only 10% consumed soley spirits. When per-population was not part of the exposure measure, then the model also adjusted for population density within a 1.6 km area (operationalized into quartiles).
Here, exponentiated beta coefficients represents a relative value. Thus, in cross-sectional data the exp(beta) 1.35 can be interpreted as 35% higher drinking days per month when living in the highest quartile of outlet density (0.29-1.7 per 10,000 population) relative to the lowest quartile (the referent group).
c Odds ratios derived from logistic regression. High consumption relative to others in cohort (top quintile >= 8 drinks per week) Supplement Table 6. Longitudinal sensitivity analysis -sample subset to drinkers.
Multinomial regression, adjusted a within-person change in alcohol consumption (change in days, drinks, binge occassions) for an increase in off-premise alcohol outlets within a 1.6 km buffer (>0 increase in outlets b The exposure is a binary variable: increase in outlets vs. no increase in outlets (referent category) using the measure 'count of outlets in 1.6 km buffer'.
The category for 'decrease' in outlets was not included because very few participants experienced a decrease in outlets.
Per population standardization was not needed for the exposure variable in longitudinal model because the exposure was within-person change in outlet exposure and population density did not change much (because participants remained in their state). Nevertheless, we included population density (quartiles) as an adjustment variable in the model.  Figure 2. Adjusted cross-sectional estimates of exposure to off premise alcohol outlets (density of and distance to outlets) and number of drinks per day, stratified by gender.

Supplement
There was no strong evidence that there were cross-sectional differences by gender in the association between alcohol outlets (density or proximity) and drinking days or high alcohol use or binge in past 30 days (P for interaction > 0.09 and no evidence from longitudinal analyses that there were differences by gender in the association between change in outlet and change in consumption (p for interaction > 0.18). The exception to this was in cross-sectional analysis, the association between alcohol outlets (density or proximity) and drinks per week was stronger for males than females (P for interaction <0.0001, Supplement Figure 2). For example, males living in the highest density area (quartile 4 vs. 1) had 35% more drinks per week vs. females 24% more drinks per week (males expβ 1.35, 95% CI: 1.18, 1.54; females expβ 1.24, 95% CI: 1.04, 1.48). The association between living in an area farthest from an alcohol outlet and lower number of drinking days was only apparent for males (tertile 3 vs. 1 males expβ 0.62, 95% CI: 0.53, 0.72). whereas for females the association was positive (which was the inverse of our hypothesis, expβ 1.25, 95% CI 1.03, 1.51).
SD: standard deviation, P25: 25th percentile, P75: 75th percentile, Col = column, NIAAA: U.S. National Institute on Alcohol Abuse and Alcoholism a Includes participants who did not consume alcohol in past 30 days (N = 207 or 27% of the cohort) b Note that in main text Table 1, consumption is shown per month (whereas here we show consumption per week).