Community alcohol sales and opioid poisoning deaths: Alcohol serving space as a harm reduction opportunity

Abstract

The concurrent use of opioids and alcohol is particularly dangerous for individuals. Alcohol is commonly seen in opioid overdose death toxicology reports and, concurrent use of alcohol and opioids is often reported by individuals across a diverse range of opioid use profiles. This study investigates whether there is a community-level relationship between alcohol sales and opioid-related overdose deaths to inform the situating of harm reduction efforts in spaces most likely to reduce substance-related harms. Using an ecological design, zip code-level data for New Hampshire were combined from the US Census Bureau’s American Community Survey (sociodemographics), the National Alcohol Beverage Control Association (alcohol retail sales), and the NH Office of the Chief Medical Examiner (zip code level opioid poisoning deaths) to investigate the relationship between alcohol sales and opioid-related poisoning deaths at a community level in a state with the third highest rate of opioid poisoning deaths for the year the current study represents. Using a spatial error regression model approach, opioid-related poisoning deaths were higher in zip codes with greater population density and on-premise alcohol sales and were lower in zip codes with greater off-premise alcohol sales and area disadvantage. The findings here co-locate higher levels of on-premise alcohol sales and opioid-related poisoning deaths at a community-level, mirroring individual-level findings on the danger of mixing these two substances. Results inform harm reduction approaches by identifying substance use spaces where overdose prevention messaging or policy change may be most effective.

Introduction

Opioid-related overdose deaths remain a critical issue in the United States. Among the 67,367 drug overdose deaths in the US in 2018, 69.4% of those deaths were opioid-related [1]. Programmatic and policy responses to the high number of opioid-related overdose deaths have often been focused on the community level and include strategies such as expansions of medication assisted treatment, naloxone availability, and prescription drug monitoring programs [2,3,4]. Research has investigated the sociodemographic correlates of these deaths to guide community responses to the opioid epidemic. A mix of community and individual factors have emerged as important predictors of opioid-related deaths. At the community level, areas with lower median age, greater population density, and higher economic disadvantage have been shown to be more likely to have higher rates of opioid-related deaths [5,6,7,8]. Among individual risk factors for opioid-related poisoning deaths, community reentry following incarceration, presence of a substance use disorder (SUD), discontinuation of SUD treatment, and the use of opioids in combination with alcohol, benzodiazepines, or cocaine have been shown as key predictors of an increased risk for opioid poisoning deaths [9,10,11].

Concurrent use of alcohol and opioids is relatively common across a variety of opioid use profiles [12,13,14]. Among individuals reporting extra-medical use of prescription opioids, 52% reported concurrent alcohol use, with 26% reporting concurrent use with opioids every time, and 26% reporting concurrent use at least 1–3 days a week [12]. In a sample of individuals who used prescription opioids daily, 15% used alcohol concurrently with their opioid prescriptions and were also likely to be classified as risky drinkers [13]. Lastly, in a community sample of individuals who experienced chronic pain, 23% were misusing opioids and engaging in risky alcohol use, as evidenced by the Current Opioid Misuse Measure and the Alcohol Use Disorders Identification Test respectively [14]. Opioids and alcohol are both central nervous system depressants, concurrent use can result in slowed breathing and lowered pulse/blood pressure. Opioid poisoning events turn fatal when the central nervous systems’ respiratory function is suppressed enough to result in respiratory failure [15, 16], so concurrent use of opioids and other CNS depressants like alcohol can be particularly lethal [10, 11, 15, 16].

Alcohol is often detected in toxicological reports of opioid-related overdose deaths [9, 15, 17]. In roughly 50% of heroin-related overdose deaths, and a third of methadone-related deaths, alcohol is detected [18]. More general examinations of the prevalence of alcohol in all opioid-related deaths have found around 20% of overdose deaths related to the concurrent use of alcohol and opioids [15]. Combining opioids and alcohol is especially risky because overdose and poisoning deaths when co-using alcohol and opioids have been shown to occur when blood alcohol concentrations are low (i.e., below 0.08) and blood concentrations of opioids are otherwise non-fatal [19, 20]. Individuals who understand safety limits related to the use of alcohol or opioids separately can make fatal miscalculations when co-using because of how these substances interact in the body. This body of research indicates that the level of risk associated with concurrent use of opioids and alcohol is high and interventions, such as targeted informational and messaging campaigns designed to promote awareness of these dangers, may be important harm reduction strategies.

The current study investigates whether rates of opioid-related overdose deaths are related to alcohol sales at a community level of analysis. Although no study to date has empirically tested this relationship, previous research has demonstrated higher densities of off-premise alcohol retailers as predictive of higher community rates of opioid overdoses [21]. We extend the work on alcohol outlet density by using the volume of alcohol sold in a community rather than density of retailers to predict substance related harms. Using sales data is considered here as representative of community interaction with a substance use space. If a relationship between rates of opioid-related poisoning deaths and alcohol sales is found to exist, this study may suggest an opportunity for targeted community interventions focusing on policy change or public messaging related to the dangers of concurrent alcohol and opioid use within the alcohol retail environment. Utilizing changes in alcohol retail policy has previously been shown as an effective way to reduce alcohol-related harms, community-level interventions limiting alcohol sales can reduce excessive alcohol consumption and alcohol-related violence [22, 23].

Using previous research on the community factors associated with opioid-related deaths, The current study extends the work on factors associated with opioid poisoning deaths by moving beyond individual-level and location density examinations to consider the impact of the volume of alcohol sold at a community level. We hypothesized that postal code rates of opioid poisoning deaths would be positively associated with the volume of alcohol sold for on-premise (i.e., bars, restaurants) and off-premise (i.e., liquor stores) alcohol outlets after controlling for area-level sociodemographics.

Methods

Data for this study were combined from several sources to include information on the number of opioid-related overdose deaths, alcohol retail sales, and demographic indicators at the zip code level in New Hampshire (NH), which has been at the forefront of the opioid epidemic in the United States. The data described below are from 2016, when NH had the third highest rate of opioid-related deaths in the country [24]. The census tract is normally used to proxy communities in US ecological studies, zip codes were used here because they provided enough detail for the spatial modeling of opioid-related poisoning deaths while maintaining individual privacy. Additionally, in a rural state like NH, the difference between the number of census tracts and zip codes is small; NH has a total of 295 census tracts and 248 populated zip codes.

Measures

Opioid poisoning death rates

The NH Office of the Chief Medical Examiner (NH OCME) provided zip-code level counts of poisoning deaths from 2016 that were linked to at least one opioid. Rates were then calculated as the number of deaths per 100,000 residents. The number of residents in each zip code area were retrieved from the 2016 US Census Bureau’s American Community Survey 5-Year Estimates (ACS) zip code tabulation area data. Counts of opioid poisoning deaths were provided by the NH OCME at the zip code level to protect individual privacy.

Alcohol sales

The National Alcohol Beverage Control Association (NABCA) provided 2016 sales data for distilled spirit and wine retail locations in NH, one of 17 US states or localities that adhere to an alcohol beverage control model where states control wholesale distribution and operate retail stores for the sale of spirits and wine [25]. In NH, the state operated a network of 79 retail stores in 2016, overseeing $678 million in gross sales that provided $156 million for general state revenue and $3 million for the prevention and treatment of alcohol abuse disorder [26].

Data were provided by NABCA at the outlet level reporting the total volume of alcohol sold in 2016. The volume data are standardized by NABCA as the number of 9 L units of alcohol sold in an outlet, this corresponds to the amount of liquid in one case of alcohol (12 bottles x 750 ml = 9 L). The total number of 9 L units of alcohol sold by each outlet in 2016 were aggregated at the zip code level for on-premise (i.e., liquor stores) and off-premise (i.e., bars, clubs, restaurants) alcohol outlets using the address of each retail location. This provided a measure of the total volume of alcohol sold in each NH zip code area in 2016.

Sociodemographic controls

The 2016 ACS zip code tabulation area data were used to construct sociodemographic controls for the analysis. Single-item measures included median age and population density measured as the number of residents per square mile of a zip code area. The Area Deprivation Index (ADI), a composite index of 17 ACS variables measuring housing, poverty, and employment domains, was used to control for socioeconomic factors associated with increased community rates of overdoses [8, 27]. Areas that rank high on the ADI indicate greater socioeconomic stress along the domains of income, education, employment, and housing [28]. The ADI was constructed using the approach of Singh [27] and Kind et al. [28] and a standardized index score was used in analyses.

Analysis plan

Data were combined and analyzed using GeoDa, an open-source software package for geospatial analyses [29] Because the data for this study are organized by postal code, we used a spatial regression approach to clarify the nature and significance of any spatially-dependent patterns in the data and their impact on model fit. Understanding the nature of any spatial dependence includes whether the outcome variable is correlated across space, merely because of proximity, and/or whether the residual error from a regression is correlated across space, indicating the presence of an unobserved variable(s) that may improve model fit. This analysis compared three models to determine which model best fit the data from the following:, (1) Ordinary Least Squares (OLS) regression; (2) spatial lag regression model; (3) spatial error regression model [30] First, an OLS regression was used to determine whether significant spatial auto-correlation existed in the outcome variable. A significant Moran’s I value here indicates the model could be improved by considering spatial structure in subsequent analyses. Then, two models were compared to determine whether controlling for specific types of spatial dependence improved model fit—a spatial lag model controlling for spatial autocorrelation in the outcome variable and a spatial error model controlling for the correlation of the residual error across space. The spatial error model includes lambda, a measure of spatially correlated errors to indicate whether there may be unobserved variables that would improve model fit.

Using the Akaike Information Criteria (AIC), model fit was compared for OLS regression, spatial lag, and spatial error models. The spatial error model was chosen based on its AIC value being the lowest and the results of the Moran’s I and Lagrange Multiplier diagnostics for spatial autocorrelation. Lastly, the spatial error model indicated non-stationarity via the Koenker-Bassett test (41.78, p < .001), meaning the relationship between the predictor and dependent variables was not constant across space. Because of this, we applied Geographically Weighted Regression (GWR) using ArcMap 10.8.1 [31] to visualize this non-stationarity and identify any patterns in the distribution. GWR provides R² values for each postal code to indicate how the proportion of variance explained changes across space.

Results

Table 1 presents descriptive statistics for the predictor and criterion variables used in this analysis. The spatial error model (Table 2) indicated population density and on-premise alcohol sales were significantly positively related to rates of opioid-related overdoses. Each 10% increase in population density is associated with a 0.3% increase in the rate of overdose fatalities and a 10% increase in the volume of alcohol sold associated with an 0.05% increase in the rate of overdose fatalities. Both the ADI and off-premise sales were significantly negatively associated with rates of opioid poisoning deaths. Lastly, lambda (spatially correlated error) was not significant in the spatial error model, indicating there is not an additional unobserved variable that would improve model fit.

Table 1 Descriptive statistics of study variables (n = 248 zip codes)

Table 2 Spatial error model of alcohol sales and community sociodemographics predicting opioid-related death rates (n = 248 zip codes)

The spatial error model produced a significant Koenker-Bassett test value of 41.78, indicating non-stationarity in the relationship between variables in the model across space. Figure 1 is used to illustrate the non-stationarity of the findings and presents the spatial distribution of local R² values across NH postal codes where higher values indicate better model fit and a greater proportion of explained variance. The GWR reveals a strong spatial pattern where the central and southern areas of NH have the best fit to the spatial error model. This pattern is indicative of the population distribution of NH. Figure 1 displays the point locations of the twenty most populous municipalities in NH, all are located in zip code areas above the R² value in the global spatial error regression model.

Fig. 1

Spatial distribution of local R² values in NH postal code areas

Conclusions

In this study, we found that population density and on-premise alcohol sales were positively related to opioid-related overdose deaths at a community level. Our findings are consistent with other work indicating population density as a predictor of opioid-related overdose deaths [7]. Additionally, findings here are the first to demonstrate a greater volume of on-premise alcohol sold was positively related to opioid overdose deaths. These findings complement individual-level findings that show the co-use of alcohol and opioids as reliably linked to overdose death and other poor health outcomes [32].

Contrary to our hypotheses, this study found a significant negative a relationship between the volume of off-premise sales (i.e., liquor stores) and rates of opioid poisoning deaths. The distribution of NH off-premise alcohol retail is important to consider in relation to this finding. NH’s liquor stores are state-owned retail outlets renowned for their large, tax free selection of spirits and easy-in, easy-in, easy-out locations to encourage out of state business (e.g., densely located on southeast border with Maine and Massachusetts and directly accessed via the state’s major north-south highway). Greater than half of NH wine and liquor off-premise retail sales come from out of state customers [33, 34]. Because of this unique policy environment, single state ecological designs like this may be limited in their ability to capture any regional connections between NH off-premise alcohol sales and non-NH opioid poisoning deaths.

The ADI was negatively associated with opioid poisoning deaths, postal codes that were relatively less disadvantaged had higher rates of deaths. While this finding runs counter to other state and national spatial analyses of structural disadvantage and overdose fatalities [5,6,7,8], contextually, it may be related to the spatial distribution of local postal code R² values seen in Fig. 1. A greater proportion of variance was explained in NH’s more populated areas, reflecting the choice of the alcohol-serving space as the focus of this study. The inclusion of features of the built environment more likely to be located in population centers with lower structural disadvantage compared to rural communities may have impacted our results.

The community-level findings here showing rates of opioid poisoning deaths higher in zip codes with a greater volume of on-premise alcohol sales have implications for harm reduction interventions, particularly policy, environmental, and information dissemination strategies. The alcohol retail environment provides opportunities for policy interventions related to advertising, price, location, and operating hours that have the potential to impact rates of opioid poisoning deaths. In terms of environmental interventions, approaches may include merging responsible alcoholic beverage server training with overdose education and naloxone distribution (OEND) approaches to preventing opioid poisoning deaths. This would include training on-premise alcohol retail staff on the identification of acute opioid intoxication and freely providing naloxone in alcohol-serving spaces, either among staff champions, in lavatories, or included alongside other life-saving kits like the Automated External Defibrillator [35]. With the rise of fentanyl as an adulterant in most substances commonly used in bars and nightclubs, the availability of naloxone in on-premise alcohol outlets is a feasible approach that may prevent some overdose deaths [36, 37].

From an information dissemination perspective, guerilla marketing campaigns (i.e., placing prevention messaging on alcohol packaging/labeling) and other messaging campaigns have the opportunity to communicate the dangers of mixing alcohol and opioids in a targeted fashion in and around alcohol-serving establishments. Targeting opioid prevention to alcohol retail spaces could also serve to support OEND models through broadening discussions of the dangers of mixing alcohol and opioids and broadening availability of naloxone by using alcohol-serving spaces as another partner in efforts to achieve community naloxone saturation [38, 39].

This study has limitations. First, it is unknown whether the individual opioid poisoning deaths involved in this analysis were concurrent users of alcohol and opioids, the ecological findings here are associational only. The results and implications from this study are applicable to communities, not individuals. Second, NABCA sales data only include liquor and wine; beer sales are not included in the totals. Future investigations could utilize grocery scanner data to include beer sales in investigations of alcohol sales and opioid-related deaths in NH. While there were significant relationships found between the volume of alcohol sold and opioid poisoning deaths, the effect was very small. Lastly, the results here are most useful for the more populous areas in NH and additional explanatory models would be needed to better understand patterns of overdose deaths within rural areas of the state [40].

The ability to intervene in substance use spaces is key to reducing substance-related harms. Because of the clear connection between alcohol and opioid co-use and fatal overdoses, prevention interventions and messaging have the ability to reduce the most severe harms associated with opioid misuse. At an individual level, avoiding the concurrent use of opioids and alcohol and carrying naloxone are important steps to take to prevent serious harms. At an environmental level, widening access to naloxone and increasing knowledge of its application are key to ensuring individuals across different spaces are equipped and ready to respond if an overdose occurs. Alcohol retail spaces may serve as prime locations for this targeted messaging on the dangers of the concurrent use of opioids and alcohol and for increasing the availability of naloxone in communities.