Alcohol Policy Scores and Alcohol-Attributable Homicide Rates in 150 Countries

Abstract

Introduction:

More comprehensive state-level alcohol policy environments are associated with lower alcohol-attributable homicide rates in the U.S., but few studies have explored this internationally. This study tests whether 3 national-level alcohol policy scores are associated with alcohol-attributable homicide rates.

Methods:

Data were from the 2016 WHO Global Survey on Alcohol and Health and the 2017 Global Burden of Disease Study (N=150 countries). In 2020, the authors calculated domain-specific alcohol policy scores for physical availability, marketing, and pricing policies. Higher scores represented more comprehensive/restrictive alcohol policy environments. Negative binomial regressions with Benjamini–Simes–Hochberg multiple testing correction measured the associations between policies and alcohol-attributable homicide rates. Authors stratified countries by World Bank income group to determine whether the associations differed among low- and middle-income countries.

Results:

A 10% increase in the alcohol policy score for pricing was associated with an 18% lower alcohol-attributable homicide rate among all the countries (incidence rate ratio=0.82, adjusted p-value or q<0.001) and with a 14% (incidence rate ratio=0.86, q=0.01) decrease among 107 low- and middle-income countries. More controls on days and times of retail sales (incidence rate ratio=0.96, q=0.01) and affordability of alcohol (incidence rate ratio=0.95, q=0.04) as well as adjusting excise taxes for inflation (incidence rate ratio=0.96, q<0.01) were associated with a 4%–5% lower alcohol-attributable homicide rate in the full sample.

Conclusions:

Countries with policies that reduce alcohol’s affordability or days/hours of sales tend to have fewer alcohol-attributable homicides, regardless of their income level. Alcohol-attributable homicide rates are highest in low- and middle-income countries; policies that raise alcohol-relative prices may hold promise for curbing these harms.

INTRODUCTION

Globally, less than half of adults drink alcohol¹; yet, alcohol is responsible for roughly 206 homicides per day.² Studies have consistently linked alcohol use with homicide perpetration and victimization³ and have demonstrated a dose–response relationship between alcohol consumption and aggressive behavior.⁴ In 2017, worldwide, 92.6% of alcohol-attributable homicide (AAH) victims were male, and 30.0% were in their 20s even though this age group only comprised 16.0% of the population; 90.7% of AAHs occurred in low- and middle-income countries (LMICs).²

Given the burden and demographic concentration of AAHs, alcohol prevention strategies that reduce AAH could enhance health equity. Still, action to adopt evidence-based policies has been slow,⁵ particularly in LMICs, where alcohol’s toll is high, resources to combat harms are minimal, and research about policies’ effectiveness is limited.^1,6,7

Alcohol policy scores (APSs) measure the combinations of alcohol policies with statistical efficiency. Brand et al.⁷ pioneered an APS combining 5 policy domains—availability, drinking context, price, advertising, and drink driving. They found that more comprehensive policy environments were associated with reduced per capita alcohol consumption. Others have calculated country-level APSs to test the association of stronger alcohol policies with disease burden¹⁰ and youth consumption.⁸ However, the APS literature remains geographically limited to either high-income countries^7,10 or some LMICs within specific geographic regions.^8,9

In 2017, the WHO European Regional Office (EURO) constructed 10 domain-specific APSs corresponding to the action areas in the WHO Global strategy to reduce the harmful use of alcohol.¹¹ Researchers have not tested the associations between these APSs and harms among geographically and economically diverse countries. To address this gap, this study explores whether the 3 EURO domain-specific APSs corresponding to the best buys—the most effective interventions reducing alcohol-related harm across populations costing <$100 U.S. per disability-adjusted life year averted¹²—are associated with AAH rates in a diverse sample of countries. This study also examines (1) whether the association between APSs differs between high-income countries and LMICs and (2) which APS subindicators (i.e., components of domain-specific APSs) are associated with AAH rates.

METHODS

Study Sample

This study used alcohol policy data from the 2016 Global Survey on Alcohol and Health (GSAH). The survey’s response rate was 89.2% (n=173 of 194 WHO Member States). The 173 participating countries comprised 98.3% of residents within the 194 Member States. Ministries of health nominated appropriate survey respondents in each participating country. Expert opinion can introduce inaccuracies, but WHO validates GSAH data at the country level to identify and correct these.¹

Measures

The calculated dependent variable was the number of homicides attributable to alcohol use per 100,000 people in 2017. Homicide data came from the Institute for Health Metrics and Evaluation Global Burden of Disease (GBD) tool (i.e., GBD Compare Tool).² The GBD applies a Levin-based population attributable fraction method to their meta-analyses.¹³ GBD collaborators calculated population attributable fractions for AAH using relative risks of alcohol consumption and interpersonal violence on the basis of a systematic review that identified 2 studies relevant to fatal violence.^14,15 Shield and colleagues (K Shield, University of Toronto, personal communication, 2020) is the only other economically diverse source for AAHs. Prevalence estimates from these 2 sources differed by <10% in 2016.¹⁶ The authors constructed models using Shield et al.’s estimates of AAHs and obtained nearly identical results.

The independent variables were 3 domain-specific APSs corresponding to the 3 best buys: physical availability, alcohol marketing, and pricing policies. Authors divided each APS by the theoretical maximum (i.e., total points in the APS), so these variables ranged from 0 to 100.¹¹ Higher values indicated more comprehensive or restrictive policies. The APS variables were scaled to range from 0 to 10; a 1-unit increase corresponded to a 10% increase in the APSs.

The domain-specific EURO APSs contained 2–6 subindicators, where each subindicator represented 1 theme within the alcohol control policy(ies). These subindicators were identified by a panel of experts convened by EURO. Briefly, the panel’s methods involved (1) identifying questions from WHO surveys that cover the most effective alcohol policies, (2) grouping the selected questions thematically into summary indicators, (3) assigning policies weights that indicate the level of underlying scientific evidence, (4) reformulating the variables as summary indicators, and (5) calculating scores as a weighted sum of the subindicators.¹¹

The 5 subindicators in the physical availability APS were legal purchasing age, government control of retail sales, restrictions on retail sales by outlet density and location, restrictions on retail sales by days and hours, and alcohol-free public environments. The marketing APS had 4 subindicators: restrictions on alcohol advertising, product placement, industry sponsorship for sporting and youth events, and producer promotions. A total of 3 subindicators comprised the pricing APS: adjusting excise taxes for inflation; affordability of beer (i.e., price divided by gross national income); and other pricing policies, including minimum unit pricing, bans on low-cost sales and volume discounts, and additional duties. Potential scores for all these subindicator variables were standardized using the same methods as the domain-specific APSs.

Owing to missing GSAH data, the authors only included the affordability of beer in the pricing APS, using Euromonitor to impute missing price data for beer. Authors were unable to impute missing data for spirits and wine because Euromonitor reported these prices by individual brand rather than by average unit prices, with high levels of missing data.

A previous analysis tested the criterion and construct validity of EURO’s overall APS (the sum of the 10 domain-specific APSs) by comparing it with other APSs. In support of the criterion validity, EURO’s APS correlated highly with Brand and colleagues’ gold standard.^7,17, In addition, more restrictive APSs were associated with lower per capita consumption after adjusting for total population, providing evidence of construct validity.¹⁷

Authors explored income group as a confounder and potential modifier because LMICs tend to have less comprehensive/restrictive policies and greater harm per liter than high-income countries.^1,18 World Bank income group was operationalized as a 3-category variable: low- and lower middle–income (combined), upper middle–income, and high-income countries (reference group) when used as a confounder. Owing to the small numbers of countries in these 3 strata, the authors used a binary variable of LMIC versus high-income countries to stratify the regressions. High-income countries were defined as those with a 2016 gross national income per capita ≥$12,476, and upper middle–income countries were defined as those with a gross national income of $4,036–$12,475.¹⁹

Other confounders included world region, percentage of the population that were male and aged 21–25 years, percentage of the population identifying as Muslim, policy enforcement, and percentage of homicides attributable to illicit drugs.

The rates of alcohol-attributable deaths and alcohol policy implementation vary regionally.¹ On the basis of the UN’s Sustainable Development Goals geographic region definitions and the geographic distribution of the sample, the 8 regions included Western Europe (reference group), Eastern Europe, Latin America, Caribbean, North America, Sub-Saharan Africa, Western Asia, and South-Eastern Asia.

Male individuals perpetrate most homicides, comprise majority of the homicide victims, and are more likely to be intoxicated at the time of homicide than female individuals.^20,21 Homicide perpetration also peaks in the early 20s.²² Data on the percentage of the population that were male and aged 21–25 years in 2016 came from the World Bank Databank.²³

Boston University’s World Religion Database provided data on the percentage of the population identifying as Muslim in 2010²⁴; its data come from government censuses and surveys. Islamic practices prohibit followers from consuming alcohol; therefore, percentage Muslim is associated with both abstinence and more restrictive alcohol policies.²⁵

The authors included 2 proxies for policy enforcement: a government effectiveness index and organized crime’s effect on business. The government effectiveness index captured “...the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies” and came from the World Bank World Governance Indicators Databank.²⁶ It ranged from −2.5 to 2.5, with higher values denoting more effective governance. Estimates of organized crime’s effect on business came from the World Economic Forum’s Global Competitiveness Index.²⁷ Countries with stronger policy enforcement tend to have a lower impact of organized crime on business.²⁸ The estimates are based on answers from 12,775 business executives in 133 economies to the question: In your country, to what extent does organized crime (mafia-oriented racketeering, extortion) impose costs on businesses? Responses were measured on a 7-point Likert-type scale and ranged from a huge cost (1) to none (7).²⁷

Drug trafficking may be associated with steadily high levels of violence or may spark violence when there are disputes over territorial control or other matters of power.²⁰ In addition, some drugs may amplify alcohol’s disinhibiting effects, leading to violence.²⁹ The percentage of homicides attributable to illicit drugs came from the Institute for Health Metrics and Evaluation’s GBD Compare Tool.² A similar variable for the percentage of homicides attributable to guns as a proxy for gun availability was tested but not retained in the final models because it explained the variability in a minority of high-income countries and was therefore not significant.

Before imputation, 59 countries (39.3%) lacked at least 1 variable used to calculate an APS; missingness was highest for pricing policies (30%) and lower for marketing (15.3%) and availability (6.7%). The primary imputation method was dubbed the best data approach because it used the most complete and similar data available at the country level. This method replaced the values missing in the 2016 GSAH with data from the 2012 GSAH and Euromonitor. All but 4 marketing variables could be imputed using this method.

Two sensitivity analyses explored the robustness of this imputation method. The first imputed the missing data using median values for each variable within each World Bank income group; the second used truncated multiple chained regressions. Results are in the Appendix (available online).

Statistical Analysis

A total of 3 sets of negative binomial regression models accounting for overdispersion in the rates of AAH, with the natural log of the 2017 population as the offset, estimated the association between APSs and the AAH rate. APSs (2016) and AAHs (2017) were lagged by 1 year to allow time for policy implementation. The first model contained all the 3 domain-specific APSs (i.e., the best buys) and tested the overall association between alcohol policy and AAHs while adjusting for country-level income. The second was stratified by country-level income to test whether this modified the association between alcohol policies and the AAH rate. The final set comprised 3 models exploring which component(s) of each APS (if any) were associated with the AAH rate; each was specific to 1 best buy and included the respective policy subindicators.

Countries within the regions may be more similar than those from different regions, so Moran’s Index (Moran’s I) determined the independence of observations.³⁰ Finally, a Benjamini–Simes –Hochberg multiple testing correction minimized Type I error rate³¹ and provided an adjusted p-value (called a q-value), where all the q-values equal to or less than the critical value (0.05) are statistically significant.³²

RESULTS

After excluding countries with a total ban on alcohol (n=9) or with missing covariates (n=14), there were 150 countries in the final sample (Table 1). Included countries had an average population of 44.5 million in 2015, whereas excluded countries were smaller (mean=14.2 million, p<0.01). Included and excluded countries did not differ on World Bank income group (p=0.59), WHO Region (p=0.60), or whether they had a national written alcohol policy (p=0.58).

Table1.

Description of Study Sample, APSs, and Policy Subindicators, Overall and by Income Level

	Full sample (N=150)			LMIC (n=107) Median (SD)a,b	High income (n=43) Median (SD)a,b
Variables	Median (SD)a	Min	Max
APS and subindicators
Physical availability APS	5.4 (2.5)	0	9.4	5.2 (2.4)	5.7 (2.5)
Minimum legal purchase age	7.5 (3.4)	0	10	7.5 (3.4)	7.5 (3.4)
Licensing and monopolies	10.0 (4.3)	0	10	10.0 (4.0)	10.0 (4.8)
Days and times of retail sales	2.5 (4.1)	0	10	2.5 (4.0)	2.5 (4.1)
Density and location of retail sales	7.5 (3.8)	0	10	7.5 (3.8)	2.5 (3.8)
Alcohol sales at special events	6.7 (4.7)	0	10	0.0 (4.6)	10.0 (4.6)
MarketingAPS	4.4 (3.5)	0	10	4.4 (3.7)	4.4 (2.8)
Advertising restrictions	5.0 (3.5)	0	10	5.0 (3.7)	7.5(2.7)
Product placement	5.0 (4.0)	0	10	5.0 (4.1)	5.0 (3.8)
Sponsorship	5.0 (4.2)	0	10	5.0 (4.3)	5.0 (3.9)
Producer promotions	2.5 (4.0)	0	10	0.0 (4.2)	2.5 (3.3)
Pricing policies APS	2.0 (1.5)	0	9.2	2.7 (1.6)	2.0 (1.4)
Tax adjustment	0.0 (4.2)	0	10	0.0 (4.4)	0.0 (4.0)
Affordability of beer	5.0 (3.2)	0	10	7.5 (3.4)	5.0 (2.2)
Other pricing policies	0.0 (1.6)	0	10	0.0 (1.2)	0.0 (2.3)
Covariates
Region, n (%)
Western Europe	28 (18.7)	—	—	5 (4.7)	23 (53.5)
Eastern Europe	9 (6.0)	—	—	5 (4.7)	4 (9.3)
Latin America	17 (11.3)	—	—	15 (14.0)	2 (4.7)
Caribbean	10 (6.7)	—	—	7 (6.5)	3 (7.0)
North America	3 (2.0)	—	—	1 (0.9)	2 (4.7)
Sub-Saharan Africa	51 (34.0)	—	—	48 (44.9)	3 (7.0)
Western Asia	15 (10.0)	—	—	12 (11.2)	3 (7.0)
Southeast Asia	17 (11.3)	—	—	14 (13.1)	3 (7.0)
Organized crime	4.55 (0.9)	1.5	6.8	4.4 (0.8)	5.5 (0.7)
% Male aged 21–25 years	8.0 (1.4)	5.3	17.9	8.2 (1.0)	6.9 (1.9)
Drug availability	0.7 (0.5)	0.1	3.7	0.7 (0.5)	0.8 (0.5)
Government effectiveness	−0.1 (1.0)	−1.8	2.2	−0.5 (0.6)	1.1 (0.5)
% Muslim	3.1 (32.4)	0	99.0	9.6 (35.5)	1.5 (14.0)

^aAppropriate for APSs, subindicator scores, and all covariates except for region.

^bAppropriate for region.

APS, alcohol policy score; LMIC, low- and middle-income country; max, maximum; min, minimum.

The median rate of AAHs was 40.8 homicides per 100,000 people. LMICs had approximately 1.75 times higher median rate of AAHs (59.7 per 100,000) than high-income countries (31.7 per 100,000). The median APS was highest for physical availability (5.4 of a possible 10.0, SD=2.5), followed by marketing (4.4, SD=3.5) and pricing policies (2.0, SD=1.5) (Figure 1). Countries with higher marketing APSs tended also to have higher physical availability APSs (r=0.50), but there was no association between pricing policy APSs and APSs for availability (r= −0.07) or marketing (r= −0.01).

Figure 1.

Maps showing 2016 APSs and 2017 AAH rate per 100,000 persons by country. APSs are scaled from 0 to 100. Higher scores indicate more comprehensive or restrictive alcohol policies. AAH rate data were obtained from the Global Burden of Disease Study.

Note: Figure 1A–C depict the distribution of APSs for the 3 best buys—physical availability, marketing, and pricing policies—across 150 countries. The darker the color, the higher the APS and the more comprehensive and restrictive the policy domain. Sub-Saharan Africa, Eastern Europe, and South-Eastern Asia have some of the highest physical availability and marketing APSs, whereas North America, Western Europe, and Sub-Saharan Africa have the highest pricing APSs. Figure 1D shows the distribution of the rates of AAH. The darker the color, the higher the AAH rate. Brazil, Mexico, and South Africa have the highest rates of AAH. Circle size is proportional to the number of countries with that value; colors were determined by natural Jenks.

AAH, alcohol-attributable homicide; APS, alcohol policy score.

A 10% increase in the pricing policy APS was associated with an 18% lower rate of AAHs (incidence rate ratio [IRR]=0.82, 95% CI=0.75, 0.90, q<0.001) (Table 2). In this model, upper middle–income countries’ rate of AAHs was 135% higher (IRR=2.35, 95% CI=1.35, 4.10, q<0.01) than that of their high-income counterparts, and low- and lower middle–income countries had a rate >200% higher (IRR=3.16, 95% CI=1.56, 6.39, q<0.01). AAHs had slight positive spatial dependence (Moran’s I=0.02, p=0.02) before the regressions, but the regression covariates explained all of it. Neither the physical availability nor the marketing APSs were associated with AAH rates.

Table 2.

Negative Binomial Regression for the Associations Between AAH and Alcohol Policy Scores, Overall and Stratified by Income Level

Variables	All countries (N=150), IRR (95% CI)	Low- and middle-income countries (n=107), IRR (95% CI)	High-income countries (n=43), IRR (95% CI)
Alcohol policy score
Physical availability	0.97 (0.92,1.03)	0.92 (0.85, 1.01)	1.03 (0.97, 1.09)
Marketing	0.98 (0.93,1.04)	0.99 (0.93,1.05)	1.02 (0.95, 1.09)
Pricing	0.82** (0.75, 0.90)	0.86* (0.77, 0.97)	0.95 (0.87, 1.04)
Region
Western Europe	ref	ref	ref
Eastern Europe	1.16 (0.63, 2.16)	1.89 (0.62, 5.75)	0.56** (0.38, 0.82)
Latin America	1.65 (0.89, 3.08)	2.09 (0.83, 5.28)	1.04 (0.55, 1.96)
Caribbean	2.80** (1.47, 5.33)	1.94 (0.73, 5.15)	4.64** (2.63, 8.17)
North America	0.88 (0.32, 2.43)	1.19 (0.20, 7.09)	1.30 (0.53, 3.18)
Sub-Saharan Africa	0.90 (0.52, 1.58)	1.26 (0.55, 2.90)	0.65 (0.38, 1.14)
Western Asia	1.32 (0.65, 2.67)	1.84 (0.64, 5.24)	3.78** (1.60, 8.91)
Southeast Asia	0.31** (0.17, 0.56)	0.50 (0.21,1.20)	0.56* (0.34, 0.93)
Income level
High income	ref
Upper middle income	2.35** (1.35, 4.10)	—	—
Low or lower middle income	3.16** (1.56, 6.39)	—	—
Organized crime	0.79* (0.65, 0.96)	0.75 (0.59, 0.96)	1.11 (0.88, 1.41)
% Male aged 21–25 years	1.21* (1.06,1.39)	1.28* (1.07,1.52)	1.55** (1.34, 1.80)
Drug availability	1.94** (1.36, 2.77)	1.72* (1.11, 2.67)	1.37 (0.99, 1.91)
Government effectiveness	1.07 (0.80, 1.44)	1.22 (0.87,1.72)	0.61* (0.43, 0.89)
% Muslim	0.11** (0.06, 0.21)	0.13** (0.06, 0.26)	<0.01** (<0.01, <0.01)
Moran’s index	−0.01	<0.001	<0.01

Note:

Boldface indicates statistical significance

^*(q<0.05

^**q<0.01).

q-Value is also referred to as adjusted p-value.

AAH, alcohol-attributable homicide; IRR, incidence rate ratio.

In adjusted models stratified by the World Bank income group, a 10% increase in the pricing APS was associated with a 14% lower AAH rate among LMICs (IRR=0.86, 95% CI=0.77, 0.97, q=0.04).

In the physical availability model, the days and hours of retail sales indicator were associated with a 5% lower rate of AAHs (IRR=0.95, 95% CI=0.91, 0.98, q=0.01) (Table 3). Minimum legal purchasing age; government control of retail sales; and restrictions on alcohol outlet density, location, and sales at special events were not associated with AAH rates.

Table 3.

Negative Binomial Regression for the Association Between AAH and Alcohol Policy Scores by Policy Type and Subindicator (N=150)

Variables	Physical availability, IRR (95% CI)	Marketing, IRR (95% CI)	Pricing policies, IRR (95% CI)
Physical availability
Minimum legal purchase age	1.02 (0.97,1.06)	—	—
Licensing and monopolies	1.00 (0.97,1.04)	—	—
Days and times of retail sales	0.95* (0.91, 0.98)	—	—
Density and location of retail sales	1.02 (0.97,1.06)	—	—
Alcohol sales at special events	1.01 (0.98, 1.05)	—	—
Marketing score
Advertising restrictions	—	0.97 (0.90, 1.04)	—
Product placement	—	1.03 (0.97, 1.09)	—
Sponsorship	—	1.00 (0.95, 1.05)	—
Producer promotions	—	0.96 (0.90, 1.02)	—
Pricingscore
Tax adjustment	—	—	0.95** (0.93, 0.98)
Affordability of beer	—	—	0.94* (0.89, 0.99)
Other pricing policies	—	—	0.92 (0.85, 1.01)
Region
Western Europe	ref	ref	ref
Eastern Europe	0.78 (0.41, 1.50)	1.03 (0.55, 1.94)	1.07 (0.57, 1.99)
Latin America	2.00 (1.10, 3.63)	1.56 (0.81, 3.03)	1.87 (1.05, 3.31)
Caribbean	3.89** (2.01, 7.52)	3.33** (1.67, 6.66)	3.14** (1.69, 5.83)
North America	0.89 (0.32, 2.50)	0.81 (0.27, 2.42)	0.90 (0.32, 2.50)
Sub-Saharan Africa	1.06 (0.61, 1.86)	0.87 (0.48, 1.56)	1.07 (0.63, 1.84)
Western Asia	1.07 (0.53, 2.14)	1.23 (0.58, 2.58)	1.36 (0.69, 2.71)
Southeast Asia	0.43* (0.24, 0.77)	0.40* (0.21, 0.73)	0.33** (0.19, 0.58)
Income level
High income	ref	ref	ref
Upper middle income	0.97 (0.63,1.49)	1.03 (0.66, 1.61)	0.76 (0.50, 1.15)
Low or lower middle income	0.49 (0.23, 1.03)	0.45 (0.21, 0.97)	0.33** (0.16, 0.66)
Organized crime	0.80 (0.65, 0.98)	0.81 (0.66, 0.99)	0.78* (0.64, 0.94)
% Male aged 21–25 years	1.25** (1.09, 1.45)	1.24* (1.06, 1.44)	1.22* (1.06, 1.40)
Drug availability	1.81** (1.27, 2.60)	1.82** (1.26, 2.62)	1.83** (1.29, 2.59)
Government effectiveness	0.94 (0.69, 1.28)	0.96 (0.69, 1.32)	1.07 (0.79, 1.45)
% Muslim	0.09*** (0.05, 0.18)	0.11** (0.05, 0.21)	0.10*** (0.05, 0.18)
Moran’s index	−0.01	—a	−0.01

Note: Boldface indicates statistical significance

^*(q<0.05,

^**q<0.01).

q-Value is also referred to as adjusted p-value.

^aMoran’s index could not be calculated owing to missing values. AAH, alcohol-attributable homicide; IRR, incidence rate ratio.

In the pricing policy model, adjustment for inflation was also associated with a 5% lower AAH rate (IRR=0.95, 95% CI=0.93, 0.98, q<0.01), whereas affordability of beer was associated with a 6% lower rate (IRR=0.94, 95% CI=0.89, 0.99, q=0.02). Other pricing policies (e.g., minimum unit pricing) were not significantly associated with the rates of AAH.

Comparing the best data imputation with the alternatives yielded similar distributions and IRRs for the adjusted regressions (Appendix, available online).

DISCUSSION

This study builds on a previous analysis¹⁷ exploring the association between APSs and country-level alcohol consumption by showing that countries with higher pricing APSs have lower AAH rates and that this association was statistically significant among LMICs. The authors also extend findings from Naimi et al.²¹ and Lira and colleagues,³³ who examined this association subnationally within the U.S. and concluded that more comprehensive alcohol control policy environments were associated with fewer alcohol-involved homicides and reduced odds of alcohol involvement in homicides. Using the APS subindicators, this study further showed that countries with lower beer affordability, inflation-adjusted alcohol excise taxes, and fewer days/hours of alcohol retail sales also had lower AAH rates. To the best of the authors’ knowledge, this is the first study to link APSs to country-level outcomes in a diverse sample of countries and to associate APSs with AAH rates at the country level.

Nonsignificant findings for the marketing APS likely resulted from the measurement design. Both the availability (70) and the pricing (94) APSs had more available points than the marketing APS (48).¹¹ After rescaling the APSs by dividing them with the maximum, there was still notably less variability in the marketing scores. The authors are unaware of any longitudinal analyses of alcohol marketing exposure and homicide; however, 2 cross-sectional analyses suggest that alcohol advertisements can affect violent crime.^34,35 One compared 4 types of violence and found the strongest association for homicide.³⁵ Further, preferably longitudinal, research is needed in this area.

Few studies explore the association of alcohol policies with the rate of AAH over time; however, several time-series studies^36–39 show that homicides rise and fall in response to changes in alcohol consumption. Recent policy evaluations are consistent with the present findings: A seminal meta-analysis of longitudinal alcohol excise tax studies estimated that a 10% increase in the price of alcohol would result in a 2.2% decrease in the overall violence rate (p<0.001).⁴⁰ In addition, a policy that established an 11:00PM closing time for 24-hour bars in 1 Brazilian city was associated with a 44% decrease in homicide.⁴¹

Findings from the models stratified by income group suggest that pricing policies, the most effective tools to reduce consumption and related harms,^5,42,43 may hold promise for preventing AAHs in LMIC settings, despite a lack of research in these countries to date. If corroborated by future longitudinal studies, this line of research could offer urgently needed evidence to fill gaps in policy coverage in LMICs, especially in the Americas where the burden of AAHs is extremely high. Pricing policies would be a win–win–win for LMICs with high levels of AAH, raising revenue, reducing the level of harms, and advancing equity.⁴⁴

Limitations

Limitations of this analysis include its cross-sectional and ecologic design. Results should not be interpreted as causal and are only generalizable at a country level and not to individuals. Associations between alcohol and homicide are complex and result from individual-, community-, and societal-level factors (along with interactions among them),^45–47 and this analysis focused only at the societal level. The variable of males aged 21‒25 years summarized the most important individual-level predictors (i.e., sex and age), and the authors considered others (e.g., income inequality, unemployment, poverty) during the model-building process. APS formulas did not include data on policy enforcement, which is essential to policy effectiveness.⁴² The measure of AAH was also estimated. However, the GBD methodology has been refined over ≥25 years,^13,48,49,50 and the use of these data allowed the study outcome to be restricted to AAHs, the subset of homicides that would be most responsive to alcohol policies, while using a geographically diverse sample. Finally, imputation requires assuming that data are missing at random; if this assumption is not met, findings may be biased. However, sensitivity analyses indicated that the primary imputation method was robust.

CONCLUSIONS

The findings around price suggest the importance of strategies such as alcohol taxes, which are widely underutilized globally, failing to keep up with inflation in most countries, and consequently losing their preventive value.^1,6,12 Stratification of countries by income group was a first step toward addressing the gap in evidence regarding what works to prevent alcohol’s harms in LMIC settings. Future research should evaluate whether recent policy changes are associated with changes in the rate of AAHs. Given that the per capita prevalence of AAH increased since 1990 by 19.5% and 39.6% in lower middle–and upper middle–income countries, respectively,² this line of research could identify an important tool to help address the inequitable distribution of AAH across income groups and countries.