Developing an alcohol policy assessment toolkit: application in the western Pacific

Abstract Objective To demonstrate the development and feasibility of a tool to assess the adequacy of national policies aimed at reducing alcohol consumption and related problems. Methods We developed a quantitative tool – the Toolkit for Evaluating Alcohol policy Stringency and Enforcement (TEASE-16) – to assess the level of stringency and enforcement of 16 alcohol control policies. TEASE-16 was applied to policy data from nine study areas in the western Pacific: Australia, China excluding Hong Kong Special Administrative Region (SAR), Hong Kong SAR, Japan, Malaysia, New Zealand, the Philippines, Singapore and Viet Nam. Correlation and regression analyses were then used to examine the relationship between alcohol policy scores and income-adjusted levels of alcohol consumption per capita. Findings Vast differences exist in how alcohol control policies are implemented in the western Pacific. Out of a possible 100 points, the nine study areas achieved TEASE-16 scores that ranged from 24.1 points for the Philippines to 67.5 points for Australia. Study areas with high policy scores – indicating relatively strong alcohol policy frameworks – had lower alcohol consumption per capita. Sensitivity analyses indicated scores and rankings for each study area remained relatively stable across different weighting schemes, indicating that TEASE-16 was robust. Conclusion TEASE-16 could be used by international and national regulatory bodies and policy-makers to guide the design, implementation, evaluation and refinement of effective policies to reduce alcohol consumption and related problems.


Introduction
Globally, alcohol consumption contributes to an estimated 9.6% of all disability-adjusted life-years. 1 Fortunately, effective evidence-based policies for alcohol control can protect population health and well-being, save lives, reduce health-care costs and increase productivity. [2][3][4] Indeed, every European country has some form of national alcohol control policy framework. 3 Policy development, however, is only one step. The World Health Organization (WHO) and other leading international agencies have repeatedly called for countries to assess, compare and refine their national alcohol control policy frameworks. 3,5,6 There is limited guidance, however, on how these activities should be carried out. A reliable and valid tool for quantifying alcohol policy stringency and enforcement would yield a standardized, succinct summary of a country's policy framework and facilitate investigation of the relationship between alcohol policies and consumption per capita. It would also enable meaningful comparisons across countries and jurisdictions and comparisons within countries over time. Further, it could highlight weak policies and provide estimates of the impact of policy improvements on consumption.
A small number of scales have been developed to assess national alcohol control policies. 7 Although informative, these scales have several limitations. These include the failure to: (i) comprehensively assess enforcement, [7][8][9][10][11][12][13][14] (even though enforcement varies considerably across policies and countries and is likely to impact effectiveness); 15 (ii) demonstrate the scale's feasibility through a practical application; 10 (iii) demonstrate the scale's robustness through sensitivity analyses; [7][8][9][10][11][12] (iv) rank countries according to their degree of alcohol control; 12 (v) relate alcohol policy scores to corresponding per capita consumption estimates [7][8][9][10][11][12] or adjust for income -which shows considerable between-country variation -in per capita consumption estimates; 13,14 and (vi) provide complete assessment of policies and use up-to-date literature. 13,15,16 We aimed to address these limitations and develop a comprehensive and practical tool to measure the stringency and enforcement of national alcohol control policies.
To demonstrate our tool's feasibility and practical value, we applied it to nine study areas in the western Pacific. Despite recent increases in alcohol consumption in the region, 17 previous alcohol policy scales have been largely applied to Europe. Only two previous studies have evaluated alcohol control policies in the western Pacific and both focused on high-income countries. 12,13 In this paper, we compared the relative strength of national alcohol policy frameworks across a range of developed and developing study areas in the region. We used up-to-date policy data, conducted comprehensive sensitivity analyses to demonstrate the tool's robustness and investigated the relationship between alcohol policy scores and income-adjusted levels of alcohol consumption per capita.

The assessment tool
We developed the Toolkit for Evaluating Alcohol policy Stringency and Enforcement-16 (TEASE-16), which builds on previous policy evaluation scales. 13 It is the first tool to assess levels of stringency and enforcement comprehensively. As summarized in Table 1, TEASE-16 has five main components: (i) five regulatory domains; (ii) 16 evidence-based alcohol control policies or policy topics; (iii) effectiveness star ratings Objective To demonstrate the development and feasibility of a tool to assess the adequacy of national policies aimed at reducing alcohol consumption and related problems. Methods We developed a quantitative tool -the Toolkit for Evaluating Alcohol policy Stringency and Enforcement (TEASE-16) -to assess the level of stringency and enforcement of 16 alcohol control policies. TEASE-16 was applied to policy data from nine study areas in the western Pacific: Australia, China excluding Hong Kong Special Administrative Region (SAR), Hong Kong SAR, Japan, Malaysia, New Zealand, the Philippines, Singapore and Viet Nam. Correlation and regression analyses were then used to examine the relationship between alcohol policy scores and income-adjusted levels of alcohol consumption per capita. Findings Vast differences exist in how alcohol control policies are implemented in the western Pacific. Out of a possible 100 points, the nine study areas achieved TEASE-16 scores that ranged from 24.1 points for the Philippines to 67.5 points for Australia. Study areas with high policy scores -indicating relatively strong alcohol policy frameworks -had lower alcohol consumption per capita. Sensitivity analyses indicated scores and rankings for each study area remained relatively stable across different weighting schemes, indicating that TEASE-16 was robust. Conclusion TEASE-16 could be used by international and national regulatory bodies and policy-makers to guide the design, implementation, evaluation and refinement of effective policies to reduce alcohol consumption and related problems.

Research
Alcohol control policies in the western Pacific Natacha Carragher et al. (i.e. ratings of the effectiveness of the policies in reducing the adverse effects of alcohol, which were based on expert reviews of the literature); 2 (iv) level of stringency; and (v) level of enforcement. Further details on the conceptual framework of TEASE-16 are provided in Appendix A (available at: https://ndarc. med.unsw.edu.au/resource/appendixtease-16-supplementary-details).
Following Brand et al., 13 we examined five broad regulatory domains that were identified in a WHO-sponsored comprehensive analysis of alcohol policies. 2 Within these domains, we focused on 16 policies that have been implemented around the world and evaluated by experts as being effective in reducing the adverse effects of alcohol. 2 Like Brand et al., 13 we excluded policies that: have limited effectiveness (e.g. warning labels on containers for alcoholic drinks and/or that relate to the treatment of problem drinkers) because we wished to focus on preventive public health strategies; and we were not implemented in any of the nine study areas under investigation (e.g. minimum pricing).
Each policy was rated according to level of stringency and enforcement (Appendix A). Briefly, stringency refers to the relative strictness of a given policy. For example, limiting the age of those who can purchase alcohol to 16,17,18,19 or ≥ 20 year-olds, reflects increasingly stringent policy positions on controlling the availability of alcohol. Enforcement refers to the strength at which a given policy is implemented in practice. We divided levels of enforcement into three categories: (i) poor -reflecting policies that were rarely or poorly enforced, or instances where no legislation or enforceable powers were in place; (ii) moderate -referring to policies that had limited or occasional enforcement or were enforced only when violations were reported or blatant; and (iii) strong -reflecting widely enforced policies.
Although TEASE-16 builds on Brand et al.'s Alcohol Policy Index, 13 the scales differ in terms of policy conceptualizations, effectiveness ratings, inclusion of enforcement ratings, the development of three alternative weighting schemes combining stringency and enforcement ratings, and the use of income-adjusted estimates of alcohol consumption per capita (Appendix A).

Scoring and sensitivity analyses
Each of the 16 policy topics was allocated a maximum potential number of points based on peer-reviewed assessments of effectiveness in reducing the adverse effects of alcohol (Table 1). Proportionate points were then allocated according to the particular level of stringency and enforcement. For each study area, scores across all 16 policy topics were collated to yield an overall score ranging between 0 and 100. Then study areas were rank ordered.
To examine the robustness of TEASE-16, we applied different weighting methods to each policy topic according to its effectiveness rating and subsequently calculated the corresponding proportionate point scores. We tested alternative weighting methods to avoid the risk of study areas rejecting TEASE-16 on the grounds that a particular weighting scheme was unfairly punitive.
In total, four different weighting methods were used to assign stringency and enforcement points: baseline, heavy, equal and area-specific. In baseline weighting, weights of 1, 2 and 3 were applied to policy topics with one-, two-and three-star effectiveness ratings, respectively. Heavy weighting used corresponding weights of 1, 3 and 5 whereas equal weighting assigned the same weight to all policies regardless of effectiveness ratings. Area-specific weights were also derived -using data envelopment analysis 18 and implemented with the Solver add-in for Excel 2010 (Microsoft, Redmond, United States of America) -in a manner that optimized a study area's relative performance. Area-specific weights were constrained to reflect effectiveness ratings (i.e. a three-star policy topic received a greater weight than a two-star policy and a two-star topic received a greater weight than a one-star policy). Additional constraints were specified to ensure that the area-specific weights were plausible and to avoid instances where a study area might be awarded a perfect rating because zero weights had been allocated to policy topics that had minimal stringency. For example, the area-specific weights were constrained so that the maximum weight was less than 12-fold higher than the minimum weight.
For each of the four weighting schemes, three methods were used to yield combined ratings for stringency and enforcement (see Appendix A). In one method -50:50 combinationequivalent points were allocated to stringency and enforcement. In another method -25:75 combination -a quarter of the points were assigned to stringency Domain, policy topic Effectiveness star rating a

Level of stringency Level of enforcement b
Mandatory penalties for exceeding the legal maximum blood alcohol concentration 2 No penalty Poor, moderate or strong Fine Penalty points Disqualification or licence suspension Imprisonment Other Graduated licensing for young drivers 2 No Poor, moderate, or strong Yes a Policies that were considered to be effective in reducing the adverse effects of alcohol were given one-, two-and three-star ratings, respectively. 2 b Poor reflects policies that were rarely or poorly enforced, or instances where no legislation or no enforceable powers were in place; moderate refers to policies that had limited or occasional enforcement, or were enforced only when violations were reported or blatant; and strong reflects widely enforced policies. c Price indexes were calculated using guidelines by Brand et al. 13 The price index refers to the retail price (including alcohol taxes) for a standard size beverage container (0.5 L beer, 0.75 L wine or 0.75 L spirits), taking into account a country's standard of living. This adjustment involves dividing the retail price by the per capita share of a country's gross domestic product and subsequently multiplying the result by 10 000 to yield a price index with an approximate range of 0-20.

Research
Alcohol control policies in the western Pacific Natacha Carragher et al. and the remainder to enforcement. In the third method -multiplicative combination -stringency ratings were multiplied by a third of the raw enforcement rating. All calculations were conducted in Excel 2010.

Data retrieval
Since low-income nations generally have a greater disease burden per unit of alcohol consumption than high-income nations, 19 we retrieved alcohol policy and consumption data from both developed areas of the western Pacific (Australia, China excluding Hong Kong Special Administrative Region (SAR), Hong Kong SAR, Japan, New Zealand and Singapore) and developing areas (Malaysia, the Philippines and Viet Nam). 20 These nine study areas are economically diverse and geographically widespread; they have different epidemiological profiles and reflect a range of cultural, religious and social practices relating to alcohol use.

Policies
We obtained data on stringency and enforcement from peer-reviewed papers and WHO reports published between 2008 and 2012, as well as government and related public health websites. A full list of the data sources is available from the corresponding author. Where information was unclear or outdated, we verified policies with the relevant public health and government officials in August-October 2012. This ensured that we included the most up-to-date legislation and that policy topics were correctly weighted. Extensive efforts were undertaken to cross-reference data to ensure accuracy. Complete information on stringency and enforcement was retrieved for all 16 policy topics in each of the nine study areas.

Alcohol consumption
For each study area, an estimate of the average percentage of alcohol by volume was used to convert total volume of alcoholic drinks sold in 2011 [21][22][23][24][25][26][27][28][29] into total volume of alcohol consumed in pure alcohol. The result was then multiplied by 1 000 000 and divided by the population estimate, in millions, for the study area-obtained via the websites of the relevant national statistics agency and verified by officials -to yield an estimate of the mean volume of alcohol consumed per capita in 2011. Since alcohol consumption is positively related to income, 2 we divided each estimate by the relevant gross domestic product per capita -reported in international dollars using purchasing power parity exchange rates 30 -to yield an incomeadjusted estimate of alcohol consumption per capita for the year 2011 in each study area.

Analysis
For each study area, we calculated alcohol policy scores using 12 sets of assumptions -reflecting the four weighting methods and three methods for combining ratings of stringency and enforcement. To facilitate comparisons, we calibrated the scores generated under each set of assumptions to yield equivalent ranges. Subsequently, we identified the median rank and overall score for all 12 assumptions for each study area. We then calculated Pearson's or Spearman's correlation coefficients -as appropriate -to compare these medians with the corresponding baseline ranks (i.e. those produced using baseline weighting). Additionally, we calculated correlations using the extreme values -rather than medians -to provide a measure of the robustness of TEASE-16. To evaluate the relationship between policy scores and income-adjusted alcohol consumption per capita, we performed a simple linear regression in SPSS (SPSS Inc., Chicago, USA).

Strength of policy frameworks
We compared the comprehensiveness of alcohol control policies in nine study areas in the western Pacific by calculating a rating for each regulatory domain (Table 2). Overall, the median rating was 56.4 points -out of a possible 100 points -with ratings ranging from 24.1/100 in the Philippines to 67.5/100 in Australia. The nine study areas received median domain ratings of 5.9/28.9 points for physical availability policies, 3.9/10.5 points for drinking context policies, 18.4/23.7 points for alcohol pricing policies, 0.4/2.6 points for alcohol advertising policies and 23/34.2 points for motor vehicle regulations.

Policy scores and alcohol consumption
A strong inverse relationship was observed between income-adjusted levels of alcohol consumption per capita and alcohol policy scores (r = −0.88; P = 0.001; Fig. 1). To exclude price demand influences, we recalculated alcohol policy scores after removing alcohol prices from the model. This resulted in a minor change to the observed relationship (r = −0.83; P = 0.003). Based on the slope of the regression line, a one-point increase in alcohol policy score equated to a 1.8% decrease in income-adjusted

Sensitivity analyses
Rankings remained relatively stable across the 12 assumptions tested, with those of only three study areas -Hong Kong SAR, Japan and Malaysia -differing by three positions (Table 3). Correlation analyses confirmed that median ranks (r = 0.98; P < 0.0001) and ratings (r = 1.0; P < 0.0001) differed minimally from the baseline ranks and ratings. Indeed, even when baseline ranks and ratings were compared with the extreme corresponding ranks (r = 0.97; P < 0.0001) and ratings (r = 0.97; P < 0.0001) from the 12 assumptions, correlation coefficients remained high. The negative relationship between median ratings and consumption remained robust (r = −0.86, P = 0.003).

Discussion
Studies comparing alcohol policy frameworks and consumption across countries are scarce. 15 To address this gap, we developed and applied a tool to evaluate policies across nine study areas in the western Pacific. We found striking variations in how policies were implemented. Among the nine study areas evaluated, the Philippines had the weakest alcohol policy framework whereas Australia had the strongest. Australia was particularly strong in relation to policies limiting driving while under the influence of alcohol and alcohol pricing policies. In Australia -as in most of the other study areas -alcohol advertising policies were relatively weak. In the Philippines, all regulatory domains were generally weak, particularly those relating to drinking context and driving policies. Although we used TEASE-16 to evaluate alcohol control policies and consumption at a particular point in time, the tool could be used to evaluate policy changes and consumption within a study area over time. Under a log regression function, an increase in alcohol policy score would have a greater impact on consumption for study areas with weak policy frameworks than for study areas with strong policy frameworks. For example, targeted policy improvements resulting in a seven-point increase in the TEASE-16 score would reduce per capita consumption in the Philippines -per 1000 international dollars of gross domestic product -by 0.19 litres but the corresponding reduction in Japan would only be 0.09 litres.
Like Brand et al., 13 we found that study areas with more stringent -and strongly enforced -alcohol policies had significantly lower levels of consumption. Although we recognize that there are many structural and contextual factors influencing the extent and patterns of alcohol consumption, the results indicate that alcohol consumption rates -whatever their causes and even after controlling for differences in income -are at least partially related to the strength of national alcohol control policies.
In addition to the need for ongoing social and treatment programmes for individuals and communities at high risk of alcohol-related harm, WHO encourages its Member States to regularly assess and refine their alcohol control policy frameworks. 4,6 This paper highlights considerable scope for strengthening policies in the western Pacific, particularly in relation to the advertising and physical availability of alcohol.
Our study has several limitations. First, considerable heterogeneity exists in alcohol policies and cultural differences may affect the level of alcohol consumption per capita. These differences are not captured in TEASE-16, which focuses on formal, national

Research
Alcohol control policies in the western Pacific Natacha Carragher et al.
alcohol policies. Nevertheless, where possible, we attempted to minimize the effects of differences within study areas. For Australia, for example, we collected data from all eight states and territories for each policy topic and then used the general consensus to reflect the national position. Second, although TEASE-16 assesses a panoply of alcohol control policies, it does not cover the full spectrum of policies. However, as Karlsson and Österberg point out, 11 it would be laborious -if not impossible -to do this, as there are over 100 relevant policies.
Third, it is well recognized in the literature that there is a close relationship between national affluence and alcohol consumption. 31 For this reason, we adjusted consumption estimates to take account of each study area's per capita gross domestic product. However, the extent and patterns of alcohol consumption in any nation are not entirely determined by regulatory framework and affluence. 31,32 Many other factors (e.g. socioeconomic factors, physical environment, biological and genetic factors, access to health-care services and facilities, and individual characteristics) are involved. Any observed disparities in consumption patterns may therefore result from the complex interplay of a variety of structural and contextual factors. In designing and implementing effective alcohol control policies, it is important to account for this panoply of mitigating factors and to adopt a coordinated response. Fourth, the cross-sectional nature of this study means that a causal relationship between alcohol policy scores and income-adjusted estimates of per capita alcohol consumption cannot be inferred. Fifth, cross-national comparisons will necessarily restrict sample size because the collection of policy data and the cross-referencing of sources are so time-consuming. The use of small sample sizes reduces statistical power and increases the likelihood of potential bias from outliers. Accordingly, caution should be exercised in extrapolating this study's findings beyond the study areas examined. However, since our finding of an inverse relationship between alcohol policy scores and alcohol consumption remained robust across 12 alternative weighting schemes -and matches the conclusions drawn by Brand et al., 13 who analysed alcohol policies in 30 countries -we can conclude there was little bias in our study.
Sixth, although enforcement is a critical component of policy evaluation, its measurement presents a challenge due to the difficulty in securing objective data. To minimize bias, we attempted to verify enforcement data by cross-referencing information with numerous officials and against relevant statistics. Finally, while TEASE-16 appears to be reliable and have content, face and criterion validity; construct validity and test-retest reliability have yet to be established.
Despite its shortcomings, TEASE-16 has numerous benefits. First, TEASE-16 provides an updated, empirical synopsis of national policies across several study areas. Second, by reducing a vast amount of data into a single score, the tool is useful for facilitating communication with the general public, public health advocates and policy-makers. Third, TEASE-16 overcomes limitations of previous alcohol policy scales. In the future, TEASE-16 could be used to conduct a more nuanced examination of the relationship between targeted policies (e.g. measures taken against driving while under the influence of alcohol) and specific outcomes (e.g. numbers of alcohol-related road traffic accidents and fatalities).

Research
Alcohol control policies in the western Pacific Natacha Carragher et al.