Article Text

The effect of an annual temporary abstinence campaign on population-level alcohol consumption in Thailand: a time-series analysis of 23 years
  1. Udomsak Saengow1,2,3,
  2. Roengrudee Patanavanich4,
  3. Paibul Suriyawongpaisal4,
  4. Wichai Aekplakorn4,
  5. Bundit Sornpaisarn5,6,7,
  6. Huan Jiang5,6,8,
  7. Jurgen Rehm5,6,8,9,10,11
  1. 1Center of Excellence in Data Science for Health Study, Walailak University, Tha Sala, Nakhon Si Thammarat, Thailand
  2. 2Research and Innovation Institute of Excellence, Walailak University, Tha Sala, Nakhon Si Thammarat, Thailand
  3. 3School of Medicine, Walailak University, Tha Sala, Nakhon Si Thammarat, Thailand
  4. 4Department of Community Medicine, Faculty of Medicine, Ramathibodi Hospital, Mahidol University, Bangkok, Thailand
  5. 5Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
  6. 6Institute for Mental Health Policy Research, Centre for Addiction and Mental Health, Toronto, Ontario, Canada
  7. 7Faculty of Public Health, Mahidol University, Bangkok, Thailand
  8. 8Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, Ontario, Canada
  9. 9Department of Psychiatry and Institute of Medical Science, University of Toronto, Toronto, Ontario, Canada
  10. 10Center for Interdisciplinary Addiction Research (ZIS), Department of Psychiatry and Psychotherapy, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
  11. 11Program on Substance Abuse & WHO CC, Public Health Agency of Catalonia, Barcelona, Spain
  1. Correspondence to Dr Udomsak Saengow; saengow.udomsak{at}gmail.com

Abstract

Rationale A small number of earlier studies have suggested an effect of temporary abstinence campaigns on alcohol consumption. However, all were based on self-reported consumption estimates.

Objectives Using a time series of 23-year monthly alcohol sales data, this study examined the effect of an annual temporary abstinence campaign, which has been organised annually since 2003 during the Buddhist Lent period (spanning 3 months), on population-level alcohol consumption.

Methods Data used in the analysis included a time series of monthly alcohol sales data from January 1995 to September 2017 and the midyear population counts for those years. Generalised additive models (GAM) were applied to estimate trends as smooth functions of time, while identifying a relationship between the Buddhist Lent abstinence campaigns on alcohol consumption. The sensitivity analysis was performed using a seasonal autoregressive integrated moving average with exogenous variables (SARIMAX) model.

Intervention The Buddhist Lent abstinence campaign is a national mass media campaign combined with community-based activities that encourages alcohol abstinence during the Buddhist Lent period, spanning 3 months and varying between July and October depending on the lunar calendar. The campaign has been organised annually since 2003.

Main outcome Per capita alcohol consumption using monthly alcohol sales data divided by the midyear total population number used as a proxy.

Results Median monthly per capita consumption was 0.43 (IQR: 0.37 to 0.51) litres of pure alcohol. Over the study period, two peaks of alcohol consumption were in March and December of each year. The significant difference between before-campaign and after-campaign coefficients in the GAM, −0.102 (95% CI: −0.163 to –0.042), indicated an effect of the campaign on alcohol consumption after adjusting for the time trend and monthly seasonality, corresponding to an average reduction of 9.97% (95% CI: 3.65% to 24.18%). The sensitivity analyses produced similar results, where the campaign was associated with a decrease in consumption of 8.1% (95% CI: 0.4% to 15.7%).

Conclusions This study demonstrated that the temporary abstinence campaign was associated with a decrease in population-level alcohol consumption during campaign periods. The finding contributed to a growing body of evidence on the effectiveness of emerging temporary abstinence campaigns.

  • Epidemiology
  • Public Health
  • Prevention strategies
  • Mental Health & Psychiatry
  • Health education and promotion

Data availability statement

Data are available upon reasonable request. Data may be obtained from a third party and are not publicly available. The data used in this study was acquired from the Excise Department and the Bank of Thailand with the necessary permissions. Sharing this data requires obtaining permission from both organisations upon a reasonable request to the corresponding author.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • We conducted a systematic search without time limitations for all reviews on ‘alcohol’ and ‘temporary abstinence’ in multiple databases, and received only one result. Widening the search terms to include reviews on ‘alcohol’, ‘abstinence’ and various terms for ‘initiative’, ‘campaign’ resulted in 95 results, of which we retrieved three narrative reviews. The reviews and key articles reviewed indicated decreases in consumption and some health benefits of temporary abstinence, but were based on self-report and self-selected samples. No population-level associations were reported.

WHAT THIS STUDY ADDS

  • This is the first study to show population-level evidence for a repeated temporary abstinence campaign across the years studied. Using monthly per capita data on alcohol consumption for Thailand based on taxation records over 23 years, we were able to demonstrate that the Buddhist Lent campaign in effect since 2003 was associated with average reductions of drinking of approximately 10%, compared with the average drinking amount during Buddhist Lent before the campaign was initiated.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Combining an abstinence campaign with cultural traditions in a country resulted in repeated reductions in the consumption of alcohol. Further research is necessary to determine the underlying processes and longer-term impact of such campaigns.

Introduction

Alcohol consumption has been identified as a leading modifiable health risk in all comparative risk assessments to date1; in 2016, it accounted for about 3 million deaths globally.2 While most of these deaths are due to non-communicable diseases, alcohol also has a strong impact on intentional and unintentional injuries, making it the most important risk factor for mortality in adolescents and young adults.3 Several effective and cost-effective measures have been recommended to control alcohol consumption and attributable harm, including increasing taxes and prices, regulating availability and imposing marketing restrictions.4 5

As an alternative to these recommended measures at the population level, campaigns challenging people to temporarily refrain from drinking have emerged in recent years. Examples include ‘Dry January’ in the UK, ‘Dry July’ in Australia and the ‘Buddhist Lent abstinence campaign’ in Thailand, among others.6–9 Most campaigns persuade participants to abstain for 1 month, with the exception of the Thai campaign, which encourages 3 months of abstinence.

The Buddhist Lent abstinence campaign has been organised annually in Thailand since 2003. Its goal is to encourage a 3-month abstinence period during the Buddhist Lent period—a period between July and October with exact dates varying each year depending on the lunar calendar. A national-level mass media campaign is accompanied by community-based activities (eg, peer group support, a ceremony for the public’s commitment to abstain and the presentation of awards for successful abstinence) to promote drinking abstinence.10 11 The theme of the campaign is changed every year, and has alternated between four major themes: peer-supported abstinence, abstaining for your health, abstaining for your family and Buddhism-related themes. The campaign has been included in Thailand’s National Alcohol Strategy as a ‘norm-altering’ activity.12 Such activities can be classified as education and persuasion strategies.13 This policy area has been widely regarded as ineffective on its own, and recommended only to support other effective policy options.13–15

Nevertheless, temporary abstinence campaigns may differ from other options under the same policy area as a small but growing body of evidence suggests some effectiveness in reducing consumption and improving health.9 16 17 A prospective study of Dry January participants, who self-enrolled on the campaign website, demonstrated that the campaign led to a reduction in drinking frequency, drinking quantity and intoxication episodes. An increase in drink refusal self-efficacy was also observed.18 Another study on Dry January prospectively followed campaign participants and non-participants. The study found improvement in self-reported physical and mental health in successful campaign participants; these effects were not observed in non-participants.19 Qualitative studies suggested that positively motivating participants by portraying them as heroes and offering an avenue for ethical self-improvement may underpin the effectiveness of the campaign.20 21

For the Buddhist Lent abstinence campaign in Thailand, a nationally representative survey estimated that 5.8 million drinkers completely refrained from drinking during a 3-month campaign period in 2016.8 Campaign-related activities carried out at the community level helped in maintaining abstinence beyond the campaign period.22

The effectiveness of temporary abstinence campaigns has still been questioned as the supporting evidence has been based solely on self-report measurements,23 often assessed in self-selected samples. In this study, we used a time series of 23 years of monthly alcohol sales data to evaluate the effect of the Thai temporary abstinence campaign. An advantage of using sales data is that it is more accurate in estimating alcohol consumption at the population level compared with an estimate arrived at from survey responses.24 The objective of this study was to examine whether the campaign was associated with a reduction in population-level alcohol consumption.

Methods

Data sources

The data used in the analysis included monthly alcohol sales data from January 1995 to September 2017 and midyear population levels for these years. Monthly alcohol sales data were used as a proxy for alcohol consumption. The sales data were obtained from two sources according to data availability at each period. Primarily, sales data is collected by the Excise Department, Ministry of Finance and sent to the Bank of Thailand and other government agencies for economic reporting and forecasting purposes. According to its internal protocol, the Excise Department was able to provide data from January 2010 to September 2017 in response to our formal data request. Alcohol sales data from January 1995 to December 2009 were obtained from the Bank of Thailand database as an alternative data source. Midyear populations of those aged 15 years and above were obtained from the Department of Provincial Administration database.25

Study design

To evaluate the effectiveness of temporary abstinence campaigns, a quasi-experimental study design was undertaken—more specifically, an interrupted time-series design was used.26 Alcohol sales data was collected before and after the campaigns from January 1995 to September 2017. It includes a total number of 273 data points, providing sufficient statistical power to detect an effect.27 Given that the data set seemed to have more complex trends than linear ones, generalised additive models (GAM) were applied to estimate trends as smooth functions of time, while identifying a relationship between the Buddhist Lent abstinence campaign on alcohol consumption.28

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Data analysis

Alcohol sales data originally reported in litres of alcoholic beverages were converted to litres of pure alcohol. The pure alcohol data was divided by the midyear population of those aged 15 years and older to obtain monthly per capita consumption. Buddhist Lent periods are based on the Thai lunar calendar. The period covers approximately 3 months with the exact start and end dates varying between the months of July to October. The impact of the Buddhist Lent months was captured as the days of Buddhist Lent in a month divided by the total number of days in that month. As a result, possible values ranged from 0 to 1. For example, in 2016, the Lent period fell between July 20 and October 16; the number of days in the Lent period months were 12, 31, 30 and 16 days for July, August, September and October, respectively. As a result, values for the Buddhist Lent variable were 0.39 (July), 1.00 (August), 1.00 (September), and 0.52 (October). For other months of 2016, values were zero.

The variable of interest was the effect of the Buddhist Lent campaign on consumption. Conceptually, this variable has two components: (1) the pure effect of Buddhist Lent without any campaign; and (2) the effect of the campaign itself. These components were operationalised as follows: the pure effect of Buddhist Lent was measured by the Buddhist Lent variable described above in the years without a campaign; as the campaign was run in all years since 2003, this would be the years prior to 2003. The coefficient associated with this variable would measure how Buddhist Lent periods differed from other periods in the year. Since 2003, the values of the Buddhist Lent variable captured the composite effect of Buddhist Lent plus the campaign, as the campaign took place in all years since 2003. The net effect of the campaign could then be obtained by subtracting the effect of Buddhist Lent before the campaign from the composite effect, that is, the Buddhist Lent variable after the campaign started.

Accordingly, operationally, the Buddhist Lent variable was divided into two variables depending on the time axis, that is, Buddhist Lent v1 before 2003, and Buddhist Lent v2 for 2003 and after. The first variable only had values for the proportion of Buddhist Lent days in each month up to December 2003 and was 0 thereafter, the second variable had 0 for each month up to December 2003 and then the proportion of days that fell within Buddhist Lent thereafter. An example of the coding of the two variables (for the years 2003 and 2004) is presented in table 1.

Table 1

Example of the coding of the two Buddhist Lent variables in 2003 and 2004

At first, an exploratory analysis was conducted to break the time series down into three components: seasonality, trend and remainder, to show the overall trend and how the seasonal components changed over time. To further test for an effect of the campaign on alcohol consumption, we performed interrupted time-series analyses by employing a GAM, which included both before-campaign and after-campaign variables and smooth functions of time and monthly seasonality. The outcome, alcohol consumption per capita, was log-transformed to reduce the skewness of the variable. This also ensures that data meets the underlying assumption of normally distributed error terms with constant variance. Types of smooth functions were selected to achieve the best fit of the data based on the Akaike information criterion and Bayesian information criterion statistics.29 Residuals were examined for autocorrelation using the Box-Jenkins method. The autoregressive integrated moving average (ARIMA) model was applied where necessary; ARIMA (p, q) was added with p auto-regressive term and q moving average term using the auto.arima() function (forecast package) in R .3.6.3.30 31

Once the final GAM was determined, a comparison of the coefficients for before-campaign and after-campaign variables was performed to test whether the campaign had an effect on population-level alcohol consumption. To quantify the effect of the campaign on consumption, the difference of the coefficients (and 95% CI) were calculated using the Wald χ2 test.32 Since the preliminary investigation found a notable hike in alcohol consumption in December 1999 and it was due to the privatisation of a state-owned alcohol company in January 2000,33 the analyses excluded data from December 1999.

As a sensitivity analysis of campaign effects, the seasonal autoregressive integrated moving average with exogenous variables (SARIMAX) model was used to confirm the findings with the differenced time series as a dependent variable and the campaign variable as an independent variable. Additionally, a second sensitivity analysis was conducted, encompassing the entire study period (which includes the data from December 1999). All the described analyses assume an immediate effect resulting from the campaign, which would continue as there were campaigns in every year since 2003. In the third sensitivity analysis, we examined potential slope changes, that is, the effect of the campaign would weaken or strengthen over time. Finally, in the fourth sensitivity analysis, we included the two most important alcohol control policy changes in the main time series model: the Alcoholic Beverage Control Act (including availability restrictions, and a ban on advertising, promotion and sponsorship), which came into effect on 14 February 2008, as well as the major change in the excise taxation system implemented on 4 September 2013. Both were included as immediate and lasting effects (for classifying the potential effects of alcohol control policies, see the study by Rehm et al34).

Results

The time series covered a period of 23 calendar years. There were 56 months in the data set which overlapped with campaign periods. Median monthly per capita consumption over the study period was 0.43 litres of pure alcohol (IQR: 0.37 to 0.51). Table 2 summarises the variables used in the analysis.

Table 2

Characteristics of variables in the time series

A plot of the time series is presented in figure 1. Overall, per capita consumption tended to increase over the study period, especially from the year 2000 onward, and then to level out after 2015. An increase in volatility of data from 2011 onward could be attributed to the restructuring of alcohol taxation in 2013 and the preceding adjustment in tax rates for various categories of alcoholic beverages in 2012. Campaign periods are illustrated by solid lines and tend to be lower than the non-Lent months of the same year.

Figure 1

Monthly per capita consumption of alcohol (January 1995–September 2017) with campaign period (solid line).

Figure 2 shows the time series broken down into three components: seasonality, trend and remainder. The seasonal component has two peaks in March and December of each year. For the remainder component, there was a notable hike in December 1999, a consequence of the privatisation of a state-owned alcohol company that became privately owned in January of 2000.33 This company released a large amount of alcoholic beverages into the market late in 1999 in order to resolve taxation issues arising from the privatisation.35

Figure 2

Decomposition of time series: original time series (top panel), seasonal component (second panel), trend (third panel) and remainder (bottom panel).

The GAM (table 3) shows that both the monthly seasonality and the overall time trend had a significant impact on alcohol consumption. The plot of smooth functions (online supplemental figure 1) further confirms that there were two peaks in March and December of each year and that alcohol consumption increased over the study period. After adjusting for seasonal effects and time trends, both before-campaign and after-campaign variables were not significantly different from 0. However, these effects moved in different directions. Before the campaign, the effect of the Buddhist Lent on alcohol consumption was estimated to be 0.029 (95% CI: −0.073 to 0.131), corresponding to an increase of 2.94% (95% CI: −7.03% to 14.00%) in alcohol consumption during Lent. The effect was −0.073 (95% CI: −0.169 to 0.023), indicating a decrease in alcohol consumption of 7.03% (95% CI: −2.27% to 15.55%) during Lent from 2003 (the first year of the campaign) onward. This suggests a potentially significant change in the difference in alcohol consumption between these two periods, and this difference is interpreted as the campaign effect. To formally test the hypothesis of the difference in alcohol consumption during the Buddhist Lent with and without the campaign, a χ2 test was performed; the difference is statistically significant (p<0.001). The estimate of the difference was −0.102 (95% CI: −0.163 to –0.042). This estimate corresponds to an average reduction in overall alcohol consumption of 9.97% (95% CI: 3.65% to 24.18%) during the Buddhist Lent with the campaign compared with the before-campaign period (see the study by Yang36 for details of converting regression coefficients and the online supplemental appendix for the calculation).

Supplemental material

Supplemental material

Table 3

GAM statistics and Wald χ2 test for the effects of the campaign on alcohol consumption per capita

The first two sensitivity analyses produced similar results. With SARIMAX model, after transforming the coefficient to the original scale, the campaign was significantly associated with a 0.0347 litre reduction in per capita alcohol consumed per month (95% CI: −0.0676 to –0.0018), which corresponded to a reduction of 8.1% of the median monthly per capita consumption (95% CI: 0.4% to 15.7%), with a large overlap with the main analysis described above (for details, see online supplemental appendix sensitivity analysis I).

The second sensitivity analysis, which comprised the entire study period, also generated similar estimations as the main analysis (see online supplemental appendix sensitivity analysis II). In addition, a formal test for changes in the impact of the campaign over time did not find any significant results, that is, the impact of the campaign, as examined via a change in the slope parameter, did not increase or decrease significantly over time. Moreover, the addition of the potential slope changes worsened the fit of the model overall (see online supplemental appendix sensitivity analysis III). Finally, the inclusion of the two major alcohol control policy changes did not affect the major results and conclusions of our study.

Discussion

This study evaluated the effect of a temporary abstinence campaign on per capita consumption of alcohol in Thailand using a time series of alcohol sales data. The main finding was that the Buddhist Lent abstinence campaign accounted for a 9.97% reduction in alcohol consumption at the population level compared with the same period before the campaign. Interestingly, the campaign did not decrease long-term consumption; in each year, consumption was decreased during the Buddhist Lent period, and resumed outside of the period.

This finding was consistent with previous survey-based assessments of the same campaign. Based on a national survey in 2016, one-third of current drinkers successfully abstained for 3 months, and reduced their consumption by 35% during the campaign period. Those who drank less frequently prior to the campaign had a 2.9 to 4.1 greater likelihood of actively participating in the campaign compared with regular drinkers.8

Contrary to our results showing no potential long-term decrease in consumption, a previous study found that the campaign’s effect lasted longer than the Lent period, that is, for about 6 months. In this community-based study, the rate of abstinence at the 6 month follow-up was higher in communities that participated in campaign activities, compared with control communities that did not provide these activities. The difference between the two types of communities disappeared by the ninth month.22 The differences between our results and this earlier study are likely due to the longer time frame of our study, as well as the objective measurement of alcohol consumption on a national level.

Significant decreases in drinking frequency, drinking quantity and episodes of drunkenness were also observed after 6 months in Dry January participants as well as improvement in self-reported mental and physical health scores.18 19 However, these results may be biased since self-reports can be biased (for an overview, see the study by Althubaiti37).

Our study design avoided the various biases of self-report but there are potential limitations associated with the use of aggregate data. Thus, our study was not able to identify the degree to which each group within the Thai population contributed to the decrease in alcohol consumption. Moreover, this study cannot determine the extent to which the decrease in overall consumption was attributed to complete abstinence, a decrease in drinking frequency or a decrease in quantity consumed per drinking occasion. Finally, the motivational and cognitive pathways leading to abstinence are not clear.

Although the campaign was associated with a decrease in consumption, the overall level of consumption in Thailand actually increased during a portion of the study period. An increasing trend in alcohol consumption in low- and middle-income countries (LMICs) has long been recognised.4 38 39 The slope was steeper compared with historical data from high-income countries. A lack of alcohol control policies and the multinational alcohol industry played a crucial role in augmenting alcohol consumption in LMICs.39 40 The Thai government has implemented the WHO’s recommended alcohol control policies such as increasing excise tax, restricting availability and enforcement of a partial ban on marketing activities.12 However, these policies along with the temporary abstinence campaign were not able to counter the increasing trend. It would be interesting to estimate the impact all of these alcohol control policies had, for example, by comparing the impact to neighbouring countries that implemented fewer policies and had a larger increase in their consumption.41 42

How widely can these results be generalised? Thailand is still a country where the majority of adults abstain,43 so abstaining does not create the same social pressures as in countries where a large majority of people consume alcohol, such as Australia.44 As well, it seems that Buddhist Lent alone (ie, without campaigns or community support) is not sufficient to reduce drinking. Future research, including experimental research, should attempt to better identify the most successful components employed as part of the Thai Lent campaigns.

In conclusion, the present study confirmed that changes in consumption at the individual level during the temporary abstinence campaign period in Thailand observed in previous surveys actually translated into a reduction in overall consumption at the population level. The findings of this study thus contribute to a growing body of evidence on the effectiveness of emerging temporary abstinence campaigns, and helps quantify the potential population effects of such campaigns if rooted in community support and cultural and religious values.

Supplemental material

Data availability statement

Data are available upon reasonable request. Data may be obtained from a third party and are not publicly available. The data used in this study was acquired from the Excise Department and the Bank of Thailand with the necessary permissions. Sharing this data requires obtaining permission from both organisations upon a reasonable request to the corresponding author.

Ethics statements

Patient consent for publication

Ethics approval

The study protocol was approved by the Human Research Ethics Committee of Walailak University (Approval Number WUEC-19-009-01).

Acknowledgments

The authors would like to thank the Bank of Thailand and Excise Department for provision of monthly alcohol sales data. We would also like to thank Dr Thammasin Ingviya, Faculty of Medicine, Prince of Songkla University, for his initial advice on time-series data analysis.

References

Supplementary materials

Footnotes

  • Handling editor Helen J Surana

  • X @SaengowUdomsak

  • Correction notice The published version misspelled co-author’s name as Paibul Suriyawongpaisul. The correct name should be Paibul Suriyawongpaisal.

  • Contributors US helped design the study, helped analyse the data and wrote the first draft of the manuscript. RP helped design the study, obtained permission to access the data and helped analyse the data. PS and WA helped design the study and provided advice on data analysis. HJ conducted the main data analysis. BS assisted in interpreting the results. JR provided advice on the data analysis and drafted the second version of the paper. US, WA, BS and JR acquired grants to help finance the study. US is a guarantor of this article. All authors contributed and agreed to the final version of the paper.

  • Funding This work is supported by the Thai Health Promotion Foundation (contract number 60-00-1746 and contract number 65-00345), the Center for Alcohol Studies, Thailand (grant number 61-02029-0104, 61-02029-0113 and 65-10068-12), and the US National Institute on Alcohol Abuse and Alcoholism (NIAAA) (grant number 1R01AA028224).

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.