Abstract
Background and aims
Alcohol use has been identified as a major risk factor for burden of mortality and disease, particularly for countries in eastern Europe. During the past two decades, several countries in this region have implemented effective alcohol policy measures to combat this burden. The aim of the current study was to measure the association between Lithuania’s alcohol control policies and adult all-cause mortality.
Design
Interrupted time–series methodology by means of general additive models.
Setting
Lithuania.
Participants
Adult population of Lithuania, aged 20 years and older.
Measurements
Alcohol control policies were ascertained via a document review of relevant legislation materials. Policy effects were evaluated as follows: (1) slope changes in periods of legislative (non-)activity with regard to alcohol control policy (analysis 1); (2) level changes of three interventions following recommendations of the World Health Organization (analysis 2); and (3) level changes of seven interventions judged a priori by an international panel of experts (analysis 3). Mortality was measured by sex-stratified and total monthly age-standardized rates of all-cause mortality for the adult population.
Findings
During the period 2001–18, effective alcohol control policy measures were implemented on several occasions, and in those years the all-cause mortality rate declined by approximately 3.2% more than in years without such policies. In particular, the implementation of increased taxation in 2017 was associated with reduced mortality over and above the general trend for men and in total for all analyses, which amounted to 1452 deaths avoided (95% confidence interval = −166 to −2739) in the year following the implementation of the policy.
Conclusions
Alcohol control policies in Lithuania appear to have reduced the overall adult all-cause mortality over and above secular trends.
Keywords: Alcohol, alcohol control policy, all-cause, best buys, death, mortality
INTRODUCTION
Alcohol use is well established as one of the major risk factors for mortality and burden of disease, ranking among the top 10 avoidable risk factors in every global comparative risk assessment to date [1,2]. As such, there are a number of proven effective and cost-effective policy measures available to reduce this burden [3].
However, the impact of such alcohol control policies is rarely strong enough for all-cause mortality for the entire population to be positively affected immediately following implementation. For example, in the classic systematic review by Wagenaar and colleagues [4] on the effects of alcohol taxation and price policies on morbidity and mortality only one study reported on all-cause mortality, and the effect of taxation in this study was not significant. However, all-cause mortality in a population is arguably the most important yardstick for measuring the impact of alcohol policies, as it is directly linked to the life expectancy of a population and as it cannot be argued that the increases in cause-specific mortality are balanced by decreases in other cause-specific mortality.
The situation in the Russian Federation provides one of the few previous contradictory examples of where all-cause mortality has been affected by the implementation of alcohol control policies [5,6]. In Russia, alcohol use has long been identified as having a strong influence on all-cause mortality [7], and natural experiments such as the Gorbachev Reform [8] and ecological studies (e.g. [9]) have shown an impact from alcohol control policies on all-cause mortality.
Accordingly, the aim of the current study was to test the relationship between the implementation of alcohol control policies and adult all-cause mortality in Lithuania. Lithuania is a post-Soviet country in northeastern Europe, where alcohol has also been identified as a major risk factor for disease and mortality burden [10] and where the World Health Organization’s (WHO) recommended effective and cost-effective alcohol control policies (i.e. the so-called ‘best buys’ of taxation increase, availability restrictions and bans on marketing [3,11]) have been implemented during the past several years [12,13].
In short, the hypothesis underlying this analysis was that the implementation of alcohol control policies reduced all-cause mortality for Lithuania on a population level. In addition, we postulated stronger effects for men compared to women, based on the higher level of use and the more harmful patterns of drinking in the former. We aimed to separate two effects: (i) a change in all-cause mortality in the years of implementation of alcohol control policies which have been independently judged to be very effective (specifically, taxation increases above a given threshold, availability reductions and a ban on advertisement and marketing; analysis 1) compared to years without such policy implementation and (ii) an immediate and continuing effect of alcohol control policy implementation using interrupted time–series analyses ([14], analyses 2 and 3).
MATERIALS AND METHODS
The current study is an observational design, based on secondary all-cause mortality data for Lithuania. Monthly mortality and population data for 2001–18 (n = 216 months) were obtained from the Lithuanian Institute of Hygiene. Data was obtained for adults aged 20 years and older.
Identification of alcohol policies
All alcohol control policy measures during the time-period under consideration are described in detail elsewhere [13] and are listed in Table 1. The procedures used to select the policy being modeled are described in detail elsewhere (see [15]), but we provide a short overview in the following. In analysis 1, we examined whether the overall trajectory of all-cause mortality rates followed the pattern which would be expected based on the legislative activity with regard to alcohol policies. Specifically, we defined the following four periods, analogously to a previous time–series of alcohol-related motor vehicle accidents in Lithuania [16]; see also Table 1 below and [15]): (1) January 2001–December 2007—no effective alcohol control policy implemented, liberalization of the market; (2) January 2008–December 2009—two key alcohol control policies implemented, period of economic recession; (3) January 2010–March 2014—period of inaction with regard to alcohol control policies, period of economic growth; and (4) April 2014–December 2018—implementation of the WHO’s three ‘best buys’ of alcohol control policy. In this analysis, we only evaluated slope changes of the four distinct periods, while not adjusting for any economic indicators.
Table 1.
Alcohol control policies implemented between 2001 and 2018 in Lithuania.
| No. | Date of policy implemented | Type of policy | Analysis 1 | Policy score for analysis 3 |
|---|---|---|---|---|
| 1 | 1 June 2001 | Taxation/price (increase excise tax by 6% for ethyl alcohol, change of taxing format) | 0 | 0.5 |
| 2 | 28 November 2001 | Availability (liberalization of sales and production, sales of alcohol in petrol stations allowed) | 0 | 1.2 |
| 3 | 28 June 2002 |
|
0 | 0.0 |
| 4 | 1 July 2002 | Taxation/price (exemption of excise tax for small breweries) | 0 | 0.3 |
| 5 | 1 May 2003 | Drink-driving (criminal liability restored in certain cases when individuals are harmed or property is damaged significantly) | 0 | 2.0 |
| 6 | 16 July 2003 | Marketing/advertising (liberalization of advertising by expanding the range of display places) | 0 | 0.0 |
| 7 | 1 January 2004 | Taxation/price (decline of excise tax of a Subcategory of fermented beverages when equalizing the tax with other similar categories) | 0 | 0.0 |
| 8 | 1 May 2004 | Taxation/price (four alcohol beverage categories formed, licensing changes when joining the EU) | 0 | 0.3 |
| 9 | 1 January 2008 |
|
1 | Three measures Avg. 4.6 |
| 10 | 1 January 2009 |
|
1 | Two measures Avg. 5.7 |
| 11 | 1 April 2014 | Taxation/price (increase excise tax by 10–47%; 1% for ethyl alcohol) | 1 | 6.0 |
| 12 | 1 January 2015 | Drink-driving (0% BAC for select drivers) | 0 | 4.2 |
| 13 | 1 March 2015 | taxation/price (increase excise tax by 10–16%; 2% for ethyl alcohol) | 1 | 6.5 |
| 14 | 1 January 2016 | Availability (banned sales at petrol stations) | 1 | 5.0 |
| 15 | 1 March 2016 | Taxation/price (increase excise tax by 8%; 2.5% for ethyl alcohol) | 0 | 4.7 |
| 16 | 1 January 2017 | Drink-driving (> 0.15% BAC = criminal offense) | 0 | 3.3 |
| 17 | 1 March 2017 | Taxation/price (increase excise tax: 111–112% for wines and beer; 23% for ethyl alcohol) | 1 | 9.5 |
| 18 | 1 January 2018 |
|
1 | Two measures, average 6.1 |
Gray-highlighted rows indicated time-points with expert judgment of ≥ 5 for at least one of the policies implemented. BAC = blood alcohol concentration.
In two additional analyses (analyses 2 and 3), we linked mortality changes to specific alcohol control policies. For this purpose, we selected measures with established evidence of effectiveness (and cost-effectiveness) according to WHO’s ‘best buy’ evaluations, which are linked to immediate effects for large parts of the population, i.e. restrictions in availability and increases in price (which, in the context of Lithuania, concerned increases in taxation). For the latter, we further restricted the selection to measures linked to actual decreases in affordability, which is a key variable in the pathway to actual effects. This resulted in three key alcohol control policies being selected (1 January 2008, 1 March 2017 and 1 January 2018), which were analysed in analysis 2. For analysis 3, a group of alcohol control policy experts, not familiar with the mortality data in Lithuania and thus not biased by perceived associations between policies and mortality in Lithuania, rated all policies on an 11-point scale based on their perception of whether the policy would have an immediate impact on alcohol use and health [15]. Based on their expert assessment, we identified seven policy dates where policies had an average rating of 5 or higher, including the three dates identified for analysis 2, plus four additional dates (additional policy dates: 1 January 2009, 1 April 2014, 1 March 2015 and 1 January 2016).
Dependent variables
Monthly sex-stratified and total age-standardized all-cause mortality per 100 000 people were the dependent variables. We used this measure because it is the most important indicator for the potential impact of alcohol policies directly linked to life expectancy. No sensitivity analyses for 100% alcohol-attributable outcomes were presented, as coding for these diseases underestimates their real occurrence [17,18], and the guidelines in Lithuania for coding changed during the observation period [19].
Potential confounding variables
To adjust the time–series models in analyses 2 and 3, we considered the following (confounding) variables: centered monthly gross domestic product (GDP), expressed in Euros per capita (re-calculated from quarterly data obtained from Statistics Lithuania [20], centered unemployment rate (available by sex [21]), as well as a second-order polynomial for the trend (i.e. the square of the consecutive months from 1 to 216). The last variable helped to achieve stationarity in the time–series.
Data analyses
Interrupted time–series analyses, following the recommendations of Beard et al. [14], were conducted. We used general additive models (GAMs), which adjusted for seasonality using smoothing splines in all analyses. As illustrated in Supporting information, Figs S7 and S8, all-cause mortality rates peaked in winter months (December and January) and were lowest in summer months (June to September) for women, men and the total population. Further, autocorrelation was checked and corrected for (see Supporting information, Figs S7 and S8 for details). As illustrated in Supporting information, Fig. S1, the all-cause mortality rates were approximately normally distributed; thus, the response distribution of all models were specified as Gaussian.
We performed three sets of analyses, sex-stratified, and for the total population: in analysis 1 we examined slope changes in four distinct periods, as defined by alcohol control policy legislation activity; in analysis 2 we examined the impact of the first set of policies expected to have the largest impact; and in analysis 3 we examined the impact of the seven policies (including those three from analysis 2) rated by experts to be impactful.
For analysis 1, the models aimed to describe the underlying trends in all-cause mortality in the four identified periods of alcohol policy legislation. The models only included four covariates for each period, coded 0 outside the respective periods and 1–n for the n months that defined the period (e.g. 1–84 for the first 7 years, January 2001 to December 2007). As the residuals were non-stationary in these models (visual inspection and augmented Dickey–Fuller test), the models were fitted on the first-order differenced time–series of the age-standardized mortality rate (see Supporting information, Fig. S2).
For analyses 2 and 3, the models aimed to evaluate the association of all-cause mortality with select alcohol policies, which were entered as dummy variables. The variables were coded 0 for all months preceding the implementation and 1 for all months starting with the implementation of the respective policies, assuming an immediate and sustained impact. For instance, to measure the sets of policies implemented on 1 January 2008, the associated variable was coded 0 for all months between January 2001 and December 2007 and coded 1 for all months between January 2008 and December 2018. The final models were determined based on the following criteria: (a) accounting for at least one economic indicator (GDP or unemployment rate); (b) presence of non-stationarity in residuals (checked visually and confirmed with augmented Dickey–Fuller test and applying a more conservative threshold of P = 0.025); and (c) simplicity of models (i.e. the fewer variables the better). For a comprehensive description of the economic indicators, see Supporting information, Figs S3–S6; for more details on model selection procedures and model specifications (including autocorrelation), see Supporting information, Figs S7 and S8, Tables S1 and S2.
Please note that we estimated women, men and total population separately in all time–series analyses as an additional check to determine whether the combined effects for women and men were consistent with the effects for total population.
All analyses were performed using R version 4.0.2 [22] and the complete data and R code can be found in Supplementary File S1.
RESULTS
Overall, all-cause mortality decreased in the time-period observed, but particularly since 2008, when the implementation of effective alcohol control policy measures began (see Table 2). The overall decrease in age-standardized mortality rates between 2008 and 2018 were 22%, higher in men (−24%) than in women (−19%). Results of analysis 1 confirm the specific hypotheses of a decrease of all-cause mortality rates in Phases 2 and 4, i.e. years when alcohol control policies were implemented. However, slope changes in the time–series models were only significant for men, but not for women or both sexes combined (see Table 3).
Table 2.
Sex-stratified and total annual age-standardized all-cause mortality rates per 100 000, adult alcohol per-capita consumption and relative changes.
| Year | Women | Men | Total | APCb | ||||
|---|---|---|---|---|---|---|---|---|
| Rate (SD) | % changea | Rate (SD) | % changea | Rate (SD) | % changea | % changea | ||
| 2001 | 69.6 (5.7) | 145.1 (11.5) | 101.0 (7.9) | 10.5 | ||||
| 2002 | 68.5 (7.8) | −1.5 | 144.4 (11.0) | −0.5 | 100.0 (8.6) | −1.0 | 11.1 | 5.7 |
| 2003 | 66.3 (7.3) | −3.3 | 143.4 (12.9) | −0.7 | 98.0 (9.1) | −2.0 | 11.3 | 1.8 |
| 2004 | 65.5 (6.2) | −1.2 | 141.7 (9.9) | −1.2 | 97.3 (7.0) | −0.7 | 12.2 | 8.0 |
| 2005 | 67.7 (5.7) | 3.4 | 151.6 (10.8) | 7.0 | 102.9 (7.0) | 5.7 | 12.5 | 2.5 |
| 2006 | 68.9 (6.3) | 1.7 | 153.5 (11.3) | 1.3 | 104.4 (8.0) | 1.5 | 13.2 | 5.6 |
| 2007 | 68.0 (7.3) | −1.2 | 158.8 (10.3) | 3.4 | 106.0 (8.0) | 1.6 | 13.9 | 5.3 |
| 2008 | 66.4 (5.4) | −2.4 | 146.7 (13.0) | −7.6 | 99.9 (8.2) | −5.8 | 13.9 | 0 |
| 2009 | 61.8 (5.8) | −7.0 | 136.9 (9.9) | −6.7 | 92.8 (7.0) | −7.1 | 13.1 | −5.8 |
| 2010 | 61.1 (3.4) | −1.1 | 134.5 (8.2) | −1.8 | 91.6 (5.2) | −1.3 | 13.5 | 3.1 |
| 2011 | 59.2 (4.3) | −3.1 | 129.8 (6.9) | −3.5 | 88.5 (5.2) | −3.4 | 14.7 | 8.9 |
| 2012 | 58.1 (3.6) | −1.9 | 127.4 (7.7) | −1.8 | 86.9 (5.1) | −1.8 | 14.7 | 0 |
| 2013 | 58.3 (6.1) | 0.4 | 127.1 (10.5) | −0.2 | 86.9 (7.4) | 0.0 | 14.5 | −1.4 |
| 2014 | 55.7 (3.4) | −4.6 | 122.1 (6.3) | −4.0 | 83.2 (4.2) | −4.3 | 14.2 | −2.1 |
| 2015 | 57.9 (5.1) | 4.0 | 123.7 (7.7) | 1.4 | 85.1 (5.8) | 2.4 | 14.0 | −1.4 |
| 2016 | 56.2 (4.6) | −2.9 | 122.4 (7.3) | −1.1 | 83.5 (5.7) | −1.9 | 13.2 | −5.7 |
| 2017 | 55.5 (5.4) | −1.3 | 114.4 (10.9) | −6.5 | 79.5 (7.4) | −4.8 | 12.3 | −6.8 |
| 2018 | 53.9 (5.7) | −2.8 | 112.1 (9.8) | −2.0 | 77.8 (6.8) | −2.1 | 11.2 | −8.9 |
Years in which alcohol policies with expected high impact were implemented are highlighted in graya and calculated as the % change from the previous year. bRecorded annual pure alcohol consumption per capita among those aged 15 years and above. Official data from Statistics Lithuania [23]. APC = alcohol per capita; SD = standard deviation.
APC = alcohol per capita; SD = standard deviation.
Table 3.
Model results for the impact of different sets of alcohol policies on all-cause mortality.
| Analysis la | Women | Men | Total |
|---|---|---|---|
| Interceptb | 0.05 (−0.33, 0.43) | 0.21 (−0.36, 0.79) | 0.11 (−0.28, 0.51) |
| Period 1 | −0.003 (−0.01, 0.01) | −0.002 (−0.01, 0.01) | −0.002 (−0.01, 0.01) |
| Period 2 | −0.01 (−0.05, 0.03) | −0.07 (−0.14, −0.01)* | −0.04 (−0.08, 0.01) |
| Period 3 | −0.01 (−0.02, 0.01) | −0.02 (−0.05, 0.003) | −0.01 (−0.03, 0.004) |
| Period 4 | −0.005 (−0.02, 0.01) | −0.02 (−0.04, −0.002)* | −0.01 (−0.03, 0.002) |
| R 2 | 0.562 | 0.616 | 0.647 |
| Autocorrelationc | MA2 | MA2 | AR2, MA2 |
| Analysis 2a | Women | Men | Total |
| Interceptb | 61.59 (59.18, 64.00)*** | 159.53 (149.49, 169.57)*** | 93.04 (88.54, 97.54)*** |
| Months squared | NA | −0.001 (−0.002, −0.001)*** | NA |
| Unemployment rate (centered) | −0.71 (−1.02,−0.40)*** | −0.57 (−1.09,−0.05)* | −0.99 (−1.46,−0.52)*** |
| GDP per capita in 1000s Euros (centered) | −7.08 (−9.46,−4.70)*** | 8.78 (−0.86, 18.42) | −9.78 (14.14, 5.41)*** |
| Policy implementation dated | |||
| 1 January 2008 | 0.59 (−2.84, 4.02) | −6.01 (−15.11, 3.10) | 0.47 (−6.28, 7.22) |
| 1 March 2017 | −1.45 (−4.99, 2.08) | −9.13 (−17.23, −1.03)* | −7.52 (−13.42, −1.62)* |
| 1 January 2018 | 1.76 (−2.48, 6.00) | 6.44 (−4.01, 16.88) | 3.72 (−4.34, 11.78) |
| R2 | 0.801 | 0.842 | 0.812 |
| Autocorrelationc | AR1 | AR1, MAI | AR2, MAI |
| Analysis 3a | Women | Men | Total |
| Interceptb | 65.78 (63.64, 67.92)*** | 172.92 (163.92, 181.91)*** | 114.07 (108.68, 119.46)*** |
| Months^2 | NA | −0.002 (−0.003, −0.002)*** | −0.001 (−0.002, −0.001)*** |
| GDP per capita in 1000s Euros (centered) | −2.45 (−4.50, −0.40)* | 20.79 (12.99, 28.58)*** | 10.33 (5.69, 14.97)*** |
| Policy implementation dated | |||
| 1 January 2008 | 1.17 (−2.55, 4.88) | −8.74 (−16.85, −0.63)* | −4.38 (−9.05, 0.30) |
| 1 January 2009 | −7.31 (−10.40, −4.23)*** | 1.54 (−8.00, 11.08) | −0.78 (−6.38, 4.82) |
| 1 April 2014 | −0.74 (−3.88, 2.41) | 4.23 (−3.54, 12.00) | 3.35 (−1.45, 8.15) |
| 1 March 2015 | 0.58 (−3.36, 4.52) | 7.46 (−0.58, 15.50) | 4.12 (−0.82, 9.06) |
| 1 January 2016 | 0.15 (−3.82, 4.13) | 7.75 (−1.08, 16.57) | 4.15 (−0.97, 9.27) |
| 1 March 2017 | 2.29 (6.17, 1.58) | −11.07 (−18.88, −3.27)** | −5.36 (−10.10, −0.61)* |
| 1 January 2018 | 0.64 (−3.55, 4.84) | 6.97 (−2.12, 16.05) | 3.89 (−1.36, 9.13) |
| Ra | 0.804 | 0.857 | 0.866 |
| Autocorrelationc | AR1 | AR2 | AR1 |
P < 0.05;
P < 0.01;
P < 0.001;
bold numbers denote P < 0.05. Results from generalized additive models using the age-standardized all-cause mortality rates per 100 000 as dependent variable and adjusted for autocorrelation and seasonality. NA = not applicable, as this variable was not included in the final model.
In analysis 1, the model was performed on the first-order differenced mortality rates (i.e. the change from one month to the next). In analyses 2 and 3, the outcome variable remained unchanged.
In analysis 1, the intercept is close to 0, as the model was performed on the differenced time–series. In analyses 2 and 3, the intercept denotes the level of mortality in the beginning of the time–series in 2001.
The terms used to correct for autocorrelation in the time–series: AR (number) = autoregressive term of order (number), MA (number) = moving average term of order (number).
The values refer to the immediate and continuing changes after the time-points indicated.
Similarly, the recorded alcohol per capita (APC) consumption continued to increase until 2008 (see Table 2). The average APC increase during 2002–07 was 5.3%. In contrast, during the years with at least one expected high-impact alcohol control policy implemented (2008, 2009 and 2014–2018), there was an average decline of 4.4%.
The annual and monthly trajectories of age-standardized all-cause mortality rates between 2001 and 2018 are presented in Table 2 and Fig. 1, respectively.
Figure 1.
Age-standardized all-cause mortality rates by sex for the period 2001–18 in Lithuania. Gray lines indicate observed mortality rates, green lines indicate modelled mortality rates (from analysis 3). Vertical lines indicate alcohol policy implementation dates; vertical solid lines indicate three sets of policies with the highest expected impact; vertical dashed lines indicate additional four sets of policies rated to be impactful by experts. Light blue shaded areas indicate periods 1 and 3 during which no effective alcohol policies implemented, dark blue shaded areas indicate periods 2 and 4 during which several effective alcohol policies were implemented in each of the years
For each target population (women, men and total) the time–series models of analyses 3 (testing the impact of each of the seven policies separately) were found to demonstrate the best fit for the data.
As summarized in Table 3, the results of analysis 3 suggest that for both sexes together only the increases in excise taxation of March 2017 were significantly associated with a reduction of all-cause mortality. In March 2017 there were significant increases in excise taxes: 111–112% for fermented beverages, wines and beer, 91–94% for intermediate products and 23% for ethyl alcohol. As predicted, the effect was much stronger in men compared to women. In absolute numbers of deaths, this translated to 1346 [95% confidence interval (CI) = −2294 to −397] fewer all-cause deaths for men in the year after the policy implementation. Consistent with this, the estimated number of all-cause deaths in the time–series for the total population was estimated to be 1452 (95% CI = −2739 to −166) fewer deaths. Analysis 2 found similar results for the impact of these taxation increases.
Additionally, for men, the policy changes on 1 January 2008, were significantly associated with decreases in all-cause mortality, whereas the policy changes on 1 January 2009 were significantly associated with all-cause mortality decreases in women.
The high, and sex-specific, reductions in all-cause mortality associated with the implementation of stark increases in excise taxation on 1 March 2017 were corroborated by similar reductions in 100% alcohol-attributable mortality (see Fig. 2) and by a reduction of adult per-capita consumption in 2017 of 6.8%. In this year, all-cause mortality rates decreased by 1.3, 6.5 and 4.8% for women, men and both sexes combined, respectively. In all other years, average changes in all-cause mortality rates amounted to −1.5, −1.1 and −1.3% for women, men and both sexes combined, respectively.
Figure 2.
Sex-specific proportional changes in age-standardized all-cause and alcohol-specific mortality rates relative to March in each year for the period 2001–18 in Lithuania (black thin lines for all years other than 2017, blue thick lines for 2017; vertical blue line indicates March, i.e. month of implementation of 2017 policy)
Similarly, 100% alcohol-attributable mortality declined by 31.3, 8.7 and 14.8% for women, men and both sexes combined, respectively. In all other years, average changes in 100% alcohol-attributable mortality amounted to −0.5, −3 and −2.5% for women, men and both sexes combined, respectively.
DISCUSSION
The alcohol control policies which were anticipated to be highly impactful had a profound effect on all-cause mortality rates. In the years when such policies were in effect, the age-standardized rate decreased by approximately 3.2% more than in years when such policies were not in place. The marked excise taxation increase of 1 March 2017 was the single policy most associated with significant differences over and above the secular downward trend for Lithuania, leading to a decrease in the rate of mortality by 5.4 deaths per 100 000 (2.3 decrease in females, 10.1 decrease in males; resulting in an annualized decrease of approximately 4.8% or 1453 deaths). In sum, marked changes in population-level all-cause mortality seem to have resulted from the implementation of these alcohol control policy measures.
Our findings suggest a strong policy impact on all-cause mortality; however, such causal language would require more rigorous testing for alternative explanations. Even though the design controlled for seasonality, secular trends and economic variables, the use of ecological data is clearly a limitation for our study. There may be alternative explanations for some of the changes observed, including the economic recession which began at the end of 2008, but only fully impacted the country in 2009, resulting in reduced purchasing power [24]. However, most of the effects of the economic recession would be covered by the inclusion of the economic variables of per-capita GDP and unemployment.
While unlikely, it could be that in all 7 years involving the implementation of expected high-impact policies, other events took place which led to reductions in both alcohol use and all-cause mortality. Moreover, these confounders would have to have been absent in all the years without effective policy implementation. As indicated, such a pattern would be unlikely, but in any ecological analysis alternative explanations cannot be fully excluded (for further discussion, see [25]).
The next paragraph discusses the causal pathway via alcohol consumption. While adult per-capita data for Lithuania were not available on a monthly basis, we could compare our results with other results found in the literature. For countries in northern Europe, we would estimate a reduction of approximately 3% of all-cause mortality with each litre of alcohol adult per-capita consumption reduced [26,27]. How does this compare with the effects found in 2017? Overall, adult per-capita consumption was reduced by approximately 1 litre in the year after the taxation increase and the all-cause mortality decreased by 4.8%: a slightly higher, but still consistent, effect.
As observed in the time–series of all-cause mortality during the period of 2001–18, the peak was reached in 2007 (standardized all-cause mortality: 1089.4 deaths per 100 000), and has been declining since then (in 2018, it was 792.8 deaths per 100 000) [28]. The downward trend may have been influenced not only by the policies targeting alcohol control, but also by others, such as cardiovascular disease (CVD) prevention and cancer screening programs which were started in 2004 and 2005 [29], policies implemented to control tobacco use [30], to improve general traffic safety (e.g. the mandatory use of seatbelts since 2006) and to reduce suicide mortality [31], as well as possibly other qualitative and technological improvements achieved in the health-care system. All the above mortality categories are also known to be interlinked with the use of alcohol, but nonetheless are also influenced by their own policies. What we found in this analysis is that the specific alcohol control policies may have resulted in a considerably higher reduction in mortality over and above the general downward trend, and began in the exact year and month in which it was implemented.
We realize that better control could be achieved in simultaneous modeling of all the Baltic countries, using each country as a control for the other two, while also including cross-border trade. However, for such a model more preparatory work still needs to be performed internationally, and the data are not currently available.
Significant differences in mortality between the sexes in Lithuania are also notable. In 2017, Lithuania had a 9.8-year difference in life expectancy between males (life expectancy of 70.7 years) and females (80.5 years), one of the largest gaps seen across the European Union (EU). The large number of premature deaths for males is driven by risk factors such as alcohol use [32]. As avoidable mortality for Lithuanian males may be closely interlinked with excessive drinking, the implemented alcohol control policies provide a relatively larger effect on reducing male mortality than for females. The overall effect on all-cause mortality may also be larger due to a relatively worse starting-point for males in Lithuania compared to other countries (e.g. in terms of lower male life expectancy). Therefore, as observed, the major rise of life expectancy in Lithuania began after 2007 when the major alcohol control policies began to be introduced in Lithuania, and until 2018 life expectancy steadily increased by 6.41 years for males, but only by 3.42 for females [33,34]. The gap in mortality between the genders (together with the more detrimental drinking patterns of males) could play a role in a more statistically visible decline in all-cause mortality in males in 2008 and 2017, as observed in our analysis.
The most effective alcohol policy interventions—which led to an immediate change in mortality—were part of the WHO’s so-called ‘best buys’ [11] and involved a marked increase in excise taxation, which led to a reduction in affordability [15]. In fact, interventions of this kind were also the most cost-effective in the most recent analyses of 16 countries [3]. As these cost-effectiveness calculations are based on meta-analyses and models, our study corroborated these underlying models; however, additional studies that are better controlled (e.g. controls established by other countries; for availability, randomization into regions [35] with different availability) will be required for further empirical corroboration for the impact of the WHO’s ‘best buys’ on all-cause mortality. Taxation to increase prices is not only a ‘best buy’ with respect to reducing alcohol-attributable disability life-years [3]; as shown in Lithuania, such alcohol control measures can also have an immediate impact on all-cause mortality and thus on life expectancy. The immediate effect was due to the fact that prices increased on the first day of the tax implementation, impacting consumers’ decisions to buy alcohol [36,37], unless retailers still had a great deal of old stock available at the earlier prices. Overall, however, alcohol is a commodity with a relatively short shelf-life—i.e. the beverages are generally sold out quickly, as illustrated in a 2017 press article summarizing the effect of a major increase in excise tax [38]. When alcohol is no longer available throughout the night, except in bars and/or restaurants, the decision to drink is immediately impacted, and thereby reduces availability.
The third ‘best buy’ (i.e. a ban on marketing and advertising), which was also implemented in Lithuania, did not show an immediate impact, neither in 2008 nor in 2018. The ban in 2008 was only a partial one, restricting advertising via broadcast media during the day; it would not, therefore, have been expected to have an immediate strong effect, particularly given the rapid rise in the use of digital marketing at the time [39], which was not addressed in the 2008 policy. The policy change in 2018 was more complete, and included a ban on digital marketing. However, we would not expect such a ban to have an immediate impact. The postulated pathway is for more long-term effects on the purchasing of alcohol and its level of use [40].
In summary, our study shows sizeable effects for alcohol control policies in general, and for pricing policies in particular, on all-cause mortality. This demonstrates that even general indicators such as all-cause mortality and life expectancy (which is simply a weighted average of all-cause mortality rates) can be impacted with the right public health policies. Given the overall toll on alcohol-attributable mortality [41], and the immense costs resulting from the consumption of alcohol—not only to the health-care system [42] but also more widely in society [43]—such measures should be implemented much more often. Lithuania provides us with a great example of what is possible in a modern western democracy—even when its policy choices are limited by its membership in the European Union—and provides us with a useful model for the implementation of the WHO’s ‘best buys’.
Supplementary Material
Supplementary File S1 data and R code
Figure S1 – Density distribution and QQ-plots of age-standardized all-cause mortality rates.
Figure S2 Characteristics of final models used in Analysis 1 (ACF = autocorrelation function; pACF = partial autocorrelation function).
Figure S3 Monthly GDP per capita between 2001 and 2018 (red vertical lines indicate the implementation of the three alcohol policies of interest for Analysis 2, yellow vertical lines indicate the implementation of the remaining four alcohol policies for Analysis 3).
Figure S4 Monthly unemployment rates between 2001 and 2018 (red vertical lines indicate the implementation of the three alcohol policies of interest for Analysis 2, yellow vertical lines indicate the implementation of the remaining four alcohol policies for Analysis 3).
Figure S5 Proportional change in GDP by year, with October as reference. For Analysis 2 (upper facet), coloured lines indicate years of three alcohol policies of interest (orange: 2008, yellow: 2017, green: 2018) while coloured vertical lines indicate month of implementation (green for 2008 and 2018: January, yellow for 2017: March). For Analysis 3 (lower facet), coloured lines indicate years of four additional alcohol policies of interest (light blue: 2009, dark blue: 2014, light green: 2015, dark green: 2016) while coloured vertical lines indicate month of implementation (dark green for 2009 and 2016: January, light green for 2015: March, dark blue for 2014: April).
Figure S6 Proportional change in unemployment rate by year, with October as reference. For Analysis 2 (left facets), coloured lines indicate years of three alcohol policies of interest (orange: 2008, yellow: 2017, green: 2018) while coloured vertical lines indicate month of implementation (green for 2008 and 2018: January, yellow for 2017: March). For Analysis 3 (right facets), coloured lines indicate years of four additional alcohol policies of interest (light blue: 2009, dark blue: 2014, light green: 2015, dark green: 2016) while coloured vertical lines indicate month of implementation (dark green for 2009 and 2016: January, light green for 2015: March, dark blue for 2014: April).
Figure S7 Characteristics of final models used in Analysis 2 (ACF = autocorrelation function; pACF = partial autocorrelation function).
Figure S8 Characteristics of final models used in Analysis 3 (ACF = autocorrelation function; pACF = partial autocorrelation function).
Table S1 Candidate models for Analysis 2.
Table S2 Candidate models for Analysis 3.
Acknowledgements
Research reported in this publication was also supported by the National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health (NIAAA), grant number 1R01AA028224-01. This research was conducted as part of the project ‘Evaluation of the impact of alcohol control policies on morbidity and mortality in Lithuania and other Baltic states’ and we would like to thank the whole team for their input to wider discussions in generating the research reported in this paper. Content is the responsibility of the authors and does not reflect official positions of NIAAA or the National Institutes of Health.
Footnotes
Pre-registered hypothesis
OSF Registries. Lithuania Mortality analysis: https://osf.io/dhjq3; successful application for grant 1R01AA028224–01 (2019); see also [15].
Declaration of interests
C.F.-B. is a staff member of the WHO Regional Office for Europe, M.N. is a WHO consultant and I.Z. works in the WHO Country Office of Lithuania. The authors alone are responsible for the views expressed in this publication and these do not necessarily represent the decisions or the stated policy of the WHO.
Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
References
- 1.Rehm J, Imtiaz S Alcohol consumption as a risk factor for global burden of disease. A narrative review. Substance abuse treatment. Prev Policy 2016; 11: 37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Global Burden of Disease (GBD) 2019 Risk Factors Collaborators Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the global burden of disease study 2019. Lancet 2020; 396: 1223–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Chisholm D, Moro D, Bertram M, Pretorius C, Gmel G, Shield K, et al. Are the ‘best buys’ for alcohol control still valid? An update on the comparative cost-effectiveness of alcohol control strategies at the global level. J Stud Alcohol Drugs 2018; 79: 514–22. [PubMed] [Google Scholar]
- 4.Wagenaar AC, Tobler AL, Komro KA Effects of alcohol tax and price policies on morbidity and mortality: a systematic review. Am J Public Health 2010; 100: 2270–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Neufeld M, Ferreira-Borges C, Gil A, Manthey J, Rehm J Alcohol policy has saved lives in the Russian Federation. Int J Drug Policy 2020; 80: 102636. [DOI] [PubMed] [Google Scholar]
- 6.World Health Organization Regional Office for Europe. Alcohol Policy Impact Case Study: The Effects of Alcohol Control Measures on Mortality and Life Expectancy in the Russian Federation. 2019. Available at: https://apps.who.int/iris/bitstream/handle/10665/328167/9789289054379-eng.pdf?sequence=1&isAllowed=y (accessed 25 June 2020).
- 7.Nemtsov A, Neufeld M, Rehm J Are trends in alcohol consumption and cause-specific mortality in Russia between 1990 and 2017 the result of alcohol policy measures? J Stud Alcohol Drugs 2019; 80: 489–98. [PubMed] [Google Scholar]
- 8.Leon DA, Chenet L, Shkolnikov VM, Zakharov S, Shapiro J, Rakhmanova G, et al. Huge variation in Russian mortality rates 1984–94: artefact, alcohol, or what? Lancet 1997; 350: 383–8. [DOI] [PubMed] [Google Scholar]
- 9.Neufeld M, Rehm J Alcohol consumption and mortality in Russia since 2000: are there any changes following the alcohol policy changes starting in 2006? Alcohol Alcohol 2013; 48: 222–30. [DOI] [PubMed] [Google Scholar]
- 10.Shield KD, Rylett M, Rehm J Public Health Successes and Missed Opportunities. Trends in Alcohol Consumption and Attributable Mortality in the WHO European Region, 1990–2014. Copenhagen, Denmark: WHO European Region; 2016. [Google Scholar]
- 11.World Health Organization. ‘Best buys’ and other recommended interventions for the prevention and control of noncommunicable diseases. Updated (2017) Appendix 3 of the global action plan for the prevention and control of noncommunicable diseases 2013–2020. 2017. Available at: https://www.who.int/ncds/management/WHO_Appendix_BestBuys.pdf (accessed 19 June 2020).
- 12.Rehm J, Štelemėkas M, Badaras R Research protocol to evaluate the effects of alcohol policy changes in Lithuania. Alcohol Alcohol 2019; 54: 112–8. [DOI] [PubMed] [Google Scholar]
- 13.Miščikienė L, Midttun NG, Galkus L, Belian G, Petkevičienė J, Vaitkevičiūtė J, et al. Review of the Lithuanian alcohol control legislation in 1990–2020. Int J Environ Res Public Health 2020; 17: 3454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Beard E, Marsden J, Brown J, Tombor I, Stapleton J, Michie S, et al. Understanding and using time–series analyses in addiction research. Addiction 2019; 114: 1866–84. [DOI] [PubMed] [Google Scholar]
- 15.Rehm J, Štelemėkas M, Ferreira-Borges C, Jiang H, Lange S, Neufeld M, et al. Classifying alcohol control policies with respect to expected changes in consumption and alcohol-attributable harm: the example of Lithuania. Int J Environ Res Public Health 2021; 18: 2419. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rehm J, Manthey J, Lange S, Badaras R, Zurlyte I, Passmore J, et al. Alcohol control policy and changes in alcohol-related traffic harm. Addiction 2020; 115: 655–65. [DOI] [PubMed] [Google Scholar]
- 17.Puffer R, Griffith G Patterns of Urban Mortality, Scientific Publication no. 151. Washington, DC: Pan American Health Organization; 1967. [Google Scholar]
- 18.Rehm J, Hasan OSM, Imtiaz S, Neufeld M Quantifying the contribution of alcohol to cardiomyopathy: a systematic review. Alcohol 2017; 61: 9–15. [DOI] [PubMed] [Google Scholar]
- 19.Lithuanian Institute of Hygiene. Metodinės rekomendacijos suteiks daugiau žinių, kaip tinkamai užpildyti medicininį mirties liudijimą [The methodological guidelines will provide more knowledge on how to properly complete a medical death certificate]. Vilnius 2019. Available at: https://www.hi.lt/news/1545/1203/Metodines-rekomendacijos-suteiks-daugiau-ziniu-kaip-tinkamai-uzpildyti-medicinini-mirties-liudijima.html (accessed: 27 January 2021).
- 20.Lithuanian Department of Statistics. Official Statistics Portal. GDP per capita, at current prices. Vilnius 2020. Available at: https://osp.stat.gov.lt/statistiniu-rodikliu-analize#/ (accessed: 28 January 2021).
- 21.Lithuanian Department of Statistics. Official Statistics Portal. Unemployment rate in Lithuania. Vilnius 2020. Available at: https://osp.stat.gov.lt/statistiniu-rodikliu-analize#/ (accessed: 28 January 2021).
- 22.R Core Team. R: A Language and Environment for Statistical Computing, 3.5.1 edn. Vienna, Austria: R Foundation for Statistical Computing; 2018. [Google Scholar]
- 23.Statistics Lithuania. Official Statistics Portal. 2020. Available at: www.stat.gov.lt/home (accessed: 28 January 2021).
- 24.De Goeij MCM, Suhrcke M, Toffolutti V, Van de MheenD., Schoenmakers T. M., Kunst A. E. How economic crises affect alcohol consumption and alcohol-related health problems: a realist systematic review. Soc Sci Med 2015; 131: 131–46. [DOI] [PubMed] [Google Scholar]
- 25.Shadish WR, Cook TD, Campbell DT Experimental and Quasi-Experimental Designs for Generalized Causal Inference. New York, NY: Houghton Mifflin Company; 2002. [Google Scholar]
- 26.Rossow I, Mäkelä P Public health thinking around alcohol-related harm: why does per capita consumption matter? J Stud Alcohol Drugs 2021; 82: 9–17. [PubMed] [Google Scholar]
- 27.Norström T, Ramstedt M Mortality and population drinking: a reviewof the literature. Drug Alcohol Rev 2005; 24: 537–47. [DOI] [PubMed] [Google Scholar]
- 28.Lithuanian Institute of Hygiene. The Health Indicators of Lithuania database. 2021. Available at: https://www.hi.lt/en/the-health-indicators-of-lithuania.html (accessed: 23 January 2021).
- 29.National Health Insurance Fund (Lithuania). Lithuanian National Screening Programmes. Vilnius 2021. Available at: http://www.vlk.lt/veikla/veiklos-sritys/prevencines-programos/apie-prevencines-programas/visosltvykdprevprogramos (accessed: 23 January 2021).
- 30.Liutkutė V, Veryga A, Štelemėkas M, Goštautaitė Midttun N Burden of smoking in Lithuania: attributable mortality and years of potential life lost. Eur J Public Health 2017; 27: 736–41. [DOI] [PubMed] [Google Scholar]
- 31.Leinsalu M, Baburin A, Jasilionis D, Krumins J, Martikainen P, Stickley A Macroeconomic fluctuations and educational inequalities in suicide mortality among working-age men in the Baltic countries and Finland in 2000–2015: a register-based study. J Psychiatr Res 2020; 131: 138–43. [DOI] [PubMed] [Google Scholar]
- 32.OECD/European Observatory on Health Systems and Policies. Lithuania: Country Health Profile 2019, State of Health in the EU. Brussels: OECD Publishing, Paris/European Observatory on Health Systems and Policies; 2019. [Google Scholar]
- 33.Stumbrys D, Telksnys T, Jasilionis D, Liutkutė Gumarov V, Galkus L, Goštautaitė Midttun N, et al. Alcohol-related male mortality in the context of changing alcohol control policy in Lithuania 2000–2017. Drug Alcohol Rev 2020; 39: 818–26. [DOI] [PubMed] [Google Scholar]
- 34.Lithuanian Institute of Hygiene. Demographic Statistics 1996–2019. Vilsnius 2021. Available at: https://hi.lt/php/dm1.php?dat_file=dem1.txt (accessed: 16 January 2021).
- 35.Norström T, Skog OJ Saturday opening of alcohol retail shops in Sweden: an experiment in two phases. Addiction 2005; 100: 767–76. [DOI] [PubMed] [Google Scholar]
- 36.Sornpaisarn B, Shield KD, Österberg E, Rehm J Resource Tool on Alcohol Taxation and Pricing Policies. Geneva, Switzerland: World Health Organization and others; 2017. [Google Scholar]
- 37.O’Donnell A, Anderson P, Jané-Llopis E, Manthey J, Kaner E, Rehm J Immediate impact of minimum unit pricing on alcohol purchases in Scotland: controlled interrupted time series analysis for 2015–18. BMJ 2019; 366: l5274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Budreikiene J, Migonyte S Statistics of excise: less booze, higher tax income [Akcizų statistika: svaigalų mažiau, mokesčio surinkimas—didesnis]. Business News [Verslo žinios]. 2020. Issue date 10 October 2017. [Google Scholar]
- 39.Lobstein T, Landon J, Thornton N, Jernigan D The commercial use of digital media to market alcohol products: a narrative review. Addiction 2017; 112: 21–7. [DOI] [PubMed] [Google Scholar]
- 40.Babor TF, Caetano R, Casswell S, Edwards G, Giesbrecht N, Graham K, et al. Alcohol: No Ordinary Commodity. Research and Public Policy, 2nd edn. Oxford: Oxford University Press; 2010. [Google Scholar]
- 41.Shield K, Manthey J, Rylett M, Probst C, Wettlaufer A, Parry CDH, et al. National, regional, and global burdens of disease from 2000 to 2016 attributable to alcohol use: a comparative risk assessment study. Lancet Public Health 2020; 5: e51–e61. [DOI] [PubMed] [Google Scholar]
- 42.Rehm J, Mathers C, Popova S, Thavorncharoensap M, Teerawattananon Y, Patra J Global burden of disease and injury and economic cost attributable to alcohol use and alcohol-use disorders. Lancet 2009; 373: 2223–33. [DOI] [PubMed] [Google Scholar]
- 43.Thavorncharoensap M, Teerawattananon Y, Yothasamut J, Lertpitakpong C, Chaikledkaew U The economic impact of alcohol consumption: a systematic review. Substance abuse treatment, prevention and. Foreign Policy 2009; 4: 20. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary File S1 data and R code
Figure S1 – Density distribution and QQ-plots of age-standardized all-cause mortality rates.
Figure S2 Characteristics of final models used in Analysis 1 (ACF = autocorrelation function; pACF = partial autocorrelation function).
Figure S3 Monthly GDP per capita between 2001 and 2018 (red vertical lines indicate the implementation of the three alcohol policies of interest for Analysis 2, yellow vertical lines indicate the implementation of the remaining four alcohol policies for Analysis 3).
Figure S4 Monthly unemployment rates between 2001 and 2018 (red vertical lines indicate the implementation of the three alcohol policies of interest for Analysis 2, yellow vertical lines indicate the implementation of the remaining four alcohol policies for Analysis 3).
Figure S5 Proportional change in GDP by year, with October as reference. For Analysis 2 (upper facet), coloured lines indicate years of three alcohol policies of interest (orange: 2008, yellow: 2017, green: 2018) while coloured vertical lines indicate month of implementation (green for 2008 and 2018: January, yellow for 2017: March). For Analysis 3 (lower facet), coloured lines indicate years of four additional alcohol policies of interest (light blue: 2009, dark blue: 2014, light green: 2015, dark green: 2016) while coloured vertical lines indicate month of implementation (dark green for 2009 and 2016: January, light green for 2015: March, dark blue for 2014: April).
Figure S6 Proportional change in unemployment rate by year, with October as reference. For Analysis 2 (left facets), coloured lines indicate years of three alcohol policies of interest (orange: 2008, yellow: 2017, green: 2018) while coloured vertical lines indicate month of implementation (green for 2008 and 2018: January, yellow for 2017: March). For Analysis 3 (right facets), coloured lines indicate years of four additional alcohol policies of interest (light blue: 2009, dark blue: 2014, light green: 2015, dark green: 2016) while coloured vertical lines indicate month of implementation (dark green for 2009 and 2016: January, light green for 2015: March, dark blue for 2014: April).
Figure S7 Characteristics of final models used in Analysis 2 (ACF = autocorrelation function; pACF = partial autocorrelation function).
Figure S8 Characteristics of final models used in Analysis 3 (ACF = autocorrelation function; pACF = partial autocorrelation function).
Table S1 Candidate models for Analysis 2.
Table S2 Candidate models for Analysis 3.