SEARCH

SEARCH BY CITATION

Keywords:

  • Alcohol Policy;
  • Alcohol-Related Mortality;
  • Russia;
  • Interrupted Time Series

Abstract

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References

Background

The aim of this study was to determine the impact of a set of 2006 Russian alcohol policies on alcohol-related mortality in the country.

Methods

We used autoregressive integrated moving average interrupted time series techniques to model the impact of the policy on the number of sex-specific monthly deaths of those aged 15+ years due to alcohol poisoning, alcoholic cardiomyopathy, alcoholic liver cirrhosis, and alcohol-related mental and behavioral disorders. The time series began in January 2000 and ended in December 2010. The alcohol policy was implemented in January 2006.

Results

The alcohol policy resulted in a significant gradual and sustained decline in male deaths due to alcohol poisoning (ωo = −92.631, p < 0.008, δ1 = 0.883, p < 0.001) and in significant immediate and sustained declines in male (ω0 = −63.20, p < 0.05) and female (ω0 = −64.28, p < 0.005) deaths due to alcoholic liver cirrhosis.

Conclusions

The 2006 suite of alcohol policies in Russia was responsible for an annual decline of about 6,700 male alcohol poisoning deaths and about 760 male and about 770 female alcoholic liver cirrhosis deaths. Without the alcohol policy, male alcohol poisoning deaths would have been 35% higher and male and female alcoholic liver cirrhosis deaths would have been 9 and 15% higher, respectively. We contextualize our findings in relation to declining mortality in Russia and to results from recent studies of the impact of this law on other causes of death.

In recognition of high levels of alcohol consumption and its associated public health problems, Russia implemented a suite of national alcohol policies in 2006 to address alcohol-related harm, with specific interventions meant to reduce the production and distribution of ethyl alcohol and improve product quality and safety. The policies included strict regulations requiring registering products, tracking imported goods in a database, limiting sales to licensed locations, and monitoring raw materials used in production (Levintova, 2007). The goal of the present study is to determine whether this set of policies had an effect on alcohol-related mortality in Russia.

Russian male and female life expectancies at birth increased 3 and 2.5%, respectively, between 2003 and 2010 (Shkolnikov et al., 2013). Despite this positive growth, the gap in life expectancies between Russia and the United Kingdom (a European Union country with a middle-range life expectancy) greatly increased in the last half-century. Russian life expectancies went from 3 years lower for men and 1.5 years lower for women in 1965 to 15 and 7 years lower in 2010 (Shkolnikov et al., 2013). Alcohol-related mortality is a predominant explanation for high mortality rates in Russia, especially for males. Men and colleagues (2003) found that some of the largest relative changes in Russian mortality trends are directly linked to alcohol, including deaths from liver cirrhosis and unintentional alcohol poisoning. Zaridze and colleagues (2009) drew similar conclusions, finding that the largest relative changes in Russian mortality rates between 1991 and 2006 were related to alcohol intake. The World Health Organization (2004) reported that approximately 47,000 people in Russia died in 2001 from alcohol poisoning after consuming poor-quality products. A recent news release on alcohol and health analyses highlighted the fact that 1 in 5 Russian men die from alcohol-related causes (World Health Organization, 2011a). The burden of harmful alcohol use falls not only on those who die or suffer health problems as a result of misuse, but also on those close to them and society as a whole (World Health Organization, 2011b).

Per Capita Alcohol Consumption and Alcohol-Related Mortality

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References

Research consistently links per capita alcohol consumption with alcohol-related mortality, including deaths from liver cirrhosis and other alcohol-related diseases (see Norström and Ramstedt, 2005). Rehm and colleagues (2004) estimate that the alcohol-related burden of disease is 3.2% of global mortality, and the ratio of males to females dying from the effects of alcohol across all diseases is 10:1. Research on Russia also reveals a greater burden of alcohol consumption on males than females (Leon et al., 2009).

While moderate alcohol consumption can have positive health benefits, heavy consumption can lead to a multitude of health problems (de Lorimier, 2000). In a meta-analysis of 156 studies, Corrao and colleagues (2004) concluded that a J-shaped relationship exists between alcohol consumption and mortality and disease. For instance, they found there was a minimum risk of ischemic heart disease at 20 g/d of alcohol, a protective effect at up to 72 g/d, and an increased risk at 89 g/d. These findings are particularly important in Russia, where not only is the volume of consumption among the highest in the world, but the pattern of drinking is among the most hazardous (Gmel et al., 2001; Rehm et al., 2001). Bobak and colleagues (1999) estimated that approximately one-third of Russian males binge drink at least once per month.

The effect of alcohol consumption on alcohol-related mortality varies by type of alcohol and patterns of drinking. For instance, there is a beverage-specific effect of distilled spirits sales on cirrhosis mortality rates in the United States. Gruenewald and Ponicki (1995) found that a 1% increase in spirits sales was associated with a 0.28% increase in cirrhosis mortality rates, while beer and wine sales were not significant, and they suggested the spirits–mortality association is due to the relationship between chronic heavy drinking and consumption of spirits. Roizen and colleagues (1999) also found an association with cirrhosis mortality rates for spirits, but not for beer and wine in the United States. These findings are suggestive for Russia, where rates of chronic heavy drinking are high and distilled spirits, mainly vodka, are the beverage of preference. Norström (2001) found that in 8 of 14 European countries, there was a positive and significant association between alcohol consumption and all-cause mortality, but the strength of this effect is conditioned by volume of consumption. A 1-l increase in alcohol consumption was associated with a 3% increase in all-cause mortality in low-consumption countries and a 1% increase in medium- and high-consumption countries. In another study, per capita alcohol consumption was significantly related to liver cirrhosis mortality in 8 European countries (Ramstedt, 2007). Specifically, a 1-l increase in alcohol consumption was associated with 3 to 4 additional cirrhosis deaths per 100,000 men and 1 additional death per 100,000 women. In Russia, alcohol consumption has long been linked to alcohol-related mortality. Although historical data are sparse, dating to at least the 1970s the Former Soviet Union and Russia had much higher rates of alcohol poisonings than the United States and other western countries (Treml, 1997). More recently, scholars link changes in Russian mortality rates in general and alcohol-related mortality rates specifically to fluctuations in alcohol consumption (Leon et al., 1997; Nemstov, 2002).

Alcohol Policies and Alcohol-Related Mortality

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References

Many countries have implemented alcohol policies to reduce the public health burden of alcohol-related morbidity and mortality, and research has consistently shown that these policies have beneficial effects. In a study of the effect of alcohol taxes in the United States on cirrhosis mortality rates, Ponicki and Gruenewald (1996) found a significant negative effect for distilled spirits taxes, but not for wine or beer taxes. A $1 increase in distilled spirits taxes resulted in a long-term drop in cirrhosis mortality of 2.6%. State-specific policies also had significant effects. A study of alcohol beverage tax increases from 1976 to 2004 in Alaska found reductions in alcohol-related disease mortality of 29% in 1983 and 11% in 2002 (Wagenaar et al., 2009). A similar study in Florida from 1969 to 2004 found that a 10% increase in alcohol tax was associated with a 2.2% decrease in alcohol-related mortality (Maldonado-Molina and Wagenaar, 2010). Herttua (2010) found that a 2004 reduction in alcohol prices in Finland was associated with a 17% increase in alcohol-related mortality among men aged 40 to 49 (2.5 deaths per 100,000 person-years) and an 11% increase among women aged 40 to 49 (0.3 deaths per 100,000 person-years). For those aged 50 to 69, the effect was smaller for men and higher for women.

Russia also has attempted to alleviate problems associated with alcohol consumption by implementing alcohol-related policies. In the 18 months following the introduction of the 1985 anti-alcohol campaign, government alcohol sales in Moscow decreased by 38% and deaths from liver cirrhosis, alcohol poisoning, and blood alcohol-positive violent deaths decreased 33, 51, and 51%, respectively (Nemstov, 1998). More recently, Russia implemented another comprehensive alcohol policy, the potential effect of which is the focus of this study. Levintova (2007) provides a detailed description of the policy and its creation and implementation. In brief, on July 21, 2005, President Putin signed laws on the regulation of production and sale of ethyl alcohol and alcohol-containing products, effective January 1, 2006. Among other items, the law (i) made it illegal to produce, distribute, and sell alcohol without a license from the government, the cost of which represented a substantial increase over prior licensing costs, (ii) required all production facilities to purchase recording equipment regulated by law to measure the amount of ethyl alcohol used and produced, (iii) required the amount of alcohol and alcohol products used and produced to be provided to a government agency, (iv) required all alcohol products to display an excise stamp stating they are bound for sale in the Russian domestic market, (v) required a license to sell alcohol products, (vi) prohibited alcohol sales at sites like educational and athletic facilities and on public transport, and (vii) prohibited the sale of alcoholic beverages containing more than 15% ethanol (EtOH) alcohol by volume at or near large public places like train and metro stations and wholesale markets and in kiosks and any places not specifically licensed for such sales. This resulted in a decline in alcohol producers and distributors and increased consumer prices (Levintova, 2007).

Shkolnikov and colleagues (2013) recently suggested that the 2006 Russian alcohol policy may be partially responsible for the increase in life expectancy and the decrease in alcohol-related deaths during the last decade. The authors found steep declines from 2004 to 2010 in external causes of death in Russia, although rates remained high. The mortality decline was due mainly to decreases in mortality from alcohol poisoning and other alcohol-related accidents, violence, cardiovascular deaths, and hemorrhagic stroke among those in the working-age group. Similarly, Neufeld and Rehm (2013) found declines in Russia in all-cause mortality and deaths due to cardiovascular disease and external causes that coincided with the 2006 alcohol policy. Finally, employing interrupted time series techniques, Pridemore and colleagues (2013b) found the policy was associated with a decrease of 11% in male traffic fatalities, which represents an annual decline of over 2,400 male deaths, and Pridemore and colleagues (2013a) found that the policy resulted in a 9% reduction in male suicides, which corresponds to an annual decrease of about 3,400 male deaths. No research has yet tested directly Shkolnikov and colleagues' (2013) hypothesis that recent reductions in alcohol-related mortality were due to the implementation of the 2006 Russian alcohol policy. Our study is the first to do so. We take advantage of this natural experiment and the rigorous tests of association provided by interrupted time series techniques (Pierce, 1977) to determine whether the 2006 alcohol policy resulted in a decline in deaths due to alcohol poisonings, alcoholic cardiomyopathy, alcoholic liver cirrhosis, and alcohol-related mental and behavioral disorders.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References

Data

The outcome variables in this study were the monthly number of male- and female-specific deaths of those aged 15+ years due to accidental alcohol poisonings (X45), alcoholic cardiomyopathy (I42.6), alcoholic liver cirrhosis (K70), and alcohol-related mental and behavioral disorders (F10). We used counts instead of standardized death rates (SDR) as it is not possible to apply standardization to these monthly data because the corresponding monthly population size cannot be estimated reliably. There was only a small decline in the total population (from 146 to 143 million) during this period, and more importantly, the population of those most likely affected by the alcohol policy and alcohol-related mortality (i.e., those aged 15+) grew about 2 million (from 119 to 121 million).

Deaths are classified according to the International Classification of Diseases, 10th Revision (2004). The Russian cause-of-death data and the civil death registration system are based on medical death certificates completed by a medical doctor (who treated the deceased and/or established the cause of death), the pathologist who completed the autopsy, or a forensic medical expert. Data were obtained from anonymous death records collected by the Russian Federal State Statistics Service. Each time series began in January 2000 and ended in December 2010. The alcohol policy was implemented in January 2006.

Estimation Procedures

We used autoregressive integrated moving average (ARIMA) interrupted time series techniques to model the impact of the alcohol policy on the alcohol-related causes of death. As ARIMA procedures are well established in the literature on the impact of policy (Chamlin et al., 2008; Loftin et al., 1983; Singer and McDowall, 1988), including alcohol policy (Pridemore and Snowden, 2009), we provide only a brief discussion.

A fundamental concern associated with the evaluation of the efficacy of legislative and administrative initiatives is distinguishing their effects from other social processes that may be influencing an outcome series. ARIMA techniques, unlike simple pre- and postintervention mean or percentage difference tests, explicitly take into account the potentially confounding effects of other causal mechanisms and, as a consequence, allow one to assess the change in the level of any outcome series independent of ongoing stochastic processes (McDowall et al., 1980).

An ARIMA interrupted transfer function model consists of 2 parts. The first, the “noise” component, uses information from prior observations of an outcome series to model the systematic variation (i.e., autocorrelation) within the series. By applying the appropriate seasonal and nonseasonal differencing, along with estimating the appropriate seasonal and nonseasonal autoregressive and moving average parameters (i.e., prewhitening), one can separate the confounding influences of other causal processes from those associated with the intervention.

Once a satisfactory noise component is identified and estimated, the intervention component is added to the transfer function equation. If the inclusion of a dummy series for the intervention (coded 0 for the period prior to the onset of the intervention and 1 beginning with the observation in which the intervention occurs and thereafter) increases the explanatory power of the model above and beyond that provided by the noise component, then we can conclude that the intervention significantly affects the outcome series (Granger, 1980; McDowall et al., 1980).

Another advantage of ARIMA modeling techniques over simple pre- and postintervention change scores is that they allow one to examine the functional form of the relationship between an intervention series and an outcome series. Crude mean and percentage difference tests assume that the effect of an intervention is well represented as an abrupt, permanent change in the level of the outcome series (at least for the remainder of the observations for a given time series). While one can estimate this functional form as a zero-order transfer function using ARIMA modeling techniques, one can also examine the relative fit of competing adjustment models. It is possible that the effect of an intervention gradually reaches a new level (e.g., it takes some time for the intervention to reach its full effect) or that the effect is instantaneous but short-lived. A first-order transfer function can be estimated to model the former pattern of change in the level of a series, while a pulse function can be estimated to model the latter (McDowall et al., 1980).

ARIMA model building is an iterative process. By successively estimating the noise and intervention components, and subjecting them to a number of diagnostic tests, a final intervention model is derived. For the statistical details involved in the identification and estimation of the noise and intervention components of ARIMA interrupted times series models, we refer readers to popular and readily available published sources (McCleary and Hay, 1980; McDowall et al., 1980).

Results

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References

Monthly male and female deaths due to alcohol poisoning, alcoholic cardiomyopathy, alcoholic liver cirrhosis, and mental and behavioral disorders due to alcohol are shown in Figs 1-4, respectively. With the exception of cirrhosis, the number of deaths due directly to alcohol was lower in the period following the implementation of the 2006 alcohol policy. In the case of alcohol poisonings, the reduction was substantial. The male SDR from alcohol poisonings was 42 per 100,000 in 2000, peaked at 50 in 2004, and dropped sharply to 21 in 2010; the female SDR for alcohol poisonings was 11 per 100,000 in 2000, peaked in 2002 at 13, then dropped to 5 in 2010. The male SDR for alcoholic cardiomyopathy was 17 per 100,000 in 2000, peaked in 2005 at 44, and declined to 29 in 2010; the female SDR for alcoholic cardiomyopathy was 5 per 100,000 in 2000, peaked at 13 in 2005, and declined to 8 in 2010. The male SDR for alcoholic liver cirrhosis was 5 per 100,000 in 2000, peaked at 14 in 2005, and declined only to 13 in 2010; the female SDR for alcoholic liver cirrhosis was 2 per 100,000 in 2000, peaked at 7 in 2005, and declined only to 6 in 2010. The male SDR for mental and behavioral disorders due to alcohol was 6 per 100,000 in 2000, peaked at 8 in 2002, and declined to 5 in 2010; the female SDR for mental and behavioral disorders due to alcohol was 1.5 per 100,000 in 2000, peaked at 1.9 in 2003, and declined to 1.5 in 2010. While these trends are suggestive, it is impossible to know whether any declines in alcohol-related mortality were in fact due to the alcohol policy or to ongoing patterns in these time series resulting from other causes. To determine the impact of the alcohol policy on these 4 types of alcohol-related mortality, we must estimate the ARIMA models.

image

Figure 1. Number of monthly male and female deaths due to poisoning by or exposure to alcohol: Russia, 2000–2010. Alcohol law was introduced in January 2006.

Download figure to PowerPoint

image

Figure 2. Number of monthly male and female deaths due to alcoholic cardiomyopathy: Russia, 2000–2010. Alcohol law was introduced in January 2006.

Download figure to PowerPoint

image

Figure 3. Number of monthly male and female deaths due to alcoholic liver cirrhosis: Russia, 2000–2010. Alcohol law was introduced in January 2006.

Download figure to PowerPoint

image

Figure 4. Number of monthly male and female deaths due to mental and behavioral disorders due to alcohol use: Russia, 2000–2010. Alcohol law was introduced in January 2006.

Download figure to PowerPoint

Tables 1 and 2 present the results of model estimation. We tested for 2 main functional forms of the impact of the alcohol policy on alcohol-related mortality. Table 1 shows results for the zero-order transfer function intervention models testing for an immediate and sustained effect. Given the uneven implementation of the policy and the impact this could have with hazardous drinking and thus with alcohol-related mortality, Table 2 shows results for the first-order transfer function intervention models testing for a more gradual effect. The left column in each table provides information about the form and statistical adequacy of the final univariate (i.e., noise) models for all male and female time series. For example, looking at the intervention models for male and female alcoholic liver cirrhosis in Table 1, we see that both series required (i) first-order differencing and first-order seasonal differencing to remove drift and (ii) first-order moving average and first-order seasonal moving average parameters to remove autocorrelation. In fact, the noise models for most of the time series took this same form. The Q-statistic, which is distributed as chi-square, reveals the statistical adequacy of an ARIMA model. The significance levels for the Q-statistics for all 8 noise components are >0.05, indicating that the univariate models effectively account for the systematic variation within each of the outcome series.

Table 1. Final Noise and Zero-Order Transfer Function Intervention Models for the Effect of the 2006 Russian Alcohol Policy on the Number of Male and Female Monthly Deaths Due to Alcohol Poisoning, Alcoholic Cardiomyopathy, Alcoholic Liver Cirrhosis, and Mental and Behavioral Disorders Due to Alcohol
Noise modelIntervention model
  1. ARIMA, autoregressive integrated moving average. Ln = natural log transformation, θ = moving average parameter, B = backward shift operator, ω0 = zero-order input parameter of a transfer function, It = intervention series, Q = Ljung–Box test statistic for the null hypothesis that the model residuals are distributed as white noise.

Male alcohol poisoning deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + ω0It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–72
 It = 1 for observations 73–132
θ1 = −0.519, p < 0.001ω0 = −115.57, p < 0.16
θ12 = −0.416, p < 0.001Q = 22.4, df = 18, p < 0.22
  
Female alcohol poisoning deaths
ARIMA (0,1,1)(0,1,1)12LnYt* = at + ω0It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)LnYtIt = 0 for observations 1–72
 It = 1 for observations 73–132
θ1 = −0.594, p < 0.001ω0 = −0.04, p < 0.32
θ12 = −0.619, p < 0.001Q = 23.7, df = 18, p < 0.17
  
Male alcoholic cardiomyopathy deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + ω0It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–72
 It = 1 for observations 73–132
θ1 = −0.433, p < 0.001ω0 = −6.29, p < 0.48
θ12 = −0.595, p < 0.001Q = 8.4, df = 18, p < 0.98
  
Female alcoholic cardiomyopathy deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + ω0It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–72
 It = 1 for observations 73–132
θ1 = −0.372, p < 0.001ω0 = 50.52, p < 0.18
θ12 = −0.653, p < 0.001Q = 20.6, df = 18, p < 0.31
  
Male alcoholic liver cirrhosis deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + ω0It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–72
 It = 1 for observations 73–132
θ1 = −0.604, p < 0.001ω0 = −63.20, p < 0.05
θ12 = −0.672, p < 0.001Q = 13.1, df = 18, p < 0.79
  
Female alcoholic liver cirrhosis deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + ω0It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–72
 It = 1 for observations 73–132
θ1 = −0.492, p < 0.001ω0 = −64.28, p < 0.005
θ12 = −0.669, p < 0.001Q = 21.0, df = 18, p < 0.28
  
Male deaths from mental and behavorial disorders due to alcohol
ARIMA (0,1,1)(0,1,1)12Yt* = at + ω0It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1−72
 It = 1 for observations 73–132
θ1 = −0.580, p < 0.001ω0 = −23.24, p < 0.20
θ12 = −0.727, p < 0.001Q = 18.8, df = 18, p < 0.41
  
Female deaths due to mental and behavioral disorders due to alcohol
ARIMA1 (0,1,2)(0,1,1)12LnYt* = at + ω0It
(1−θ1B − θ2B2)(1−θ12B12) at = (1−B)(1−B12)LnYtIt = 0 for observations 1–72
 It = 1 for observations 73–132
θ1 = −0.621, p < 0.001ω0 = −0.08, p < 0.17
θ2 = −0.224, p < 0.01Q = 19.4, df = 17, p < 0.31
θ12 = −0.652, p < 0.001 
Table 2. Final Noise and First-Order Transfer Function Intervention Models for the Effect of the 2006 Russian Alcohol Policy on the Number of Male and Female Monthly Deaths Due to Alcohol Poisoning, Alcoholic Cardiomyopathy, Alcoholic Liver Cirrhosis, and Mental and Behavioral Disorders Due to Alcohol
Noise modelIntervention model
  1. ARIMA, autoregressive integrated moving average. Ln = natural log transformation, θ = moving average parameter, B = backward shift operator, ω0 = zero-order input parameter of a transfer function, δ1 = first-order output parameter of a transfer function, It = intervention series, Q = Ljung–Box test statistic for the null hypothesis that the model residuals are distributed as white noise.

Male alcohol poisoning deaths
ARIMA (0,1,[1,4])(0,1,1)12Yt* = at + (ωo/1−δ1B)It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.550, p < 0.001ωo = −92.631, p < 0.008
θ4 = −0.182, p < 0.022δ1 = 0.883, p < 0.001
θ12 = −0.385, p < 0.001Q = 18.1, df = 16, p < 0.32
  
Female alcohol poisoning deaths
ARIMA (0,1,1)(0,1,1)12LnYt* = at + (ωo/1−δ1B)It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)LnYtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.579, p < 0.001ωo = 0.013, p < 0.676
θ12 = −0.619, p < 0.001δ1 = −1.024, p < 0.001
 Q = 24.4, df = 17, p < 0.11
  
Male alcoholic cardiomyopathy deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + (ωo/1−δ1B)It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.252, p < 0.007ωo = 416.311, p < 0.001
θ12 = −0.595, p < 0.001δ1 = −0.644, p < 0.001
 Q = 9.0, df = 17, p < 0.95
  
Female alcoholic cardiomyopathy deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + (ωo/1−δ1B)It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.228, p < 0.014ωo = 209.183, p < 0.001
θ12 = −0.637, p < 0.001δ1 = −0.596, p < 0.001
 Q = 24.7, df = 17, p < 0.11
  
Male alcoholic liver cirrhosis deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + (ωo/1−δ1B)It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.544, p < 0.001ωo = 0.0001, p < 0.965
θ12 = −0.668, p < 0.001δ1 = −1.883, p < 0.013
 Q = 14.3, df = 17, p < 0.65
  
Female alcoholic liver cirrhosis deaths
ARIMA (0,1,1)(0,1,1)12Yt* = at + (ωo/1−δ1B)It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.571, p < 0.001ωo = −29.894, p < 0.003
θ12 = −0.655, p < 0.001δ1 = 0.843, p < 0.001
 Q = 21.0, df = 18, p < 0.28
  
Male deaths from mental and behavorial disorders due to alcohol
ARIMA (0,1,1)(0,1,1)12Yt* = at + (ωo/1−δ1B)It
(1−θ1B)(1−θ12B12) at = (1−B)(1−B12)YtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.540, p < 0.001ωo = 21.672, p < 0.291
θ12 = −0.713, p < 0.001δ1 = −0.949, p < 0.001
 Q = 17.4, df = 17, p < 0.43
  
Female deaths due to mental and behavioral disorders due to alcohol
ARIMA1 (0,1,2)(0,1,1)12LnYt* = at + (ωo/1−δ1B)It
(1−θ1B − θ2B2)(1−θ12B12) at = (1−B)(1−B12)LnYtIt = 0 for observations 1–70
 It = 1 for observations 71–132
θ1 = −0.700, p < 0.001ωo = 26.134, p < 0.064
θ2 = −0.194, p < 0.040δ1 = 0.008, p < 0.988
θ12 = −0.675, p < 0.001Q = 22.3, df = 16, p < 0.14

The right column in each table shows the final transfer function models assessing the impact of the 2006 Russian alcohol policy on monthly male and female deaths due to each of the 4 causes of death. The Q-statistic is the Ljung–Box test statistic for the null hypothesis that the model residuals are distributed as white noise (i.e., they are uncorrelated). As the 2006 alcohol-related initiatives were designed to reduce consumption and alcohol-related harm, p-values are for 1-tailed tests. The results show a mixture of null, zero-order, and first-order effects. For those series in which the first-order transfer function model best fits the data, we calculated the asymptotic change in the postintervention series and the amount of change realized in the first 6 postintervention observations.

For male alcohol poisoning deaths, a first-order gradual change model best fits the data and revealed a significant effect of the alcohol policy (ω0 = −92.631, p < 0.008, δ1 = 0.883, p < 0.001). In December 2005, there were 2,501 of these deaths. The decline due to the policy in the first month was 93 deaths, and by June 2006, the monthly decline reached 416. The asymptotic (i.e., eventual) monthly decline in male deaths from alcohol poisoning due to the policy was 555, which was realized within the first year. For female alcohol poisoning deaths, a zero-order transfer function model better fits the data, but the effect of the alcohol policy on these deaths was nonsignificant (ω0 = −0.04, p < 0.32).

For both male and female deaths due to alcoholic cardiomyopathy, a first-order transfer function best fits the data and revealed significant gradual effects of the policy on these deaths (males: ω0 = 416.311, p < 0.001, δ1 = −0.644, p < 0.001; females: ω0 = 209.183, p < 0.001, δ1 = −0.596, p < 0.001). Unexpectedly, however, the results show a significant increase in alcoholic cardiomyopathy deaths due to the policy. For both men and women, the largest monthly increase was in the month following the implementation of the policy. There was then a slight decline, but the overall increase was sustained for both men and women. This unexpected result is likely due to measurement artifact, which we address in the 'Discussion' below.

For both male and female deaths due to alcoholic liver cirrhosis, a zero-order transfer function best fits the data and showed significant effects. The alcohol policy resulted in an immediate and sustained monthly decrease of 63 deaths due to alcoholic liver cirrhosis for males (ω0 = −63.20, p < 0.05) and 64 deaths due to alcoholic liver cirrhosis for females (ω0 = −64.28, p < 0.005).

For both male and female deaths from mental and behavioral disorders due to alcohol, a zero-order transfer function best fits the data. In neither case, however, was the effect of the policy significant (males: ω0 = −23.24, p < 0.20; females: ω0 = −0.08, p < 0.17).

Discussion

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References

The goal of this study was to determine whether alcohol-related mortality decreased in Russia following the 2006 implementation of a suite of policies aimed at reducing alcohol-related harm. The results of our interrupted time series analyses, which provide conservative tests of association (Pierce, 1977), revealed the new alcohol law was responsible for declines in (i) male deaths due to alcohol poisoning and (ii) male and female deaths due to alcoholic liver cirrhosis. In the former, the effect was more gradual and took several months before the full and sustained effect was realized. In the latter, the effect was immediate and sustained. In both cases, by “sustained,” we mean until the end of the time series employed in our analyses (i.e., December 2010). Given the effect sizes, the 2006 Russian alcohol policy was responsible for an annual decline of about 6,700 male alcohol poisoning deaths, 760 male alcoholic liver cirrhosis deaths, and 770 female alcoholic liver cirrhosis deaths. Without the policy, male alcohol poisoning deaths would have been about 35% higher and male and female alcoholic liver cirrhosis deaths would have been about 9 and 15% higher, respectively. The impact on male alcohol poisoning deaths is notable not only for its size but because the policy included specific interventions aimed at improving product quality and safety.

The 2006 Russian alcohol law was actually a number of separate policies that varied in the success and timeliness of their implementation. While most of the policies were implemented in January, some were delayed or put in place later in the year. For instance, there was a 6-month extension on the requirement of imported alcohol products having excise stamps, and there were problems with the use of the alcohol monitoring system (Levintova, 2007). Further, information about the policies became publically available in 2005, several months before implementation. One key aim of the policy was complete registration and higher taxation of EtOH used to produce alcoholic beverages and EtOH intended for industrial use. Industrial EtOH, however, has been widely used in Russia to produce cheap vodka of poor-quality and nonbeverage products. The latter are nevertheless sometimes consumed, and research shows they contribute substantially to excess mortality due to hazardous drinking. Recognizing the higher forthcoming taxes, alcohol producers produced and stocked large amounts of industrial EtOH and alcohol products from this EtOH for future use. Thus, while the implementation of the alcohol policy at the beginning of 2006 had an effect on alcohol availability and price, the reduction in supply might have been tempered by this stockpile. It is possible that this process accelerated in July 2006 when another element of the alcohol policy—requiring licensing and detailed documentation of EtOH supplies—was introduced.

One way to account for these contingencies is to recognize the effect of the policy on alcohol-related mortality might have been gradual, and we did this by estimating first-order models. Another way to account for this is to recognize that an effect might not have been realized until all elements of the overarching policy had been implemented later in the year. We did this by re-estimating the zero-order models for all time series to account for a second intervention in July 2006 (controlling for the initial intervention in January 2006). These models revealed no further effect of the second intervention point on alcohol poisonings, alcoholic cardiomyopathy, and mental and behavioral disorders due to alcohol. There were additional effects, however, on deaths due to alcoholic liver cirrhosis for males (ω0 = −74.72, p < 0.05) and females (ω0 = −46.12, p < 0.05). These effects are cumulative; that is, they are in addition to those found for the January 2006 policy implementation, meaning that the overall effect was a reduction of 137 deaths/mo due to alcoholic liver cirrhosis for males and 110 deaths/mo for females. Not all of the different measures were implemented precisely on January 1 and July 1, 2006, and so we are unable to gauge precisely their individual effects as they were implemented. Nevertheless, our method does provide a valid assessment of the overall effects of the various policies.

While this is the first study of the impact of the 2006 law on alcohol-related mortality in Russia, 2 recent studies examined the impact of this law on other types of mortality due indirectly to alcohol. Pridemore and colleagues (2013b) found that Russian males experienced an immediate and permanent 11% reduction in traffic fatalities after the implementation of the policy, and Pridemore and colleagues (2013a) found a 9% reduction in male suicides. Taken together, the current study and the previous 2 studies provide evidence of the reduced public health burden of alcohol consumption—as measured by deaths due both directly and indirectly to alcohol—in Russia as a result of the 2006 alcohol policy. Our findings support both Shkolnikov and colleagues' (2013) hypothesis that increased life expectancies at birth in Russia over the last decade may be partially accounted for by this 2006 alcohol policy, and Neufeld and Rehm's (2013) hypothesis that the alcohol policy is responsible for declines in alcohol consumption and alcohol-related mortality in Russia.

As alcoholic liver cirrhosis is a chronic illness that takes years to develop, it may seem counterintuitive that an alcohol policy would have an immediate effect. This is not an uncommon finding, however. For example, in the 18 months following the 1985 anti-alcohol campaign in Russia, there was a 33% decline in liver cirrhosis mortality (Nemstov, 1998). Further, it is necessary to understand that in 2006, unlike in Russia during the anti-alcohol campaign in 1985, there were no alcohol shortages, but instead prices increased and EtOH-based liquids not intended for consumption became less available. The latter circumstance might have resulted in hazardous drinkers suffering fewer alcohol-related diseases or at least some extension of their lives. Alcohol researchers are also familiar with the Paris curve (see Ledermann, 1964; Skog, 1984), which revealed drastic reductions in liver cirrhosis mortality in wartime Paris following sharp decreases in wine consumption due to rationing. A similar impact on cirrhosis deaths, but with longer lags, has been found when examining the impact of alcohol taxes in the United States (Ponicki and Gruenewald, 1996) and Sweden (Andreasson et al., 2006). A reduction in consumption of distilled spirits is especially important for cirrhosis deaths (Ponicki and Gruenewald, 1996; Roizen et al., 1999), as vodka is the beverage preference in Russia and as hazardous consumption is common in the nation (Bobak et al., 1999; Gmel et al., 2001; Rehm et al., 2001).

There are 3 final items to consider. First, there was a stronger proportional impact of the alcohol policy on female alcoholic liver cirrhosis deaths relative to males. There are social, cultural, and economic features of post-Soviet Russia that may help explain this association, and the finding deserves further scrutiny. As just 1 example, males may simply be more resistant to policies aimed at reducing hazardous alcohol consumption. Second, our findings suggested the 2006 Russian alcohol policy appeared to be responsible for an increase in deaths due to alcoholic cardiomyopathy for both men and women. This is likely an artifact of measurement. In Russia, the diagnoses of cardiomyopathy (I42 in the ICD-10) and alcoholic cardiomyopathy (I42.6) only began to be regularly used in 1999. Thus, the rapid growth in alcoholic cardiomyopathy deaths shown in Fig. 2 in the first half of the decade is likely a partial consequence of the increasing popularity of this diagnosis. This also means that the monthly death counts were substantially under-enumerated and artificially low early in the decade, which would bias before and after comparisons of the alcohol policy. Third, while we did find an association for male and female alcoholic liver cirrhosis and for male alcohol poisoning, the 2006 alcohol policy did not have an impact on other types of death due directly to alcohol. This suggests the alcohol policy played a limited role in the decline in mortality and rise in life expectancy in Russia since 2003 highlighted by Shkolnikov and colleagues (2013), and other potential causes for this reversal must be considered.

The primary goal of the 2006 alcohol policy in Russia was to decrease alcohol consumption and alcohol-related harm in the country. Our study is the first to systematically evaluate the effects of this law on mortality due directly to alcohol. Our initial findings suggest the policy was responsible for an annual decline of about 6,700 male deaths due to alcohol poisoning and 1,500 deaths due to alcoholic liver cirrhosis total between men and women. If we include the subsequent effects as the full policy was implemented throughout the year, the annual reduction in deaths due to alcoholic liver cirrhosis doubles to nearly 3,000. This is a positive sign, especially when these numbers are added to the reduction in male traffic fatalities (Pridemore et al., 2013b) and male suicides (Pridemore et al., 2013a), and government and health officials should be recognized for their efforts in this area. Nevertheless, the public health burden of hazardous alcohol consumption remains extremely high in Russia, and efforts to curb this burden must continue.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References

We thank Svetlana Nikitina of the Russian Federation Federal State Statistics Service for her help with organizing the special data tabulation required for this study.

References

  1. Top of page
  2. Abstract
  3. Per Capita Alcohol Consumption and Alcohol-Related Mortality
  4. Alcohol Policies and Alcohol-Related Mortality
  5. Materials and Methods
  6. Results
  7. Discussion
  8. Acknowledgment
  9. References