The UK has among the highest per capita alcohol consumption of any country in the world [1,2], with 9.1 million adults drinking at levels above recommended limits [3,4]. The costs to society in terms of the health, social and criminal implications of excessive alcohol consumption are estimated to be more than £21 billion per annum [5]. In 2010, 3.5% of all deaths in England were wholly attributable to excessive alcohol consumption [6,7]. A further 1.1% of deaths were partially attributable [6].

The rates of harmful drinking in a given country are a function of many factors, including financial cost, cultural norms [8-11] and beliefs about alcohol-related harm [3]. In 2012 the English Government’s alcohol strategy proposed a range of policies to tackle alcohol-related harm: 1) helping individuals to change their drinking behaviour; 2) taking action locally; 3) improving treatment for alcohol dependence; 4) sharing responsibility with the alcohol industry; 5) imposing a minimum unit price; and 6) extending restrictions on advertisement to teenagers and children [5]. Since then the Government has withdrawn some policies (e.g., minimum unit pricing) but has gone ahead with some others: a ban on sales of alcohol ‘below cost’; both reductions and increases in alcohol duties; introducing screening and brief intervention for risk drinking as part of NHS Health Checks; and voluntary agreements with industry to reduce availability of high-strength canned beverages, lower the strength of existing beverages and promote low strength alternatives, and increase the number of product labels providing alcohol content information [12-14]. Alcohol charities have also promoted ‘Dry January’ ( to encourage drinkers to abstain for one month, while more than 70 local authorities have voluntary agreements with retailers to withdraw high-strength, low-cost beers and ciders from sale [15].

Timely and detailed surveillance data could help to inform and evaluate national and local alcohol policies. Several large-scale surveys collect data on alcohol use in the United Kingdom on an annual or less frequent basis [16] (e.g. Health Survey for England, General Lifestyle Survey, and Adult Psychiatric Morbidity Survey; See Table 1).

Table 1 Topics addressed by surveys in England regarding alcohol use

The ATS will fill an important gap by gathering and publishing monthly data on representative samples, including detailed information on alcohol use, attempts to reduce or cease alcohol consumption, and exposure to health professional advice on alcohol. It includes a 6-month telephone follow-up of each monthly sample which will provide data on within-individual trends and consistency in alcohol-related measures.

The ATS is modelled on, and involves the same respondents as the Smoking Toolkit Study (STS) [26], which was set up in 2006 and has already collected data on approximately 175,000 individuals. The STS has demonstrated the value of having monthly figures on key performance indicators and that it has the granularity to detect temporal trends that would otherwise be missed [27-31]. The linkage of these two surveys will provide a unique opportunity to compare trends in smoking and alcohol use behaviour at an individual, regional and national level, and to examine the interdependencies between these two behaviours.


The ATS aims to provide timely tracking of alcohol-related behaviours on a monthly basis to inform and evaluate national alcohol control policies in England. It will also permit analysis of trends as a function of major socio-demographic variables. Monthly, quarterly and annual trends will be published (overall and stratified by age group, social grade, region, gender and smoking status) on:

  1. 1.

    Prevalence of hazardous drinking (Measured by the Alcohol Use Disorders Identification Test (AUDIT))

  2. 2.

    Prevalence of hazardous drinkers who report attempting to reduce their alcohol consumption

  3. 3.

    Prevalence of different methods used by hazardous drinkers attempting to reduce their consumption

  4. 4.

    Prevalence among hazardous drinkers of receipt of advice to reduce alcohol consumption from a health professional in the past year

Changes in these variables associated with events, such as the introduction of pricing policies, will also be assessed.



The ATS involves monthly cross-sectional household computer-assisted interviews, conducted by Ipsos Mori, of approximately 1,800 adults aged 16+ and over in England. All participants who agree to be re-contacted are followed-up at 6 months by a telephone survey. Baseline data were first collected in March 2014, to be followed up in September 2014.

The baseline survey uses a type of random location sampling, which is a hybrid between random probability and simple quota sampling. England is first split into 171,356 ‘Output Areas’, comprising of approximately 300 households. These areas are then stratified according to ACORN characteristics and geographic region. ACORN is a socio-economic profiling tool developed by CACI (, and works by segmenting UK postcodes into 5 categorises (wealthy achievers, urban prosperity, comfortably off, moderate means and hard-pressed). These categorises are further divided into 17 groups and 56 types using government and consumer research data (e.g., census data and lifestyle survey records).

The areas are then randomly allocated to interviewers, who travel to their selected areas and conduct the electronic interviews with one member of the household. Interviews are conducted until quotas based upon factors influencing the probability of being at home (i.e. working status, age and gender) are fulfilled. Morning interviews are avoided to maximise participant availability.

This method of sampling is often seen as superior to conventional quota sampling because the choice of properties approached is significantly reduced by randomly allocating small output areas to the interviewers. However, as interviewers still choose the houses within these particular areas, a response rate cannot be calculated. This is because there is no definite gross sample, with units fulfilling the criteria of the quota being interchangeable.

Ethical approval for the STS was granted by the UCL Ethics Committee (ID 0498/001). Further ethical approval for the ATS was not deemed necessary by the Committee, since asking individuals about their alcohol consumption in addition to smoking does not place them at additional risk of harm.


The study is initially funded for a three year period between March 2014 and March 2017. It is anticipated that data on around 21,600 individuals will be collected each year, with a proposed 3 year sample size of 64,800. However, as with the STS, which has been collecting data for eight years, it is envisaged that the ATS will be extended beyond this initial period.


Questions were developed by an expert panel and policy-makers. The key domains addressed at baseline are: 1) prevalence and frequency of harmful alcohol consumption (AUDIT) [32]; 2) current attempts and motivation to reduce alcohol consumption below harmful levels; 3) health-care professional advice about alcohol consumption; 4) types of drinks consumed and amount spent; 4) urges to drink; 5) recent serious attempts to cut down; and 7) help sought and factors contributing to recent attempts to reduce intake. Additional questions, as with the STS, can be added if new hypotheses are generated. The STS and ATS were set up as ‘toolkits’ for researchers in alcohol and tobacco control, affording the ability to add specific questions (e.g. on mental health). Thus researchers would be provided with all the contextual measurements listed below, negating the requirement for multiple surveys.

Baseline measures

Socio-demographic characteristics

Data are collected on gender, ethnicity, socio-economic status, marital status, number of residents and children in the household, sexuality, age, disability, religion, internet access and use, and government office region (London, South East, South West, East Anglia, East Midlands, West Midlands, Yorkshire/Humberside, North West and North East). Socio-economic status is assessed through car and home ownership, employment status, educational achievements, income and by social-grade. Social grade is measured using the National Readership Survey social-grades system: A: higher managerial, administrative or professional; B: intermediate managerial, administrative or professional; C1: supervisory or clerical and junior managerial administrative or professional; C2: skilled manual workers; D: Semi and unskilled manual workers; and E: Causal or lowest grade workers, pensioners and others who depend on the welfare state for their income [33].

Smoking-related questions

All participants taking part in the ATS are also asked questions regarding their use of tobacco products and smoking behaviour. Full details of these questions and details of the methodology are available from and Fidler et al. [26].

Prevalence and frequency of hazardous drinking

The AUDIT (See Table 2: Question 1: a-k) is used to screen for alcohol use [32]. The AUDIT identifies people who could be classed as dependent, harmful or hazardous drinkers; and has proven validity, high internal consistency and good test-retest reliability across gender, age and cultures [34-38]. The full AUDIT consists of 10 questions: questions 1–3 deal with alcohol consumption, 4–6 with alcohol dependence and 7–10 with alcohol-related problems. A score of 8 or more in men (sometimes 7 in women) indicates a hazardous or harmful alcohol consumption, whilst a score greater than 20 is suggestive of alcohol dependence. The ATS uses an extended item version of the AUDIT questionnaire adopted in previous ‘Screening and Brief Alcohol Intervention in Primary Care (SIPS)’ trials; to allow exploration of higher levels of alcohol consumption in addition to the standard AUDIT scoring system [39,40]. The 2009 Adult Psychiatric Morbidity Survey reported that 24.2% of participants scored 8 or above on the AUDIT, which included 3.8% of adults whose drinking could be classified as harmful. Among males, the highest prevalence of hazardous and harmful drinking was in the 25–34 age group; whereas in females it was in those aged 16–24 [41].

Table 2 Main baseline questionnaire for the ATS

Those who score 8 or more (i.e. indicating hazardous and or harmful alcohol consumption and possible dependence) on the AUDIT or 5 or more on the AUDIT-C (i.e. indicating high-risk consumption) at baseline, are then asked the following questions.

Current attempts and motivation to reduce alcohol consumption

One of the aims of the Department of Health’s ‘Reducing Harmful Drinking’ policy, is to encourage adults to reduce their alcohol intake to recommended levels [3]. Thus participants are asked if they are currently attempting to reduce their alcohol intake (see Table 2: Question 2) and whether they are motivated to do so (see Table 2: Question 5). Question 5 was derived from the Motivation To Stop Scale (MTSS) used in the STS, which has been shown to be a reliable predictor of attempts to quit smoking [42]. The reason for using this scale is twofold: first, insofar that it proves valid and reliable, it is useful to have a single-item measure for motivation and, secondly, the scale will allow a meaningful comparison between the relative levels of motivation for changing the two behaviours.

Receipt of health professional advice on drinking

Participants are asked if they received any advice about their alcohol consumption from their GP and the form of this advice i.e. whether they were given help within the surgery or referred to a specialist service (See Table 2: Question 3). It is recommended by the National Institute for Health and Care Excellence (NICE) that GPs provide adults with brief advice on their alcohol consumption, and offer the provision of self-help materials or a referral to a specialist clinic [43]. GP brief advice also forms part of the Department of Health’s ‘Reducing Harmful Drinking’ policy [3]. Previous studies have shown that GP advice is successful in reducing alcohol intake and encouraging treatment for alcohol dependence [44].

Types of drinks consumed

In the 2009 Omnibus Survey, ‘Drinking: Adults’ behaviour and knowledge’, women were proportionately less likely to drink beer and more likely to drink wine, fortified wine, spirits and alcopops than men. There were also significant age differences, with spirits being the most popular drink among women aged 16 to 24; in comparison with wine among women aged 45–56 years of age. To determine whether relationships exist between these and other socio-demographic variables, the ATS also assesses the type of drinks consumed by participants (See Table 2: Question 4). This is measured using a similar categorisation system to the ONS Omnibus Survey, that is, which types of alcoholic drink the respondents consume most often [45].

Amount spent

Previous surveys have estimated that UK households spend around £15 billion a year on alcohol; although this is likely to be an underestimation given that the Government received £18.2 billion in alcohol duty and tax in 2013/2014 [46]. Thus expenditure on alcohol accounts for around 18% of total expenditure on all food and drink [47]. Households with individuals aged 60–64 spend more on alcohol, tobacco and narcotics, than those with individuals under 30 (£16.40 each week versus to £10.50) [48]. It is important to track expenditure over time and the association with socio-demographic characteristics to help inform policy, but also assess the impact of alcohol-related interventions. Thus, the ATS also assesses the amount spent on alcohol per week (See Table 2: Question 6).

Urges to drink

The urges to drink question (See Table 2: Question 7) was derived from the urges to smoke question used in the STS. This has been shown to be a valid measure of the severity of cigarette dependence in terms of predicting relapse following a quit attempt [49]. Several studies using a variety of methods have also shown that urges and cravings to drink predict relapse following treatment for alcohol dependency [50]. Although items exist for assessing alcohol dependence, this 1-item measure will allow comparisons with tobacco dependence.

Serious attempts to reduce intake

How many serious attempts participants have made to cut down is an important measure of response to interventions (e.g. [51]) (See Table 2: Question 8). Number of attempts to reduce alcohol consumption was included as the majority of individuals will most likely opt for safer levels of drinking as opposed to complete abstinence [52].

Help sought and motives for recent attempts to reduce intake

Questions relating to the most recent attempt to cut down on drinking include: 1) assessing what help was sought (e.g. medication, group counselling or helpline); and 2) reasons for trying to cut down (e.g. advice from doctors, an advert or health problems) (See Table 2: Questions 9 and 10).

The most recent English policy on alcohol use stated that it intended to improve treatment of alcohol dependence [5]. Current treatments recommended by NICE include medication (e.g. acamprosate, disulfiram and naltrexone), which help to prevent relapse and/or to limit the amount of alcohol consumed; and counselling in the form of self-help groups, 12-step facilitation therapy, cognitive behavioural therapy and family therapy [53]. Evidence suggests that drugs such as acamprosate significantly reduce the risk of any drinking and increase the rate of abstinence; while there appears to be some evidence for the effectiveness of psychosocial interventions, including cognitive-behavioural coping skills training and motivational interviewing [54,55].

Six month follow-up

At 6-months follow-up participants who score 8 or more (i.e. indicating hazardous or harmful alcohol consumption and possible dependence) on the AUDIT or 5 or more on the AUDIT-C (i.e. indicating high-risk consumption) at baseline are asked to complete: 1) the AUDIT and questions regarding their current attempts to reduce intake and motivation to cut down; 2) if they received GP advice on alcohol consumption; 3) the types of alcohol typically consumed; 4) the amount spent on alcohol; 5) strength of urges to drink; 6) how many attempts to cut down they have made in the previous 6 months; 7) how long ago their most recent serious quit attempt started and how long it lasted before they went back to drinking as before or more heavily; 8) what they used to help them cut down and what contributed to their most recent attempt; and 9) whether their attempt was planned or unplanned. A key motivation for the follow-up is to assess prospective associations between changes in alcohol consumption and exposure to real-world events and treatments [28,56]. The reason for limiting the follow-up to those who are at least classified as hazardous/harmful or high-risk drinkers at baseline is that these types of analyses among abstinent and moderate drinkers are less relevant.

Data analysis and dissemination

Descriptive statistics will be used to describe the basic features of the data, including the socio-demographic profile of participants and occurrence of alcohol related behaviours. These will include parameter estimates (e.g. means and percentages), number of participants, and measures of spread i.e. confidence intervals and standard deviations. When reporting prevalence data, data will first be weighted using a rim (marginal) weighting technique. This involves an iterative sequence of weighting adjustments whereby separate nationally representative target profiles are set (for gender, working status, children in the household, age, social-grade and region). This process is then repeated until all variables match the specified targets.

To assess differences among groups where the outcome is normally distributed, t-tests for two groups (e.g., men and women) and ANOVA (or related methods e.g. ANCOVA) for comparisons of more than two groups (e.g. social-grades AB, C1, C2, D and E), will be used. If parametric assumptions are violated, transformations will be performed or appropriate non-parametric tests employed (e.g., Kruskal Wallis and Mann–Whitney U). For dichotomous outcomes, log-linear regression or chi-squared tests will be implemented. When adjustment for confounding variables is required, or to examine the association between a quantitative independent variable and quantitative outcome, multivariable linear regression will be used. For the association between continuous predictors and dichotomous outcomes, Generalised Linear Models will be adopted, specifically the binomial family for odds ratios or log-binomial for relative risks.

To assess trends and changes over time as a function of policies and population-based interventions, interrupted time analysis will be used as this allows adjustment for autocorrelation and consideration of underlying trends in the time series [57]. Mediation analyses, segmented regression and non-linear regression analyses, will also be used where appropriate (e.g., segmented regression when assumptions of time-series are violated and there is no evidence of autocorrelation). Mediation analysis allows researchers to assess which factors ‘mediate’ between independent and dependent variables [58]; segmented (or piecewise) regression analysis allows the independent variable to be partitioned into separate intervals when its relationship with the dependent variable changes at a ‘break point’ [59]; while non-linear (polynomial) regression can be used when relationships represent a curved line [60].

Bayes Factors will also be calculated where appropriate. These indicate the relative likelihood of a hypothesised difference/association (hypothesis 1) versus no difference/association (hypothesis 2); and allow one to distinguish between two interpretations of a null result: there is evidence for the null-hypothesis or the data are insensitive in distinguishing an effect. The latter can be rejected if the study is adequately powered, however, there are problems with this in practice. First, calculating power requires the specification of a minimally interesting value. This is often difficult, particularly when similar studies have not been conducted. Secondly, power does not use the data itself in order to determine how sensitively that very data can distinguish the null and alternative hypotheses [61].

The Bayes factor is given by:

$$ BF=\frac{P\left( Data\Big|{H}_1\right)}{P\left( Data\Big|{H}_2\right)} $$

which is the probability of the data given hypothesis 1 (H1) over the probability of the data given hypothesis 2 (H2); and thus, is simply a ratio of the likelihoods of the two hypotheses. Bayes factors vary from 0 to ∞, where 1 indicates that the data do not favour either hypothesis; values greater than 1 indicate increasing evidence for the alternative hypothesis over the null; and values less than 1 indicate increasing evidence of the null over the alternative. Jeffreys [62] suggested that ‘substantial evidence’ be reflected by a factor less than 1/3 or greater than 3, with any value between these being only ‘anecdotal’ evidence.

Results will be disseminated using a website ( and regular updates to key English stakeholders, including Public Health England (PHE) and the Department of Health policy and communications teams. All publications will be discussed in advance with the ATS Study Advisory Group. This group was appointed to oversee and input strategically on the overall progress, priorities and planned analyses of the ATS.


The ATS has several important strengths, including the ability to examine changes in the prevalence of harmful drinking and other key performance indicators, such as attempts and motivation to cut down, in a timely manner. Tracking monthly changes permits a much more sensitive test of the possible effects of interventions than can be achieved by annual national surveys. It will also provide information on methods of reduction and how these relate to success rates and contextual variables, such as socio-demographic characteristics. The large sample size and follow-up will also allow for relationships between government policies/alcohol initiatives and reductions in consumption to be accurately estimated and tested prospectively. The data will provide quick and direct estimates of policy impacts on consumption and also inform the estimation of longer-term health benefits and cost saving by providing evidence to incorporate into existing policy appraisal models [63]. The addition of the ATS to the STS will also allow comparisons to be made between tobacco and alcohol use.

Although the ATS is restricted to data from England and cannot document the situation for the rest of the UK or other countries; it provides a framework on which other national surveys can be modelled. Moreover, given that alcohol policies and treatment in England are similar to other countries, findings from the ATS, to some extent, can be used to inform international policy [64].

The main limitation, as with all survey data, is the likelihood of underreporting alcohol intake. This is often due to a combination of poor recall, participants being untruthful, inadequacies in measurement instruments and sampling bias [65,66]. The extent of underreporting is somewhat mitigated by the use of computer assisted interviews since the presence of a computer enhances participants’ perceptions of privacy and thus increases responses to sensitive questions [67]. Recall bias is also limited in the ATS by assessing typical alcohol consumption; and including response categories (which act as triggers). The ATS also affords the ability to cross-validate responses due to the inclusion of questions measuring similar underlying constructs. For example, answers to the AUDIT are likely to be analogous to measures of urges to drink, while motivation to quit is likely to be analogous to attempts to reduce intake.