Interventions to prevent alcohol use: systematic review of economic evaluations

Abstract

Background

Alcohol use is a leading risk factor for death and disability worldwide.

Aims

We conducted a systematic review on the cost-effectiveness evidence for interventions to prevent alcohol use across the lifespan.

Method

Electronic databases (EMBASE, Medline, PsycINFO, CINAHL and EconLit) were searched for full economic evaluations and return-on-investment studies of alcohol prevention interventions published up to May 2021. The methods and results of included studies were evaluated with narrative synthesis, and study quality was assessed by the Drummond ten-point checklist.

Results

A total of 69 studies met the inclusion criteria for a full economic evaluation or return-on-investment study. Most studies targeted adults or a combination of age groups, seven studies comprised children/adolescents and one involved older adults. Half of the studies found that alcohol prevention interventions are cost-saving (i.e. more effective and less costly than the comparator). This was especially true for universal prevention interventions designed to restrict exposure to alcohol through taxation or advertising bans; and selective/indicated prevention interventions, which involve screening with or without brief intervention for at-risk adults. School-based interventions combined with parent/carer interventions were cost-effective in preventing alcohol use among those aged under 18 years. No interventions were cost-effective for preventing alcohol use in older adults.

Conclusions

Alcohol prevention interventions show promising evidence of cost-effectiveness. Further economic analyses are needed to facilitate policy-making in low- and middle-income countries, and among child, adolescent and older adult populations.

Alcohol use is a leading risk factor for death and disability worldwide, especially in young adults.¹ The Global Burden of Disease study found that alcohol use is associated with substantial health loss, particularly in males.¹ Importantly, the attributable burden of alcohol use increases monotonically with increasing alcohol consumption. Addressing alcohol-related harms is therefore a global public health priority.² There are a variety of interventions designed to prevent alcohol use at the population level (i.e. upstream interventions, such as tax increases or advertising bans) and the individual level (i.e. downstream interventions, such as school-based interventions). To facilitate successful and sustainable scale-up of effective interventions and innovative service delivery strategies, decision makers require evidence on an intervention's cost and cost-effectiveness in addition to its effect on alcohol use and associated harms. Evaluating costs alongside the health effects of alcohol prevention and control strategies is required to determine their value-for-money credentials.

The burden of alcohol use disorders is exacerbated by its comorbidity with other substance use and mental health disorders. For example, a third of adults with opioid use disorder have an alcohol use disorder.³ Depression and anxiety are also most commonly associated with alcohol,⁴ with a third of people living in the UK reporting having both a psychiatric disorder and a comorbid alcohol use disorder.^4,5

A previous review has identified 27 studies published between 2006 and 2016 that have examined economic evaluations of alcohol prevention interventions.⁶ Over half of the studies adopted a healthcare perspective, evaluating interventions over a 5-year time horizon. Most studies analysed healthcare costs, as well as costs attributable to government, social care, criminal justice, law enforcement and individual out-of-pocket payments. The studies evaluated a range of interventions, with the most common interventions comprising screening and brief interventions (SBIs), followed by upstream interventions such as tax increases, advertising restrictions and limiting retail sales. Only two school-based interventions were identified. However, this review primarily focused on economic evaluations of public health interventions and identifying methodological issues, rather than interpreting the cost-effectiveness results of broad preventive interventions for alcohol use in decision-making contexts. The evidence of economic benefit has grown rapidly since the previous review, necessitating an update. Importantly, there is also increasing evidence of economic evaluations targeting multiple health-related risk factors, including alcohol. Evidence on multifactorial prevention interventions were not included in the previous review.

Study aims

This study aims to conduct a systematic review of the evidence for the cost-effectiveness of interventions to prevent alcohol use across the lifespan. This review used narrative synthesis to evaluate the methods of published economic evaluations and the quality of the literature. A key focus of this review was to summarise the cost-effectiveness evidence for alcohol prevention interventions and to identify knowledge gaps, challenges and opportunities for future research. Alcohol use often co-occurs with other substance use and mental/physical health conditions. As such, this review also evaluated studies assessing the cost-effectiveness of preventive interventions targeting multiple health-related risk factors alongside alcohol use.

Method

Search strategy

The current review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines⁷ and was registered on the International Prospective Register of Systematic Reviews Databases (PROSPERO; identifier CRD42020147386). The protocol was amended to include additional researchers and selecting on of the Drummond checklist as the tool for quality assessment. Searches were done to identify journal articles through electronic databases hosted on the EBSCOhost platform (i.e. EMBASE, Medline, CINAHL, PsycINFO and EconLit libraries) on 1 August 2019; with an updated done on 5 May 2021. The search strategy included economic evaluation terms; prevention or treatment terms; and terms related to alcohol, smoking, illicit drug use and substance use disorders. No date restrictions were applied during literature retrieval. Grey literature were excluded to narrow the focus on rigorous, peer-reviewed evidence. Unlike pharmaceutical products, many mental health prevention and treatment interventions are not subject to formal health technology assessment requirements. Manual searches were also conducted with the Tufts Cost-Effectiveness Analysis registry, a comprehensive database containing over 10 000 cost-effectiveness studies.⁸ Further details of the search strategy are presented in the Supplementary Material available at https://doi.org/10.1192/bjo.2023.81.

Study selection

All citations were imported into a web-based systematic review software, Covidence (Veritas Health Innovation, Melbourne, Australia; www.covidence.org), which facilitated the identification and removal of duplicates. Title and abstract screening, full-text screening, data extraction and quality assessment were done independently by any two reviewers (J.F., L.K.-D.L., M.L.C., J.K.P., O.C., H.N.Q.T., M. Sultana, N.H.). Disagreements and discrepancies were resolved by a third reviewer (L.K.-D.L., M.L.C.). Studies were only included if they were full economic evaluations that compared two or more interventions in terms of their costs and outcomes.

Different economic evaluation frameworks can be used to assess the cost-effectiveness of healthcare interventions and programmes. Three commonly used frameworks include cost-effectiveness analysis (CEA), cost–utility analysis (CUA) and cost–benefit analysis (CBA). All of these frameworks measure costs in monetary terms, but differ in how outcomes are measured. For instance, outcomes are measured in CEA by using clinically meaningful units, e.g. the proportion who use alcohol, point improvements on a scale of alcohol-associated harms. The main units of outcome in CUA are generic health indices that combine measures of health-related quality of life (morbidity) and the length of life (mortality). Quality-adjusted life-years (QALYs) and disability-adjusted life-years (DALYs) are both commonly used generic health indices. In CBA, the most widely used framework beyond the health sector, all outcomes are valued in monetary terms. It follows that CBA necessitates the monetary valuation of health-related outcomes. Return-on-investment (ROI) analysis is also a commonly used partial economic evaluation framework that was included in this review. ROIs are typically a reduced form of CBA, where only the costs and cost offsets that can be attributed to healthcare interventions or programmes are considered compared with CBA, which often evaluates a wider set of health and non-health outcomes.

Economic evaluation studies can take the form of trial-based economic evaluations where the economic evaluation is conducted alongside a clinical trial. Alternatively, model-based economic evaluations synthesise multiple data sources to simulate the costs and outcomes that would occur under a scenario where an intervention is implemented versus some counterfactual scenario. All four economic evaluation frameworks (CEA, CUA, CBA and ROI) were included in this systematic review. Partial economic evaluations, cost studies, reviews, expert opinions, qualitative studies, conference papers, dissertations, book chapters and articles not in English were excluded. Studies were classified as alcohol prevention if they evaluated interventions focused on the prevention of alcohol use or the reduction of excessive alcohol use. Studies that targeted either the general population or the at-risk drinking population were included.

The mental health intervention spectrum described by Mrazek and Haggerty was used to classify prevention interventions into three types: universal, selective and indicated prevention.⁹ Universal prevention interventions target the whole population (e.g. school-based prevention). Selective prevention interventions target a subgroup of the population who are at risk for harmful alcohol use and/or binge drinking. Indicated prevention interventions target people who binge drink and/or consume harmful levels of alcohol, but do not have an alcohol use disorder or alcohol dependence. Studies were excluded if they included treatment interventions that target people diagnosed with an alcohol use disorder or alcohol dependence.

In summary, study inclusion criteria were full economic evaluations (e.g. CEA, CBA and CUA) or ROI studies aimed at prevention of alcohol use or reduction of excessive alcohol use. Exclusion criteria were partial economic evaluations and cost studies, reviews, expert opinions, qualitative studies, conference papers, dissertations, book chapters or artiwcles not in English.

Data extraction

This study used a data extraction framework that was adapted from several previous reviews of economic evaluations and the review guideline developed by the Joanna Briggs Institute (JBI).^10,11 Data extraction was completed in Microsoft Excel version 15.0 for Windows and independently performed by any two reviewers (J.F., L.K.-D.L., M.L.C., J.K.P., O.C., H.N.Q.T., M. Sultana, N.H.). Any discrepancies in data extraction were resolved by a third reviewer (J.F., J.K.P.), who was not involved in the initial extraction. Data were extracted on the target population, intervention(s) and comparator, economic evaluation framework, study design, perspective, time horizon, reference year, discount rates, currency, cost categories, outcomes measured and cost-effectiveness findings. There were no data extraction issues that warranted contacting the authors of included studies. However, if studies did not report an economic reference year, then it was assumed that the reference year was 2 years before the year of publication. To allow comparisons of value across studies, the reported intervention costs and ratios were converted into 2019 US dollars, using the EPPI-Centre cost conversion online tool.¹²

Synthesis of study findings

Results were presented for the following age groups: children and adolescents (<18 years), adults (18–65 years) and older adults (>65 years). A meta-analysis was not conducted because of the substantive heterogeneity observed between studies in relation to the population, intervention, comparator, outcome and economic evaluation frameworks. We employed narrative synthesis together with a dominance ranking framework to synthesise study methods and findings. The dominance ranking framework presents the distribution of interventions across three decision criteria (i.e. favour, unclear decision or reject an intervention). This framework was adapted from the guideline developed by the JBI.¹⁰ Two reviewers (J.F., L.K.-D.L.) conducted the dominance framework classification. Dominance ranking was based on the results reported by the studies, and traffic light colour coding was used to indicate implications for decision makers. ‘Red’ signifies study results where routine adoption of the intervention is likely to be less favoured or rejected by decision makers (i.e. costs are higher and the intervention is less effective). ‘Green’ denotes study results that suggest an intervention is potentially very acceptable or favourable to decision makers (i.e. has better health outcomes and lower costs). ‘Yellow’ signifies study results that do not provide a clear-cut decision for decision makers (i.e. the intervention is ‘more effective and more costly’ or it is ‘less effective and less costly’). In this case, some form of financial or clinical trade-off is required. Willingness-to-pay thresholds can be used here to determine whether the intervention is cost-effective and represents value for money.

Quality assessment

Reporting and quality assessment was completed with the Drummond ten-point checklist.¹³ Despite planning to use the Quality of Health Economic Studies (QHES) tool, we ultimately opted to use the Drummond ten-point checklist because it can be applied to both trial- and model-based economic evaluations (the QHES is only applicable to model-based evaluations). Two independent reviewers were involved in the quality assessment of included studies. Conflicts were resolved by a third reviewer (J.F., L.K.-D.L., M.L.C.). There are 33 sub-items attached to the ten overarching Drummond criteria, which can be answered as ‘yes’, ‘no’ and ‘cannot tell’.¹³ Items that were relevant but did not have sufficient information to judge ‘yes’ or ‘no’ were marked with ‘cannot tell’. Reporting and quality assessment were completed in Microsoft Excel, with two reviewers independently assessing the quality of included studies. To limit inconsistencies in assessment, the authors met to discuss and assess two identified studies (a trial-based and a model-based economic evaluation). An average score was calculated to gauge the quality of the studies. ‘Yes’ answers were assigned a score of 1; ‘no’ answers were assigned a score of 0 and ‘cannot tell’ were assigned a score of 0.5. Studies that scored at least 9 were considered of good quality, studies that scored 6 to <9 were considered of fair quality,¹⁴ and studies scoring <6 were deemed poor quality, but were still presented to show the entirety of the available evidence. Quality assessment was also discussed narratively to describe the characteristics of identified studies. Post-quality assessment internal consistency was measured with the Kuder–Richardson Formula 20 (KR20). For each study, a binary entry (0 for conflict or 1 for agreement) was used to represent independent quality assessment. A KR20 coefficient ranges from 0 to 1, with a score closer to 1 indicating high internal consistency.

Results

A total of 5674 articles were identified during the literature search. After removing duplicates and title and abstract screening, 488 articles remained for full-text deliberation. There were 364 articles remaining after full-text screening. Of these, 57 studies met the inclusion criteria for the prevention of alcohol use (Fig. 1). The main reasons for exclusion were as follows: being outside the scope of the review (e.g. prevention of other substances or substance use disorder treatments), did not meet the criteria of a full economic evaluation, incorrect disease population, incorrect study designs or publication type, and wrong or no outcomes. Hand searching further identified 12 studies. The reasons for missing out on the 12 articles identified through hand searching were exclusion through screening (n = 2) and from combining the concepts within the search strategy (n = 10). There were 34 model-based evaluation studies, 28 trial-based evaluations and seven studies that included both model- and trial-based economic evaluations. The economic evaluation frameworks comprised CEA (n = 21), CUA (n = 18), CBA (n = 9) and ROI (n = 4). There were also studies that used multiple economic evaluation frameworks, including CBA plus CEA (n = 4) and CEA plus CUA (n = 13). Sixty-one studies were for the general or adult population, and the remaining studies involved children (n = 7) or older adults (n = 1). Further details of included economic evaluations are found in Table 1.

Fig. 1 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram.

Table 1 Data extraction of included economic evaluations

CEA, cost-effectiveness analysis; RCT, randomised controlled trial; GBP, Great British Pound; GP, general practitioner; PFBA, personalised feedback and brief advice; eBI, electronic brief intervention; CUA, cost–utility analysis; NHS, National Health Service; PSS, personal social services; QALY, quality-adjusted life-year; ICER, incremental cost-effectiveness ratio; WTP, willingness to pay; USD, US dollars; STD, sexually transmitted disease; CBA, cost–benefit analysis; SA, sensitivity analysis; AUDIT, Alcohol Use Disorders Identification Test; SF-12, 12-item Short-Form Survey; SBI, screening and brief intervention; SBIRT, screening, brief intervention and referral to treatment; AUD, Australian dollars; DALY, disability-adjusted life-years; WHO; World Health Organization; CER, cost-effectiveness ratio; AUDIT-C, Alcohol Use Disorders Identification Test; MIFB, motivational interviewing with feedback; EQ-5D, Euro-QoL five dimension; CBT, cognitive–behavioural therapy; ROI, return on investment; %CDT, carbohydrate-deficient transferrin; RCQ, Readiness to Change Questionnaire; CORE-LD, Clinical Outcomes in Routine Evaluation; USAF, United States Air Force; BAC, blood alcohol concentration; CAD, Canadian dollar; ASSIST, Alcohol, Smoking, and Substance Involvement Screening Test; EQ-5D-3L, European Quality of Life 5 Dimensions 3 Level Version; FASD, foetal alcohol spectrum disorder; NMB, net monetary benefit.

a. Dollar conversion with the EPPI-Centre Cost Converter.

b. Quality assessment with the Drummond ten-point checklist.

Children, adolescents and young adults

Trial-based economic evaluations

Six economic evaluations were done alongside randomised controlled trials (RCTs) of preventive interventions targeting children, adolescents and young adults. One evaluation also incorporated economic modelling.¹⁵ Intervention settings varied from school-based to family-based to e-health. The time horizons of the trial-based economic evaluations ranged from 4 months to 5 years. All studies were set in high-income countries, with the majority of the studies located in the USA (n = 3). The perspectives that were adopted included societal (n = 2) and healthcare sector (n = 2).

One school-based intervention for adolescents and their parents/carers was found to be cost-saving in reducing alcohol use and binge drinking episodes compared with education as normal.¹⁶ Electronic brief intervention (eBI) and personalised feedback with brief advice (PFBA) were evaluated against screening only; with screening dominating eBI and PFBA not being cost-effective compared to screening.¹⁷ A web-based game with feedback on alcohol awareness targeting adolescents aged 15–19 years was more effective and more costly compared with care as usual, with an incremental cost-effectiveness ratio (ICER) of $83 per reduction of one glass of alcohol or $192 per reduction of binge drinking occasion per 30 days.¹⁸ A community mobilisation strategy to reduce youth substance use, delinquency, violence and other problem behaviours was cost-saving compared with a control community, with a benefit–cost ratio of $8.22 per dollar invested.¹⁹ A teenage prevention programme had a 90% probability of being cost-effective compared with attention control, using a willingness-to-pay threshold of $118 per dollar invested.²⁰

Model-based economic evaluations

Two model-based economic evaluation studies were included for children, adolescents and young adults.^15,21 Both studies used a societal perspective, with one study using a 5-year time horizon and the other using a lifetime time horizon. The two studies separately evaluated a family-based intervention, a parenting-only intervention and an intervention involving alcohol screening and counselling visits. The family-based intervention, which was about parenting skills with child involvement, produced a benefit–cost ratio of $9.60 per dollar invested, whereas the parenting-only intervention had a benefit of $5.85 per dollar invested when compared with minimal contact.¹⁵ The alcohol screening intervention plus the provision of counselling to youth identified at high risk of alcohol harms over 5 years would be cost-effective under the willingness-to-pay threshold of $978 047 per life-year saved compared with standard care, if programme efficacy was estimated at 5.6%.²¹

Adults

Trial-based economic evaluations

The review identified 30 trial-based evaluations targeting adults. Twenty-one evaluations were conducted alongside RCTs, with the remainder conducted alongside non-RCT studies. Most studies were conducted in high-income countries such as the USA (n = 15), followed by the UK (n = 7), Australia (n = 2) and The Netherlands (n = 2). Only two studies were conducted in low- and middle-income countries, including India and Thailand.^22,23 CEA (n = 11) was commonly used followed by combinations of multiple frameworks (n = 8), CBA (n = 4), CUA (n = 4) and ROIs (n = 3). Most studies adopted perspectives from societal (n = 5), healthcare sector (n = 5) or both (n = 5), with time horizons ranging from 6 weeks to 6 years.

Economic evaluations conducted alongside RCTs evaluated several interventions, including brief intervention or brief advice (n = 11), motivational interviewing and/or counselling (n = 5), and internet/computer-based interventions (n = 3). Evaluations of brief interventions reported ICERs of $0.40 to $303 per reduction in drinks per week^24–26 and a benefit–cost ratio of $39 per dollar invested.²⁷ The results of brief advice varied from being not cost-effective when compared with health and lifestyle leaflet,²⁸ to having benefit–cost ratios of $5.6 (at 6 months)²⁹ and $39 (at 12 months)³⁰ per dollar invested compared with a control group. Motivational interviewing or counselling reported a range of CEA results, from being dominated (i.e. more costly and less effective) when compared with assessment only,³¹ to cost-saving when compared with enhanced usual care.²² Motivational interviewing was found to be dominant (i.e. less costly and more effective) compared with minimal intervention.³² Incorporating a patient's significant other into motivational interviewing (SOMI) had benefit–cost ratios of $4.23 (societal) and $5.13 (healthcare) per dollar invested when compared with motivational interviewing only.³³ Web-based alcohol interventions reported ICERs ranging from being dominated when compared with minimal intervention, to $393 616 per QALY gained against measurement only.^34–36

Economic evaluations done alongside non-RCT studies involved study designs such as pre–post, quasi-experimental and retrospective. Evaluated interventions included work-based interventions (n = 3); brief intervention (n = 2); screening, brief intervention and referral to treatment (SBIRT) (n = 2); a health promotion programme called the Integrated Management of Alcohol Intervention Program (i-MAP) and a community programme. The work-based interventions generated mixed ROI ratios ranging from no cost-savings³⁷ to $3.92³⁸ for every dollar spent, and a benefit:cost ratio of 26:1³⁹ when compared with doing nothing. Different delivery methods for brief intervention were evaluated against usual care, with no impact on outcomes found.⁴⁰ Another evaluation compared brief intervention (one to five sessions from 5 min to 1 h) against brief treatment (five to 12 1-h sessions intended for patients with higher risk factors) and reported that brief intervention was not better in terms of reducing the probability of using any alcohol, but was better in reducing the proportion using alcohol to intoxication, days of alcohol use and days of alcohol use to intoxication.⁴¹ SBIRT was found to be cost-saving compared with usual care,⁴² with a 21% reduction in healthcare costs and significant reductions in 1-year emergency department visits compared with usual emergency services use.⁴³ The health promotion intervention i-MAP generated an ROI ratio of 2:1,²³ and the community-based or multicomponent alcohol prevention programme was found to be cost-saving in terms of addressing violent crime.⁴⁴

Model-based economic evaluations

There were 33 model-based economic evaluations, including an RCT study³⁵ that modelled a longer time horizon and a pre–post study⁴² utilising Monte Carlo simulation. Most evaluations were conducted in high-income countries, including the USA (n = 8), Australia (n = 6), Canada (n = 4), the UK (n = 4), Denmark (n = 2), The Netherlands (n = 2), Estonia (n = 1) and Italy (n = 1). Low- and middle-income countries included Kenya (n = 1) and Thailand (n = 1). There was also one evaluation conducted in a European context and two studies that used a global context. Most studies involved CUA (n = 13), followed by CEA (n = 7), CBA (n = 3) and ROI (n = 1). Nine studies used a combination of different economic evaluation frameworks. The majority of studies adopted a health sector perspective (n = 17) and nine studies adopted a societal perspective. Almost half of the studies used a lifetime time horizon (n = 15).

The majority of the model-based interventions included upstream interventions such as policy and taxation (n = 15). Most CEAs of policy and taxation found that it generated cost-savings.^45–48 Studies that adopted the CUA framework reported results that were cost-saving when compared with doing nothing.^49–51 Brief intervention, SBI and SBIRT were the second-most modelled interventions (n = 17). Modelling indicated that brief intervention and SBI were cost-effective.^51–66 Cognitive–behavioural therapy showed cost-savings,⁶⁷ and motivational interviewing was cost-effective under a willingness-to-pay threshold of $50 000 per QALY.⁶⁸

The models identified in this review were primarily multistage life table models, with or without a preceding decision tree. Individual-based models such as microsimulation and Monte Carlo simulation were also used, whereas simpler decision tree models were less common. Multistage life tables were used to estimate the incidence, prevalence, remission and mortality of alcohol-related diseases and injuries based on population life tables. These models can predict the demographic consequences derived from introducing new interventions (because of changes in the key parameters of incidence, prevalence remission or mortality).⁶⁶ The changes in alcohol use affect the mortality and prevalence of alcohol-related diseases (e.g. liver failure), as well as the mortality and incidence of alcohol-related injuries (e.g. road accidents). These reductions then influence overall rates of mortality and disability in the population.⁵¹ The most commonly used multistage life table models included the model developed for the Adverse Childhood Experiences Prevention study,⁶⁶ the Sheffield Alcohol Policy Model,⁶¹ the Chronic Disease Model⁶⁹ and the model developed for the WHO-CHOICE study.⁵⁵ Each multistage life table model applied potential impact fractions (PIFs) to estimate treatment effects. The PIF is an epidemiological measure of effect that calculates the proportional change in average disease incidence, prevalence or mortality after a change in the population distribution of a risk factor exposure.^70,71

There was substantial heterogeneity between the different models on the number of health conditions attributable to alcohol use. The most comprehensive models covered up to 22 health conditions attributable to alcohol use.⁶¹ Non-communicable diseases (NCDs) were the most common conditions, followed by alcohol-related injuries. Alcohol dependence/alcohol use disorders were the only mental disorders included in existing models.

Model-based economic evaluations of interventions to prevent multiple risk factors, including alcohol use

Four studies evaluated preventive interventions that included multiple risk factors, including alcohol use. Cadilhac et al⁷² modelled a hypothetical cohort to evaluate the cost–benefit of feasible reductions in six common risk factors over a lifetime (without decay). These risk factors include tobacco smoking, inadequate fruit and vegetable consumption, excessive alcohol use, high body mass index, physical inactivity and intimate partner violence. Cadilhac et al corrected the joint effects by using the joint population attributable risk fraction that was outlined in the 2003 Australian Burden of Disease and Injury study⁷³ and by the World Health Organization.⁷⁴ This formula is based on the assumption that health risks are independent.⁷⁴ Results showed that reducing these risk factors saved 2334 million Australian dollars for the 2008 Australian adult population (or 4022 million in 2019 US dollars).

A pre–post study conducted in the USA showed that the Health Risk Management programme to reduce ten risk factors in workers (i.e. poor eating habits, physical inactivity, tobacco use, excessive alcohol use, high stress, depression symptoms, high blood pressure, high total cholesterol and high blood glucose) would produce a return of $2.03 per dollar invested within 1 year of follow-up.⁷⁵ Growth curve modelling of a company health promotion and lifestyle programme was evaluated for the benefit of reducing rates of obesity, high blood pressure, high cholesterol, tobacco use, physical inactivity and poor nutrition over a 6-year time horizon. Despite having no effect on average alcohol consumption, the programme showed a return of $3.92 per dollar spent when all benefits were accounted for.³⁸ A recent economic evaluation alongside a trial conducted by Kruger et al found that within the 6-month follow-up, a theory-based online health behaviour intervention implemented in university was not cost-effective in reducing unhealthy eating, physical inactivity, binge drinking and smoking.³⁵ However, by extrapolating the efficacy of the intervention over a lifetime and rolling out the intervention to other universities, the intervention became cost-effective, with an ICER of £1545 ($2493) per QALY gained. This result is well below the UK willingness-to-pay threshold of £20 000 (around $28 653) per QALY gained.

Older adults

Only one study was found for older adults with excessive alcohol use. Although this CBA reported lower healthcare and societal costs favouring the intervention group, it also showed no significant differences in costs between the intervention and control groups.⁷⁶

Quality assessment results.

There were 26% of conflicts registered out of a possible 2277 pairings from the 33 quality assessment sub-items. Post-assessment internal consistency was calculated and resulted in a KR20 coefficient of 0.76, which indicates acceptable internal consistency. Most studies (84%) were fair (n = 43) or good (n = 15) quality. Only two economic evaluations done alongside trials, Kuklinski et al¹⁹ and Tanaree et al,²³ met all ten points of the quality checklist. Less than half of the studies lacked a clear description of the competing alternatives, relevant costs and consequences because of non-inclusion or non-reporting of capital costs. Most studies did not adequately present and discuss study results in terms of implementation, generalisability and future directions. Further details are presented in Table 2.

Table 2 Quality assessment of included economic evaluations

Colour grading of cost-effectiveness results

Figure 2 presents a summary of the classification for different interventions graded based on their results and grouped as either likely to be rejected, favoured or unclear from a decision-making perspective. Half of the interventions were found to be cost-saving for the prevention of alcohol use and 84% of studies were rated as either fair (53%) or good (32%) quality. Most interventions were delivered to adults or the general population, except for four interventions targeting children and their parents and one intervention for older adults. Specifically, universal prevention strategies restricting access to alcohol through taxation or advertising bans and selective/indicated prevention through screening with or without brief intervention accounted for most of the studies.

Fig. 2 Cost-effectiveness summary of interventions. SBI, screening and brief intervention; SBIRT, screening, brief intervention and referral to treatment.

Another 35% of interventions were categorised as ‘unclear’ because they produced improved health outcomes at a higher cost. Two-thirds of economic evaluations were fair quality, followed by studies rated as good (17%) or poor (17%) quality. Most interventions restricted exposure to alcohol through taxation with or without advertising bans for general populations or selective/indicated prevention through screening with or without brief intervention for targeting adults.

A total of 14% of interventions from the economic evaluations were categorised as ‘reject’ (i.e. less effective and more costly). Around a quarter of these studies were good quality, and two-thirds were fair quality.

Several interventions show cost-effectiveness results with a high degree of uncertainty, meaning that they comprise cost-effectiveness evidence that simultaneously indicate ‘favour’, ‘unclear’ and ‘reject’ decisions. These mixed results were affected by the variance around the choice of study elements. This suggests that a particular intervention may be acceptable or not appropriate in certain situations or contexts. For example, in universal interventions, both breath testing and brief advice had varying cost-effectiveness results. The ‘favour’ judgement for breath testing was from a modelling study showing cost-savings when compared with doing nothing.⁴⁵ However, two other studies modelled both breath testing and brief advice against different comparators, with one comparing both with current situation⁷⁷ and the other using taxation as the comparator.⁵⁶ In the first study, both breath testing and brief advice had more costs and less DALYs compared with current situation in the Estonian population.⁷⁷ In the second, a global regional modelling study, brief advice had cost-effectiveness results ranging from ‘favour’ in some regions (because it was dominant) to ‘reject’ in other regions (because it was dominated by taxation).⁵⁶ The study reported that generally, both interventions incurred higher costs than taxation. In terms of DALYs, regions with higher levels of heavy alcohol use lean toward taxation being more effective, whereas breath testing and brief advice are more effective in regions with less prevalence of heavy alcohol use.⁵⁶

In terms of ‘indicated’ interventions for adults, motivational interviewing showed cost-effectiveness judgements ranging from ‘favour’ to ‘reject’. SOMI reported ‘favour’ judgement from a positive cost–benefit ratio in both healthcare and societal perspectives when compared with standard motivational interviewing. Standard motivational interviewing reported ‘unclear’ cost-effectiveness results compared with standard care.⁶⁸ Two motivational interviewing studies reported ‘reject’ results because it was dominated by motivational interviewing with feedback³¹ and by the intervention Motivational Assessment Program to Initiate Treatment (MAPIT).⁷⁸ Motivational interviewing studies that used societal perspectives were judged ‘favour’ or ‘unclear’, whereas ‘reject’ cost-effectiveness results were found from narrower provider and probationary system perspectives.

In ‘selective’ interventions, web-based health promotion had both ‘unclear’ and ‘reject’ judgements. A study reported cost-effectiveness results for web-based health promotion judged as ‘unclear’ because of higher costs and higher utilities for the intervention compared with doing nothing, in a population of university students. The analysis was over 6-month and lifetime time horizons, and only used intervention and rollout expenses for costs.³⁵ In another study, web-based intervention was compared with minimal intervention. Reported results were mixed with ‘unclear’ judgement for CEA (lifestyle factor score improvement outcome) and ‘reject’ judgement for CUA. This evaluation was from a societal perspective and over a 2-year time horizon for the people with computer and internet access with basic internet literacy.³⁶

Discussion

The current review provides an update on the cost-effectiveness evidence for the prevention of alcohol use across the lifespan. The number of studies included in this review is nearly two and a half times more than those included in a previous review.⁶ Most of the cost-effectiveness evidence has been evaluated for interventions targeting adults. There were limited economic evaluations of interventions targeting children, adolescents and older adults. Furthermore, most studies were conducted in high-income countries, particularly using trial-based economic evaluations. Less cost-effectiveness research has been undertaken in low- and middle-income countries. Half of the evidence estimated that preventive interventions for alcohol use were cost-saving. The interventions frequently found to be cost-saving were ‘universal’ prevention, consisting largely of increasing the price of alcohol via taxation or reducing exposure to alcohol via advertising bans. Most of these interventions were compared against doing nothing or having no policy in place. Selective/indicated prevention, such as screening with or without brief intervention, was also found to be cost-saving when compared with doing nothing. It is also encouraging that school-based interventions, with and without interventions for parents (selective prevention), were found to be cost-saving, albeit in a limited number of studies. However, it is important to note that it is difficult to determine which intervention is the optimal choice, given that little evidence was established to compare different interventions within a single-study context. In terms of study quality, most studies included in this review had fair to good quality.

The results of this review provide important economic evaluation evidence to support the implementation of alcohol prevention interventions at a population level. However, it should be noted that although economic evaluation offers a useful format (with one concise indicator) for decision-making, it is not a perfect instrument. In particular, results across different economic evaluations are not comparable because of the variations in methodology, such as how utility scores for calculating QALYs were measured (differences in outcome measurement tool used) or the context in which the intervention was conducted. Other implementation considerations, such as equity, feasibility, sustainability and acceptability, may not be adequately addressed by economic evaluations. Therefore, even if there is clear evidence of effectiveness or cost-effectiveness, it does not necessarily guarantee intervention uptake.

The conflicting cost-effectiveness findings observed across several interventions were the result of substantive variations in study design. These variations limit the ability to conduct any prospective meta-analysis, highlighting a limitation of economic evaluations. Chisholm et al effectively demonstrated this issue where differences in study design or data inputs from different countries resulted in varying cost-effectiveness estimates.⁵⁶ In their evaluation, alcohol taxation was-cost saving in the USA and European countries; however, it was found to be more effective and more costly in African and Asian countries, with ICERs under a willingness-to-pay threshold of $50 000 per DALY.⁵⁶ Therefore, it is important that policy decisions be aided by adequate, context-specific research to determine which interventions can be considered value for money.

The paucity of cost-effectiveness studies on alcohol prevention among children and adolescents is in stark contrast to the literature on the cost-effectiveness of mental health promotion and prevention, where most of the existing research has focused on children, adolescents and youth.⁷⁹ Prevention of alcohol use in adolescents is important given that early use of alcohol predicts frequent drinking, leading to future alcohol-related harms.⁸⁰ Furthermore, the frequency of adolescent drinking is also predictive of substance use problems in adulthood. Further research is urgently needed to establish the value-for-money credentials of interventions to prevent or delay alcohol use in this age group.

Trial-based economic evaluations primarily evaluated indicated prevention interventions, whereas model-based economic evaluations primarily evaluated universal prevention interventions. This is sensible because universal preventive interventions are expected to have broad effects that may take years to be realised and are difficult to properly evaluate in a trial. Furthermore, model-based economic evaluations can estimate the long-term effects of alcohol use, including its effects on NCDs. Ideally, it is important to capture the full breadth of long-term effects produced by alcohol prevention interventions over the life course. However, the effect of prevention interventions over the long term becomes more uncertain, as extrapolating the longer-term effects of an intervention typically necessitates the use of assumptions that are not based on empirical evidence. Furthermore, the effects of intervention have been found to attenuate over time in long-term follow-up studies.

This review also included economic evaluations evaluating multiple risk factors, including alcohol use. However, only four studies were found that focused on combined alcohol use and risk factors for physical health conditions (e.g. obesity or NCDs). There is currently no evidence on the cost-effectiveness of preventive interventions for alcohol use and risk factors for mental health conditions. Given the high prevalence of comorbidity between alcohol use disorder, other drug use disorders and mental disorders,⁸¹ further research should explore the impact of interventions on risk factors for both physical and mental health conditions.

This review has several limitations. Only peer-reviewed articles published in the English language were included, which may have contributed to the lack of studies conducted in low- and middle-income countries. It is also common for economic evaluations, especially ROI studies, to be published in grey literature rather than in the academic literature, potentially limiting the studies identified. In addition, the involvement of multiple reviewers in screening and extraction may have resulted in inconsistencies. Meta-analysis was also not possible given the high level of methodological heterogeneity in the populations, interventions, comparators and outcomes, as well as economic evaluation frameworks across included studies. Furthermore, the majority of cost-effectiveness evidence supported the prevention of alcohol use, raising concerns of publication bias. Alternatively, a strength of this review is the use of a dominance ranking framework to summarise and provide recommendations for policy and practice.⁷⁹

In conclusion, this study found that prevention interventions for alcohol use are promising and likely provide good value for money. These findings will be of value to policy makers and other stakeholders interested in preventing alcohol use and/or excessive alcohol use. Nevertheless, policy decisions should still be aided by adequate, context-specific research on possible prevention interventions, to determine whether such interventions would be value for money. Future economic analyses are needed for low- and middle-income countries, as well as for children, adolescents and older adults. Moreover, research on cost-effectiveness with longer follow-up is also required, as it is uncertain whether the modelled longer-term effects of interventions will, in fact, be realised.