The European Journal of Public Health Advance Access originally published online on March 25, 2007
The European Journal of Public Health 2007 17(6):618-623; doi:10.1093/eurpub/ckm017
Alcohol Consumption |
A cost-effectiveness analysis of alcohol prevention targeting licensed premises
Anna M. Månsdotter1,2, Malin K. Rydberg1, Eva Wallin2, Lars A. Lindholm1,3 and Sven Andréasson1,2
1 Swedish National Institute of Public Health, SE-103 52 Stockholm, Sweden
2 Karolinska Institutet, Department of Public Health Science, Division of Social Medicine, SE-171 76 Stockholm, Sweden
3 Umeå University, Department of Public Health and Clinical Medicine, Epidemiology and Public Health Sciences, SE-901 85 Umeå, Sweden
Correspondence: Anna Månsdotter, Swedish National Institute of Public Health, SE-103 52 Stockholm, Sweden, tel: +46 8 566 135 00, fax: +46 8 566 135 05.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
Background: A multi-component alcohol prevention programme targeting licenced premises has been ongoing in Stockholm since 1996. An earlier study has established that this led to a 29% reduction in police-reported violence. The objective of the present study is to calculate the programme's cost-effectiveness from a societal perspective; the cost of implementation, the savings made as a result of fewer assaults, unlawful threats and violence towards officials, and the health gains in terms of quality-adjusted life-years (QALYs). Methods: The costs included administration, studies of alcohol serving practices, community mobilization, responsible beverage service training and stricter alcohol law enforcement. For the purpose of estimating how the decrease in violence affected savings and health gains, a survey among victims of violence (N = 604) was performed. Results: The cost of the programme was estimated at Euro 796 000. The average cost of a violent crime was estimated at Euro 19 049, which implies overall savings of Euro 31.314 million related to the judicial system (78%), production losses (15%), health care issues (5%) and other damages (2%). Accordingly, the base case cost-saving ratio was 1 : 39. The average loss of health state weighting among the victims at 0.09 translates into 236 gained QALYs for society as a whole, which should be compared with the modest proportion of savings in the health sector. Conclusion: The most significant concern is the low response rate (35%), and caution needs to be exercised when interpreting our results. Yet, a reasonable conclusion is that the monetary and human benefits have been considerable.
Keywords: alcohol, cost-effectiveness, prevention, violence
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
Since 1996, a multi-part programme has been ongoing in Stockholm combining community mobilization, training of responsible beverage service (RBS) which means not serving alcohol to intoxicated and underage patrons, and stricter enforcement of existing alcohol laws.1 Representatives from both the authorities and the hospitality industry are running the program. The main goal is to reduce problems related to alcohol consumption at licenced premises, including violence. That is, although the causal relationship is debated, research has shown that alcohol consumption in general, and at licenced premises in particular, is associated with violence.2–4
Between January 1998 and September 2000, the programme (from here on referred to as the ‘restaurant intervention’ where ‘restaurant’ refers to all types of licenced premises) was evaluated by Wallin, Norström and Andreasson.5 At that time, the restaurant intervention was active in a geographically limited part of Stockholm City, the northern part. Södermalm, a similar, but not adjacent area of Stockholm City, was defined as the control area for the evaluation. The effects were measured using time-series analyses (ARIMA) by comparing the intervention and control areas with respect to differences in police-reported incidents of assault, unlawful threat/harassment and assault/threat towards officials such as policemen and doormen; committed indoors and outdoors between 10:00 pm and 6:00 am. The results demonstrated that the restaurant intervention led to a 29% reduction in the considered indicators of violence.
The objective of the present study is to calculate the cost-effectiveness of the restaurant intervention from a societal perspective. Specifically, we aim to estimate the costs of performing the intervention, the savings due to prevented violence, and the health gains in terms of quality-adjusted life-years (QALYs).
There is a lack of health economic analyses of community-based interventions aimed at reducing alcohol-related problems. Yet, an American cost-benefit study of increased enforcement of laws forbidding the serving of alcohol to intoxicated patrons at taverns has shown that the monetary benefits to society greatly exceed the costs.6 As far as we know, few studies have also investigated the health-related quality of life among victims of violence. One recent study establishes, however, that health among victims of violence is substantially diminished compared with optimal health and other health conditions.7
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
The restaurant intervention
The restaurant intervention represents one part of a larger programme called STAD (STockholm prevents Alcohol and Drug problems), which also includes the Youth project and the Health care project. STAD aims at developing, implementing and evaluating preventive methods in a local community setting.8 The studied intervention is divided into three main components: community mobilization aimed at increased awareness of problems connected with alcohol consumption and at seeking support for action, a 2-day RBS training course about alcohol law, medical effects and conflict management for servers, doormen and restaurant owners, and lastly, strengthened enforcement of alcohol laws.
Survey among victims of violence
The foundation of the cost-effectiveness analysis was the established effectiveness of the restaurant intervention.5 For the purpose of estimating what the 29% decline in violence meant in terms of societal savings and health gains, a survey was conducted among victims.
The respondents were randomly selected from the National Police Board's register of assaults, unlawful threats/harassment, and assaults/threats towards officials during 2003 in Stockholm City, i.e. the same indicators as in the study by Wallin, Norström and Andreasson.5 In order to obtain firm cost estimates, the proportions of respondents varied by type of crime, for instance, all proper cases of serious assault towards women were selected, while every 11th case of unlawful threat was selected. A total of 604 questionnaires were sent out to victims of serious assaults (n = 83), other assaults (n = 288), unlawful threat (n = 152) and assault/threat towards officials (n = 81). The first dispatch (May 2005) was followed by two reminders.
The total number of the types of violent crimes considered during the intervention period (January 1998 to September 2000) and area (Northern part of City) was 7368, and the established preventive effect of the restaurant intervention was 29%, i.e. the number of prevented violent crimes was set at 3009 cases. A basic assumption when calculating savings and health gains was that the proportions of prevented violence were similar to the proportions of occurred violence; this means 56% assault, 13% unlawful threat/harassment and 31% assault/threat towards officials.
Costs
The costs of the restaurant intervention include administration, studies of alcohol serving practices, community mobilization, RBS training and alcohol law enforcement.
Administration
The costs of administration relate mainly to salaries for project staff, but also offices, travel, conferences, office supplies and literature, as well as overhead costs (phone, postage, cleaning, etc.). They were estimated based on the annual account statements for STAD during 1995–2000. Since the number of personnel employed has been similar for the three main projects, the administration costs are taken to be one third of the total costs.
Studies
The study costs regard remuneration to actors for studying licenced premises as to the serving of alcohol to underage and intoxicated guests in two initial mapping studies (1996–1997) and in two follow-up studies (1998–1999), as well as training of actors/young people and the purchase of alcoholic drinks. The personnel costs have been calculated with average salaries for each profession (payrolls from trade unions and employers), plus payroll taxes (Swedish Tax Board), while the time consumption was estimated by the project manager.
Community mobilization
The key component of community mobilization is an advisory group of representatives from STAD, the Licensing Board of Stockholm, the National Police Board, the National Institute of Public Health, the County Administration, the Restaurant Owners Association and the Restaurant Staff Trade Union. In addition, smaller groups were formed aimed at discussing specific topics. The costs have been estimated based on minutes of meetings and on average salaries/payroll taxes.
RBS training
The costs regarding the 2-day course in RBS are divided into three major items. First, the costs for material, food and actors were approximated with the course fee (Swedish Kronor 500, or Euro 53.8) multiplied by the number of participants. Second, the cost for adjusting film and brochures to local conditions was set at 50% of the original production cost. Third, the production loss for the 572 course participants was calculated with participation lists, and for the lecturers with the course programme; together with average salaries/payroll taxes.
Law enforcement
The costs related to enforcing existing alcohol laws consist of two items. First, joint controls by the Licensing Board and the Police have been considered by assuming that one control is performed by two policemen and two supervisors, and requires 15 min of planning and 30 min of actual control. Second, the increase of notification letters in the intervention area compared with the control area was considered by assuming that one letter involves the work of one officer for 1 h. The information on numbers and time, which comes from the Licensing Board, was combined with average salaries/payroll taxes.
Savings
The savings were based on the survey of victims of violence, which consisted of questions aimed at calculating the monetary consequences to the judicial system (police, public prosecutor, court and prison/probation), production changes (sickness absence, and other effects on working hours/salaries), health care (transport from crime scene, emergency treatment, inpatient care, outpatient care and pharmaceuticals) and other damage (technical support, personal assistance and personal belongings).
Responses related to the judicial system were combined with data from the National Police Board, the Office of the Public Prosecutor, the National Courts Administration and the National Prisons and Probation Administration, while responses related to health care issues were combined with data from the Federation of Swedish County Councils, the National Dental Health Service, the Association of Pharmacy and the Pharmaceutical Benefits Board. Production changes were calculated based on the respondents’ reports on salaries, sickness absence and changes in working conditions before/after the crime, together with payroll taxes. Finally, the costs regarding other damages were entirely based on the questionnaires.
Health gains
The survey among victims of violence also included questions aimed at quantifying how the restaurant intervention affects health. The respondents were asked to classify their health state ‘before violence’, ‘two weeks after violence’, and ‘at present’, with the questionnaire EQ-5D (EuroQol Group).9 This is based on five dimensions of health (mobility, self-care, usual activities, pain/comfort, anxiety/depression), which in turn consist of three levels (no problems, some/moderate problems and extreme problems). The health states revealed were converted to a weighted health state index by applying values of health profiles based on a UK population sample.10 The health gains were reported in QALYs by comparing the weighted health state ‘before the crime’ with the average health state of present ‘two weeks after’ and ‘at present’ (cf. health improvements after the violence were assumed to be linear).
The respondents were also asked to rate their average health status by EQ VAS on a scale from 0 (worst state) to 100 (best state) during the period from the violent crime and 12 months after. These ratings were compared with similar ratings among the general public in the county of Stockholm in 2002.11,12
General remarks
The study was approved by the Research Ethics Committee at Karolinska Institutet in April 2004. Costs and savings were adjusted to the mean price level of 2005 with the consumer price index (Statistics Sweden), and transformed from Swedish Kronor (SEK) to Euro (EUR) by the mean exchange rate January–September 2006 (The Riksbank—Sweden's central bank). Costs, savings and health gains are presented in both undiscounted and discounted estimates (base year: 1995, discount rate: 3%). All statistical analyses have been performed in SPSS, version 13.0.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
Costs of the restaurant intervention
Table 1 shows the costs of planning and implementing the restaurant intervention per year and for the major components of project administration, studies of mapping and follow-up, community mobilization, RBS training and alcohol law enforcement (undiscounted values). It is demonstrated that the major cost refers to administration (EUR 536 105) followed by the RBS training (EUR 248 970). Studies and community mobilization costs EUR 29 825 and EUR 38 092, respectively, while law enforcement costs EUR 24 414. The total costs amount to EUR 877 406, which equals EUR 795 828 after discounting.
|
Analysis of non-response
Among the 604 questionnaires sent out, 47 did not reach the intended addressee, and 194 recipients responded. The overall response rate was hence 32% (all included) and 35% (not reached excluded). An analysis of non-response (table 2) based on individuals who received the questionnaire (n = 557) demonstrates statistically significant differences (at the 5% level) between respondents and non-respondents with regard to sex and time point of violence. Females are more inclined to answer than males, and victims between 10:00 pm to 6:00 am are less inclined to answer than victims at other times of the day.
|
Costs of violence
The average cost of a violent crime in the study was estimated at EUR 19 049. In table 3, the monetary consequences are illustrated based on: type of violence, sex, age group, time of violence, place of violence, educational level, as well as sector and subheadings. Since the type of violence constitutes the basis for calculating savings, 95% confidence intervals (CI) are presented for these estimates.
|
Violence towards females is on average more costly per case than towards males. Furthermore, violence in the 34–43 age group involves the highest cost while violence in the
64 age group involves the lowest; and violence occurring between 10:00 pm to 6:00 am is more costly than violence between 6:00 am and 10:00 pm. Moreover, the cost of violence perpetrated at home indoors is highest, and so is the cost for a victim with an educational level of 3 years or more of university studies. Finally, the overwhelming costs of violence arise from the judicial system (EUR 14 847) and in particular from prison and probation (EUR 12 372), followed by costs due to diminished production, health care including transport and pharmaceuticals and other damages.
Savings
The basis of our base case calculation of savings was that the average cost of one assault is EUR 20 145 (table 3), which is generated from 27 responses regarding serious assault and 84 responses regarding other assault (111 in total). This is most likely not a true picture of the prevented proportions, since the register of police-reported violence demonstrates that serious assault represents about 6%. Hence, the average cost for an assault was revised; 6% of 111 is 7 cases and 94% of 111 is 104 cases, which generates a cost of EUR 9935 [(63 026.45 * 7 + 6361.46 * 104)/111]. This implies undiscounted savings of EUR 35 243 858 [3009 * (0.56 * 9934.93 + 0.13 * 21 382.23 + 0.31*10 869.56)].
A sensitivity analysis based on assuming that the only cost among non-respondents was the handling of the violent crime by the Police implies overall savings of EUR 16 248 991 (0.35 * (16 740 716 + 8 364 087 + 10 139 013) + 0.65 * 2001 * 3009). In addition, a worst-case variant was calculated based on entirely ignoring the survey, i.e. the cost considered for violence was restricted to the Police handling, which generates savings of EUR 6 021 009 (2001 * 3009).
Health gains
The questions aimed at estimating health gains from the restaurant intervention by EQ-5D show that the average QALY-weights were: 0.8647 before violence, 0.6776 2 weeks after violence and 0.8729 ‘at present’ (around 2 years after violence). The average health gains associated with one prevented assault were revised in the same way as for savings, which means that the overall health gains were estimated at 266 QALYs, of which 167 concerned prevented assault (0.56 * 0.0991 * 3009), 43 unlawful threat (0.13*0.1098*3009) and 56 threat/assault towards officials (0.31 * 0.0595 * 3009).
The ratings on the EQ VAS scale from the violence and 12 months after show that the average health status was 73.3, of which: serious assault 61.8, assault 78.4, unlawful threat 64.0 and violence towards officials 84.4. This means, when revised in accordance with savings and health gains by EQ-5D, that the average health status was 77.8. That is, no health gain since 77.7 was the corresponding rating in the general public for individuals in the 35–39 age group.11
Summing up: costs, savings and health gains
Table 4 presents a summary of costs, savings by sector and health gains associated with the restaurant intervention. The net savings (discounted estimates) are EUR 30.518 million in the base case and EUR 13.641 million working on the assumption that the only violence-related consequence among non-respondents was the cost of police handling. When the responses from the survey are ignored, the cost-saving decreases to EUR 4.554 million. Corresponding health gains are 236, 83 and zero QALYs.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
Our economic analysis of an intervention targeting licenced premises in Stockholm City demonstrates a base case ratio between discounted costs (EUR 795 828) and savings (EUR 31.314 million) at 1:39. This is in accordance with the cost-benefit analysis of strengthened alcohol law enforcement in Michigan, in which the most conservative cost-saving ratio was 1:45.6 A sensitivity analysis demonstrates cost-savings at 1:18 and 1:7, respectively. The study indicates that the prevented violence has led to health gains as well; 236 discounted QALYs, which corresponds to 3–4 lives with normal health. The size of health gains may also be illustrated by the proposal that a cost of about EUR 54 000 per QALY is acceptable for the Swedish context.13 That is, even without considering savings, the restaurant intervention would have been cost-effective (795 828/236 is about EUR 3000 per QALY).
Methodological concerns
The costs were based on account statement, average cost estimates and various kinds of estimations by individuals, i.e. there are several possible flaws. There are also methodological concerns regarding, for instance, the accuracy of considering production changes in health economic analyses,14 which in our study represent about one-fourth of the costs. Moreover, since the intervention may have contributed to enforcement activities other than the joint controls by the Licensing Board and the Police Authority,15 the costs of law enforcement are most likely underestimated.
A general problem regarding both savings and health gains is that they are based on a survey in which the overall response rate was very low (35%). One likely explanation is that victims of violence are not representative of the general public; certainly not regarding age. This may also explain some of the not reached proportion (8%) since younger people move more frequently than older people; another reason is the selection of individuals with protected identities. The low response rate may have influenced the result in opposite directions; one may imagine that the most harmed victims are most inclined to respond, but also that the severely harmed hesitated most about answering the questionnaire. An indication of reliability is the finding of no statistically significant difference between respondents and non-respondents regarding the type of violence, which forms the basis for calculating savings and health gains. Another concern is that savings and health gains are entirely based on the decline of violence (29%) reported from a non-randomized study.5 This may have contributed to underestimated results (e.g. spill-over effects regarding not serving alcohol to underage patrons, and staff turnover of RBS trained servers to the control area) as well as overestimated results (e.g. problematic drinkers may have displaced to the control area).
Regarding savings, one should note the average cost of nighttime violence was EUR 3.211 higher than daytime violence. Since the effectiveness study only considered violence between 10:00 pm and 6:00 am,5 the savings are underestimated from this viewpoint. Another aspect is that there were most certainly savings excluded such as benefits for the licenced premises due to avoided damage and short-time closures. Moreover, the respondents were asked to restrict their reports to 12 months after the violence, despite the possibility of more lengthy monetary consequences. In contrast, long-term effects were considered for prisons/probations (e.g. 6 years of imprisonment for a serious assault translates to EUR 696 540), which also explains the high proportion of savings from the judicial sector vs. other sectors. The indication of underestimated savings is also strengthened by not including, e.g. the reduced administration of sick leave cases by social insurance offices, and of medication by pharmacies.
A peculiar finding is that the cost of an assault (EUR 9935) is much lower than the cost of a threat (EUR 21 382), and similar to the cost of violence towards officials (EUR 10 870). This pattern is partly valid for health gains as well; threats cause the highest QALY loss, but assaults cause more QALY loss than violence towards officials. The survey does not permit a closer investigation, but one may speculate whether unlawful threats involve more enduring violence, and hence more costs and health-damaging effects, than non-serious assaults. This is somewhat supported by the finding that, among nine reports of violence enduring for more than 1 h, seven concerned unlawful threats.
The estimation of health gains based on EQ-5D shows health gains, while EQ VAS does not. Noteworthy is that both measures were based on questions in which the respondents were asked to remember, and report, earlier health states, which may indeed be criticized. The respondents reported health states before violence similar to the general public in the same age (roughly 0.86 compared with 0.85).11 Moreover, it has been shown that the health-related utility among female victims of violence is 0.64–0.66 for less severe violence and 0.53–0.62 for severe violence,7 which corresponds to our results (2 weeks after violence: 0.66 for non-serious assaults and 0.52 for serious assaults). The main doubt is hence that the respondents report better health ‘at present’ (0.87) than before the violence (0.86), which indicates a flaw leading to underestimated health gains. Yet, the modest health gains should be compared with the small proportion of savings in the health sector (5% of total).
Finally, individuals prevented from buying alcohol, and from committing violence, presumably also experienced health and other benefits, which we did not consider.
|
Conclusion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
There are concerns in our health economic analysis of alcohol prevention targeting licenced premises, particularly in the form of a low response rate in the basic survey. Hence caution needs to be exercised when interpreting and transferring the results to other settings. Since the analyses point to consistent cost-savings and likely health gains, a reasonable conclusion is that the monetary and human benefits to society have been considerable.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
We are grateful for helpful support from the manager of the restaurant intervention and from representatives at the National Police Board. Financial support, principally regarding salaries, from the Swedish National Institute of Public Health is also gratefully acknowledged.
Conflicts of interest: None declared.
Key points
|
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionAcknowledgementsReferences |
|---|
1 Wallin E, Gripenberg J, Andréasson S. Overserving at licensed premises in Stockholm. Effects of a community action program. J Stud Alcohol (2005) 66:806–14.[Web of Science][Medline]
2 Norström T. Effects on criminal violence of different beverage types and private public drinking. Addiction (1998) 93:689–99.[CrossRef][Web of Science][Medline]
3 Bushman BJ, Cooper HM. Effects of alcohol on human aggression: an integrative research review. Psychol Bull (1990) 107:341–54.[CrossRef][Web of Science][Medline]
4 Norström T. Outlet density and criminal violence in Norway, 1969-1995. J Stud Alcohol (2000) 61:907–11.[Web of Science][Medline]
5 Wallin E, Norström T, Andréasson S. Alcohol prevention targeting licensed premises: a study of effects on violence. J Stud Alcohol (2003) 64:270–77.[Web of Science][Medline]
6 Levy DT, Miller TR. A cost-benefit analysis of enforcement to reduce serving intoxicated patrons. J Stud Alcohol (1995) 56:240–7.[Web of Science][Medline]
7 Wittenberg E, Lichter EL, Ganz ML, McCloskey LA. Community preferences for health states associated with intimate partner violence. Med Care (2006) 44:738–44.[CrossRef][Web of Science][Medline]
8 Andréasson S, Lindewald B, Hjalmarsson K, et al. Exploring new roads to prevention of alcohol and other drug problems in Sweden: the STAD project. In: Alcohol & Public Health Research Unit (1999) Kettil Bruun Society Thematic Meeting; Fourth Symposium on Community Action Research and the Prevention of Alcohol and Other Drug Problems. New Zealand: University of Auckland.
9 EuroQol Group. www.euroqol.org.
10 Dolan P, Gudex C, Kind P, Williams A. The time-trade-off method: results from a general population study. Health Economics (1996) 5:141–54.[CrossRef][Web of Science][Medline]
11 Burström K, Rehnberg C. Hälsorelaterad livskvalitet i Stockholms län – en befolkningsundersökning med EQ-5D [Health-related quality of life in the County of Stockholm – a population study based on EQ-5D]. (2006) Stockholm County Council, Centre for Public Health. 1.
12 Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-effectiveness in health and medicine (1996) New York: Oxford University Press.
13 Swedish National Board of Health and Welfare. Guidelines for Cardiac Care (2004).
14 Drummond MF, O’Brien B, Stoddart GL, Torrance GW. edn. In: Methods for economic evaluation of health care programmes (2005) 3rd edn. Corby Northants: Oxford University Press.
15 Wallin E, Lindewald B, Andréasson S. Institutionalization of a community action program targeting licensed premises in Stockholm, Sweden. Eval Rev (2004) 28:396–419.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||