British Journal of Addiction (1992) 87, 873-881
RESEARCH REPORT
Levels of drunkenness of eustomers leaving licensed premises in Perth, Western Australia: a comparison of high and low ^risk' premises TIM STOCKWELL,' PHIL RYDON,' SONIA GIANATTI,' EVAN JENKINS,^ CLAUDIA OVENDEN> & DAVID SYED^ ^National Centre for Research into the Prevention of Drug Abuse, Curtin University of Technology, Bentley 6102 & '^Medical School, University of Western Australia, Nedlands 6009, Western Australia
Abstract A measure of the risk of licensed premises having customers involved in road traffic accidents and drinkdriving offences was utilised in order to identify seven 'High Risk'and eight 'Low Risk'premises in metropolitan Perth, Western Australia. This measure, or 'Risk Ratio', was defined as the ratio of incidents of alcohol-related harm to an estimate of on-premises alcohol sales for a particular establishment. A study was conducted to test the hypothesis that a High Risk status would be associated with greater levels of customer intoxication. Interviews concerning drinking behaviour and breathalyser readings were collected from 74.2% of 414 customers exiting from the chosen premises between 8 p.m. and midnight on Friday and Saturday nights. High Risk premises had three times more customers whose readings were in excess of 0.15 mg/ml ('p < O.OI). The proportion of customers with BAL's above 0.15 correlated strongly with the premises' Risk Ratio (i = 0.63, p < 0.01). There were also significantly more patrons from High than from Low Risk establishments who were rated as appearing moderately or severely intoxicated but refused to be interviewed or breath-tested. It is argued that these results support the need for strategies which aim to reduce very high levels of intoxication on licensed premises in order to reduce alcohol-related accidents, injuries and offences.
Introduction common form of intervention is for bar staff to be Increasingly, licensed drinking settings have become trained in ways to identify early signs of intoxication a focus of efforts to reduce the harm associated with so that they may be better able to prevent intoxicasociety's use of alcohol (Saltz, 1988; Single, 1990). tion—or, failing that, to reduce the likelihood of The principal goal of these efforts is to reduce levels driving by intoxicated customers (Saltz, 1988; of intoxication; in panicular, the very high levels McKnight, 1988). Other programmes have involved that are often implicated in road traffic and increased enforcement of liquor licensing laws pedestrian accidents (McKnight, 1988; McLean et regarding the service of alcohol to intoxicated al., 1984; Alexander, Cave & Lyttle, 1990). A customers (Jeffs & Saunders, 1983) or have combined this with training programmes for barstaff Correspondence to: Dr Tim Stockwell, National Centre for
Research into the Prevention of Drug Abuse, Curtin Un.vers.ty, Bentley 6102, Western Australia. 873
,
u
i
o tvi n
i
imn\
C^'g" ' ^ " s h e r & Wallack, 1979). One of the first questions which presents itself to
874
Tim Stochvell et al.
the planners of such prevention programmes is how to choose which premises to target. Where resources are limited it is frequently impossible to train all staff and monitor serving practices in all licensed premises, particularly in urban areas. One solution to this problem has been to enquire routinely of all persons charged with drink driving offences as to where they were drinking prior to this event (e.g. Stockwell, Somerford & Lang, 1991). In the California DUI Project, premises which were cited three or more times by drink driving offenders were approached and offered training courses for their bar staff (Mosher & Wallack, 1979). More recently it has been reported that a similar system has been used in one health region of New Zealand as a means of administering the 1989 Sale of Liquor Act (Durham, 1991). In this instance if premises are mentioned on two or more occasions they are visited by a Medical Officer for Health in order to examine how responsibly alcohol is served at that establishment. It has been found that drivers who had their last drink on licensed as opposed to unlicensed premises are over-represented among drink-driving offenders in Western Australia (Lang, Stockwell & Lo, 1991). It has also been found that the extent to which individual premises are identified by drink-driving offenders correlates significantly with the incidence of assaults on the same premises and also with their annual purchases of alcohol (Stockwell £( a/., 1991). This fact may be taken to mean that premises with a high volume of sales are more likely to have customers involved in drink-driving and other offences simply because they have more customers. However, it has also been shown that if sales of alcohol are controlled for it is possible to identify premises which have a far greater likelihood of having customers being involved in both drinkdriving offences and in assaults (Stockwell et al., 1991). On the basis of these findings, we have suggested that it is possible to assess the degree of risk attached to different licensed premises by examining the ratio of measurable harm involving or affecting customers as a proportion of total onpremises alcohol purchases. The purpose of the study to be described here was to test the hypothesis that the level of risk attached to licensed premises as defined above would be related to the intoxication levels of their customers. If this hypothesis can be confirmed then it strengthens the case for targeting prevention initiatives at premises identified as high risk and also for focusing on the issue of intoxication levels on these premises.
There are a number of alternative hypotheses as to why some licensed premises might be identified as being at higher risk than others. One possible source of bias in the collection of these data might occur as a result of traffic police targeting certain establishments which are believed to be high risk for other reasons or because it is easier to place random breath testing patrols in their vicinity. As has been described elsewhere (Stockwell et al., 1991) this type of bias is eliminated when data on last place of drinking is only used from persons who fail a roadside breath test after having been involved in a road traffic accident. Since January 1989 it has been the practice of the traffic police in Perth to breath test all drivers in accidents they have been called to. In order to minimize such test-site selection biases, data derived from drivers involved in accidents were given extra weight in the present study when calculating the risk scores. It might be argued that another source of bias in these data is created by the fact that premises are identified on the basis of the last place of drinking rather than the place where most alcohol has been consumed. Thus, it has been shown that establishments with late night licences, such as nightclubs, have a far higher level of risk than those like hotels and taverns which close earlier (Stockwell, Somerford & Lang, 1992). This is likely to reflect the habit among many nightclub patrons in Perth of consuming copious amounts of cheaper alcohol at hotels early in the evening and then 'kicking on' to a nightclub where they are more sparing in their consumption of highly-priced drinks. Physical and geographical features of an establishment and its locale might increase the risk of its customers being involved in traffic accidents regardless of their intoxication levels. For example, the further an establishment is from the place of residence of its customers then the further they have to travel and the greater is the probability of their having an accident once they have been drinking. Other possible influences might be traffic density in the vicinity of an establishment and the extent to which service roads are designed for road safety. For the above reasons it was deemed important to discern whether intoxication levels might make an independent contribution towards risk status of licensed establishments. The feasibility of assessing blood alcohol levels from customers of licensed establishments as they leave the premises has been demonstrated in recent years both by Davey & McLean (1988) and Werch et al. (1988). The latter study reported an 80% compliance rate of customers
Licensed premises risk levels
875
randomly approached at the time of exiting from an establishment. The procedures they described were adopted for the present study.
Method The subjects Subjects were 307 patrons out of 414 approached (74.2%) who consented to participate in the study when approached as they exited licensed premises on Friday or Saturday evenings. Of these, 150 were exiting premises designated as High Risk and 157 were exiting Low Risk premises. The majority (76.0%) were male and aged under 26 years of age (55.3%). Figure 1 illustrates the age-sex characteristics of the sample. The gender mix of patrons who refused to participate (74.8% male) was very similar to that for the study sample.
Measurement of risk The 'risk' status of a licensed establishment in the study area was calculated by dividing a measure of 'harm' involving customers of the establishment by its purchases of alcohol in the preceding financial year. The indicators of alcohol-related harm used were as follows: (i) the number of traffic accidents known to have involved a driver who had failed a roadside breath-test after drinking on those premises during the financial year 1 July 1989-30 June 1990; (ii) the number of drink-driving offenders other than from traffic accidents who had failed a roadside breath-test after drinking on those premises during the above financial year. The Police Department of Western Australia provided data on all drivers in the study area who were given a confirmatory breath test at the Perth Traffic Police headquarters after failing a roadside test. The bulk of this group (approximately 80%) comprises individuals tested by routine traffic patrols, the remainder being individuals identified by Random Breath-Test patrols (10%) and drivers involved in traffic accidents to which the police were called (10%). In the latter instance, it is the policy of the WA Traffic Police that all such drivers are breathalysed. Individuals tested were routinely asked to state the place where they last consumed alcohol and also the name of the establishment if it was licensed. The indicator of annual sales was the total
Figure 1. Age and sex characteristics of sample, I female; D male. purchases of alcohol made by each establishment in the financial year 1989/1990. This information is supplied to the Liquor Licensing Division of the WA Office of Racing and Gaming each year and is the basis for calculating liquor licence fees. As has been described elsewhere (Stockwell ei ai, 1991) it was necessary to make adjustments for off-premise sales where both these and on-premise sales occurred under the same licence. A Risk Ratio was calculated using the following formula:
where PI = percentage of citations for an establishment out of all citations from accident cases from July 1989 to June 1990; PJ = percentage of citations for an establishment out of all citation from drivers failing a roadside breath test but who had not been involved in an accident for same time period; $ r = total purchases of alcohol for that establishment for same time period. This method of calculation gives equal weighting to accident and non-accident cases despite the greater volume of the latter.
Selection of licensed premises The 15 licensed premises selected were drawn from the total sample of 114 licensed hotels and taverns in the study area according to the following criteria: (i) Premises were categorized by the Liquor Licensing Division as either a 'hotel' or 'tavern' and
876
Tim Stockwell et al.
as having a postal address falling within the operational area of the Perth Traffic Branch. (ii) Annual purchases of alcohol for the financial year 1989/1990 as notified by the Liquor Licensing Division were between AUD$ 150,000 and AUD$600,000. This was to ensure that they were sufficiently busy to permit a steady flow of customers without being atypically large. (iii) Premises had large car-parks so that it was easy to observe whether customers drove home. (One establishment in each group had no car park but both were included since they fulfilled all other criteria). (iv) Managers gave their permission for the study to take place. (v) Efforts were also made to ensure that the mean distance from Perth city centre and also the amount of alcohol purchased in the immediately preceding financial year were similar for High and Low Risk premises. (vi) Hotels which accommodated more than 50 guests were excluded as being primarily residential establishments.
patrons exiting the pub in the 4-hour period between the hours of 8 p.m. and midnight were asked to participate in the survey. The research team selected the two busiest exits from the venue for sampling. Available data (Lang et ai, 1990) suggest that these times of day and days of the week are the highest risk periods for alcohol-related traffic accidents and drink-driving offences in Perth. The interview team was comprised of two female interviewers and three male assistants using the approach described by Werch el ai (1988). On exiting the pub, patrons were approached by a female interviewer and asked if they would like to take part in a survey for Curtin University and were offered a free breath test. After completing an interview, the interviewer selected the next patron crossing the exit threshold. When patrons emerged as a group, the patron whose surname began with the letter closest to 'A' was selected for the interview. If they refused then the member of the group whose surname began with the letter next closest to 'A' was asked to take part in the interview.
Table 1 illustrates how the High and Low Risk premises were very similar in terms of mean purchases of alcohol and mean distance from Perth city centre, but markedly different with regard to accidents and drink-driving offences involving their customers.
Each interviewer was accompanied by a male assistant who made an independent rating of the patron's level of intoxication before conducting the breath analysis. The interviewers also made these ratings and an analysis of the relationship between these and blood alcohol levels will be the subject of a further paper. A fifth member of the research team acted as an unobtrusive observer of subjects' subsequent driving behaviour. Patrons were eligible to take part in the study if they were leaving the hotel for the evening and had consumed at least one drink, alcoholic or nonalcoholic, on the premises. For those who refused, estimates of patrons' intoxication levels were recorded by the assistants.
Procedure Premises were approached on the day prior to data collection and asked for their consent. Both licensees and the interview team were blind as to the risk score ranking of the pubs selected for study. Each of the 15 pubs were visited once, on either a Friday or Saturday night. High and Low Risk premises were balanced for night of the week. A random sample of
Table 1. A comparison of low (n = 8) and high (a = 7} risk premises on annual purchases of alcohol, distance from city centre, traffic accidents and drink-driving offences involving patrons during year prior to study Low risk Variable Annual purchases ($10,000) Distance (km) Accident DDO's Other DDO's
High risk
X
SD
3c
SD
34.0 6.0 0.0 4.6
12.5 3.4 0.0 1.8
39.8 6.9 3.0 29.1
12.1 2.7 0.8* 8.4*
* Difference between Low and High Risk significant at />< 0.0001, Student (-test. Differences not significant for other variables.
Licensed premises risk levels
The interview was designed to take at least 10 minutes to allow sufficient time for any alcohol in the mouth remaining after the last drink to be absorbed before taking a blood alcohol level (BAL) reading. Interviewers and assistants were instructed to pace the interview with this in mind. Patrons were also asked not to smoke during the interview as this may also artificially inflate alcometer readings. As an additional precaution patrons were also given a mouthwash of mineral water. After completing the interview patrons were asked to guess their BAL and then undertook a breath analysis. The patron was then advised of their legal status with respect to driving, asked how they intended to get home and, if over the limit, advised not to drive. The observer recorded how all patrons over 0.05 left the venue. All procedures were piloted on a sample of 52 patrons leaving two premises in a locality adjacent to the study area.
The patron interview Patrons were asked questions concerning: selfperceived level of intoxication; fitness and intentions to drive; estimated BAL and number of drinks consumed on that evening. Basic sociodemographic information concerning age, gender, occupation and place of residence was also collected.
Analyses A series of comparisons were made between patrons from the High and Low Risk premises on variables of interest. The significance of all differences obtained was tested by two-sample i-tests for continuous data, employing logarithmic transformations where distributions were highly skewed, and by x^ tests where variables were categorical. Yates corrected x^ where one-tailed significance tests were justified. Non-parametric correlations were also calculated between average BAL's, proportion of persons with BAL's in excess of 0.15 mg/ml and Risk Ratios across individual premises. One-tailed tests were used for all comparisons involving levels of drinking and intoxication.
Results Comparison of patrons who refused to participate Significantly more patrons of High Risk establishments refused to participate in the study {x^ = 4.49, df = 1, p < 0.05) and a significantly larger proportion of these refusers were rated by the interviewers as being 'moderately' or 'extremely' intoxicated than were refusers from Low Risk premises (/^ = 9.93, df = I. p < 0.005; see Table 2). The proportion of men and women refusing to participate in the study was very similar in the two Risk groups. Table 2. A comparison of patrons refusing co participate in the study from High and Low risk premises Low risk
Measures
Estimates of patrons' BAL's were obtained using the Drager Alcotest 7410 (Drager Laboratories, Germany). At the time of testing, the model superceded those which were in use by the Western Australian Traffic Police and was considered the most accurate and reliable of the available hand-held breathalyser units. Available technical data indicate that using an ethanol standard, measurement accuracy for BAL's below 0.1 mg/ml is within 0.005 mg/ml and above 0.1 mg/ml is within 0.05 mg/ml. The units were calibrated before starting data collection and have a recommended recalibration interval of 6 months. Readings provided by the unit were in the form of a three-digit display in % BAL. Ratings of patron levels of intoxication were conducted by both interviewers and assistants using a four-point scale comprising 'not at all', 'slightly', 'moderately' and 'extremely' intoxicated and assessed on the basis of visual signs.
877
High risk
Variable
(«)
(%)
(«)
(%)
Sex Male Female
31 11
73.8 26.2
49 16
75.4 24.6
Total
42
100.0
65
100.0
8 34
19.0 81.0
32 33
49.2 50.8
42
100.0
65
100.0
Intoxication* Mod/extreme None/slight Total
/^ = 9.93, p < 0.005.
Demographic
data
Comparisons between the two groups of patrons are shown in Table 3 for sex, age and occupation. Only the latter variable distinguished between the two groups to any significant degree (/^ = 27.92, ^ < 0.001). Individual x^ tests confirm that there
878
Tim Stockwell et al.
Table 3. A comparison of low (a = 157) and high (a = ISO) risk patrons on demographic variables
Low risk Variable
High risk
(n)
(%)
(«)
(%)
F
115 40
74.2 25.8
117 33
78.0 22.0
Total
155
100.0
150
100.0
0 90 42 14 9
0.0 58.1 27.1 9.0 5.8
0 78 55 9 7
0.0 52.3 36.9 6.0 4.7
155
100.0
149
100.0
Sex
M
Age
Under 18 18-25 26-35 36-45 46 + Total Occupation' Prof & Admin White Collar Blue Collar Student Other Total
25 30 45 36 11
14.7 21.0 31.5 25.2 1.1
19 41 62 6 17
13.1 28.3 42.8 4.1 11.7
143
100.0
145
100.0
N.B. There were missing data dne to refusals and coding errors so totals are not the same as those for the full sample. * Overall y^ = 27.92, p < 0.001. ** y} = 3.93, p < 0.05; '** x^ = 25.56, p < 0.001. were significantly more persons with 'Blue Collar' occupations (i.e. manual and trades) and fewer students leaving the High Risk Premises. Blood alcohol levels (BAL 's) The distribution of BALs shown in Fig. 2 across the categories created by the two legally defined BALs for traffic offences in Western Australia (0.08 and 0.15) was also significantly different between the two groups (x^ = 7.27, df = 2, p < 0.05). An individual test of the threefold difference between the groups with regard to numbers of patrons over 0.15 was also significant (Yates corrected /^ = 6.03, The High Risk patrons had a slightly higher average BAL than did Low Risk patrons though this difference was not significant (see Table 4). BAL's and Risk Ratios Spearman rank correlation coefficients were calculated between the Risk Ratios of each of the 15 premises in the study and the'proportions of patrons over 0.08 and 0.15 of all those interviewed at those
0.08 to 0.149
0.15 plus
Blood alcohol levels
Figure 2. BAL's of Low and High Risk patrons, D low risk; D high risk. premises. A significant correlation emerged for the proportion of patrons over 0.15 (r = 0.63,p
RESEARCH REPORT
Levels of drunkenness of eustomers leaving licensed premises in Perth, Western Australia: a comparison of high and low ^risk' premises TIM STOCKWELL,' PHIL RYDON,' SONIA GIANATTI,' EVAN JENKINS,^ CLAUDIA OVENDEN> & DAVID SYED^ ^National Centre for Research into the Prevention of Drug Abuse, Curtin University of Technology, Bentley 6102 & '^Medical School, University of Western Australia, Nedlands 6009, Western Australia
Abstract A measure of the risk of licensed premises having customers involved in road traffic accidents and drinkdriving offences was utilised in order to identify seven 'High Risk'and eight 'Low Risk'premises in metropolitan Perth, Western Australia. This measure, or 'Risk Ratio', was defined as the ratio of incidents of alcohol-related harm to an estimate of on-premises alcohol sales for a particular establishment. A study was conducted to test the hypothesis that a High Risk status would be associated with greater levels of customer intoxication. Interviews concerning drinking behaviour and breathalyser readings were collected from 74.2% of 414 customers exiting from the chosen premises between 8 p.m. and midnight on Friday and Saturday nights. High Risk premises had three times more customers whose readings were in excess of 0.15 mg/ml ('p < O.OI). The proportion of customers with BAL's above 0.15 correlated strongly with the premises' Risk Ratio (i = 0.63, p < 0.01). There were also significantly more patrons from High than from Low Risk establishments who were rated as appearing moderately or severely intoxicated but refused to be interviewed or breath-tested. It is argued that these results support the need for strategies which aim to reduce very high levels of intoxication on licensed premises in order to reduce alcohol-related accidents, injuries and offences.
Introduction common form of intervention is for bar staff to be Increasingly, licensed drinking settings have become trained in ways to identify early signs of intoxication a focus of efforts to reduce the harm associated with so that they may be better able to prevent intoxicasociety's use of alcohol (Saltz, 1988; Single, 1990). tion—or, failing that, to reduce the likelihood of The principal goal of these efforts is to reduce levels driving by intoxicated customers (Saltz, 1988; of intoxication; in panicular, the very high levels McKnight, 1988). Other programmes have involved that are often implicated in road traffic and increased enforcement of liquor licensing laws pedestrian accidents (McKnight, 1988; McLean et regarding the service of alcohol to intoxicated al., 1984; Alexander, Cave & Lyttle, 1990). A customers (Jeffs & Saunders, 1983) or have combined this with training programmes for barstaff Correspondence to: Dr Tim Stockwell, National Centre for
Research into the Prevention of Drug Abuse, Curtin Un.vers.ty, Bentley 6102, Western Australia. 873
,
u
i
o tvi n
i
imn\
C^'g" ' ^ " s h e r & Wallack, 1979). One of the first questions which presents itself to
874
Tim Stochvell et al.
the planners of such prevention programmes is how to choose which premises to target. Where resources are limited it is frequently impossible to train all staff and monitor serving practices in all licensed premises, particularly in urban areas. One solution to this problem has been to enquire routinely of all persons charged with drink driving offences as to where they were drinking prior to this event (e.g. Stockwell, Somerford & Lang, 1991). In the California DUI Project, premises which were cited three or more times by drink driving offenders were approached and offered training courses for their bar staff (Mosher & Wallack, 1979). More recently it has been reported that a similar system has been used in one health region of New Zealand as a means of administering the 1989 Sale of Liquor Act (Durham, 1991). In this instance if premises are mentioned on two or more occasions they are visited by a Medical Officer for Health in order to examine how responsibly alcohol is served at that establishment. It has been found that drivers who had their last drink on licensed as opposed to unlicensed premises are over-represented among drink-driving offenders in Western Australia (Lang, Stockwell & Lo, 1991). It has also been found that the extent to which individual premises are identified by drink-driving offenders correlates significantly with the incidence of assaults on the same premises and also with their annual purchases of alcohol (Stockwell £( a/., 1991). This fact may be taken to mean that premises with a high volume of sales are more likely to have customers involved in drink-driving and other offences simply because they have more customers. However, it has also been shown that if sales of alcohol are controlled for it is possible to identify premises which have a far greater likelihood of having customers being involved in both drinkdriving offences and in assaults (Stockwell et al., 1991). On the basis of these findings, we have suggested that it is possible to assess the degree of risk attached to different licensed premises by examining the ratio of measurable harm involving or affecting customers as a proportion of total onpremises alcohol purchases. The purpose of the study to be described here was to test the hypothesis that the level of risk attached to licensed premises as defined above would be related to the intoxication levels of their customers. If this hypothesis can be confirmed then it strengthens the case for targeting prevention initiatives at premises identified as high risk and also for focusing on the issue of intoxication levels on these premises.
There are a number of alternative hypotheses as to why some licensed premises might be identified as being at higher risk than others. One possible source of bias in the collection of these data might occur as a result of traffic police targeting certain establishments which are believed to be high risk for other reasons or because it is easier to place random breath testing patrols in their vicinity. As has been described elsewhere (Stockwell et al., 1991) this type of bias is eliminated when data on last place of drinking is only used from persons who fail a roadside breath test after having been involved in a road traffic accident. Since January 1989 it has been the practice of the traffic police in Perth to breath test all drivers in accidents they have been called to. In order to minimize such test-site selection biases, data derived from drivers involved in accidents were given extra weight in the present study when calculating the risk scores. It might be argued that another source of bias in these data is created by the fact that premises are identified on the basis of the last place of drinking rather than the place where most alcohol has been consumed. Thus, it has been shown that establishments with late night licences, such as nightclubs, have a far higher level of risk than those like hotels and taverns which close earlier (Stockwell, Somerford & Lang, 1992). This is likely to reflect the habit among many nightclub patrons in Perth of consuming copious amounts of cheaper alcohol at hotels early in the evening and then 'kicking on' to a nightclub where they are more sparing in their consumption of highly-priced drinks. Physical and geographical features of an establishment and its locale might increase the risk of its customers being involved in traffic accidents regardless of their intoxication levels. For example, the further an establishment is from the place of residence of its customers then the further they have to travel and the greater is the probability of their having an accident once they have been drinking. Other possible influences might be traffic density in the vicinity of an establishment and the extent to which service roads are designed for road safety. For the above reasons it was deemed important to discern whether intoxication levels might make an independent contribution towards risk status of licensed establishments. The feasibility of assessing blood alcohol levels from customers of licensed establishments as they leave the premises has been demonstrated in recent years both by Davey & McLean (1988) and Werch et al. (1988). The latter study reported an 80% compliance rate of customers
Licensed premises risk levels
875
randomly approached at the time of exiting from an establishment. The procedures they described were adopted for the present study.
Method The subjects Subjects were 307 patrons out of 414 approached (74.2%) who consented to participate in the study when approached as they exited licensed premises on Friday or Saturday evenings. Of these, 150 were exiting premises designated as High Risk and 157 were exiting Low Risk premises. The majority (76.0%) were male and aged under 26 years of age (55.3%). Figure 1 illustrates the age-sex characteristics of the sample. The gender mix of patrons who refused to participate (74.8% male) was very similar to that for the study sample.
Measurement of risk The 'risk' status of a licensed establishment in the study area was calculated by dividing a measure of 'harm' involving customers of the establishment by its purchases of alcohol in the preceding financial year. The indicators of alcohol-related harm used were as follows: (i) the number of traffic accidents known to have involved a driver who had failed a roadside breath-test after drinking on those premises during the financial year 1 July 1989-30 June 1990; (ii) the number of drink-driving offenders other than from traffic accidents who had failed a roadside breath-test after drinking on those premises during the above financial year. The Police Department of Western Australia provided data on all drivers in the study area who were given a confirmatory breath test at the Perth Traffic Police headquarters after failing a roadside test. The bulk of this group (approximately 80%) comprises individuals tested by routine traffic patrols, the remainder being individuals identified by Random Breath-Test patrols (10%) and drivers involved in traffic accidents to which the police were called (10%). In the latter instance, it is the policy of the WA Traffic Police that all such drivers are breathalysed. Individuals tested were routinely asked to state the place where they last consumed alcohol and also the name of the establishment if it was licensed. The indicator of annual sales was the total
Figure 1. Age and sex characteristics of sample, I female; D male. purchases of alcohol made by each establishment in the financial year 1989/1990. This information is supplied to the Liquor Licensing Division of the WA Office of Racing and Gaming each year and is the basis for calculating liquor licence fees. As has been described elsewhere (Stockwell ei ai, 1991) it was necessary to make adjustments for off-premise sales where both these and on-premise sales occurred under the same licence. A Risk Ratio was calculated using the following formula:
where PI = percentage of citations for an establishment out of all citations from accident cases from July 1989 to June 1990; PJ = percentage of citations for an establishment out of all citation from drivers failing a roadside breath test but who had not been involved in an accident for same time period; $ r = total purchases of alcohol for that establishment for same time period. This method of calculation gives equal weighting to accident and non-accident cases despite the greater volume of the latter.
Selection of licensed premises The 15 licensed premises selected were drawn from the total sample of 114 licensed hotels and taverns in the study area according to the following criteria: (i) Premises were categorized by the Liquor Licensing Division as either a 'hotel' or 'tavern' and
876
Tim Stockwell et al.
as having a postal address falling within the operational area of the Perth Traffic Branch. (ii) Annual purchases of alcohol for the financial year 1989/1990 as notified by the Liquor Licensing Division were between AUD$ 150,000 and AUD$600,000. This was to ensure that they were sufficiently busy to permit a steady flow of customers without being atypically large. (iii) Premises had large car-parks so that it was easy to observe whether customers drove home. (One establishment in each group had no car park but both were included since they fulfilled all other criteria). (iv) Managers gave their permission for the study to take place. (v) Efforts were also made to ensure that the mean distance from Perth city centre and also the amount of alcohol purchased in the immediately preceding financial year were similar for High and Low Risk premises. (vi) Hotels which accommodated more than 50 guests were excluded as being primarily residential establishments.
patrons exiting the pub in the 4-hour period between the hours of 8 p.m. and midnight were asked to participate in the survey. The research team selected the two busiest exits from the venue for sampling. Available data (Lang et ai, 1990) suggest that these times of day and days of the week are the highest risk periods for alcohol-related traffic accidents and drink-driving offences in Perth. The interview team was comprised of two female interviewers and three male assistants using the approach described by Werch el ai (1988). On exiting the pub, patrons were approached by a female interviewer and asked if they would like to take part in a survey for Curtin University and were offered a free breath test. After completing an interview, the interviewer selected the next patron crossing the exit threshold. When patrons emerged as a group, the patron whose surname began with the letter closest to 'A' was selected for the interview. If they refused then the member of the group whose surname began with the letter next closest to 'A' was asked to take part in the interview.
Table 1 illustrates how the High and Low Risk premises were very similar in terms of mean purchases of alcohol and mean distance from Perth city centre, but markedly different with regard to accidents and drink-driving offences involving their customers.
Each interviewer was accompanied by a male assistant who made an independent rating of the patron's level of intoxication before conducting the breath analysis. The interviewers also made these ratings and an analysis of the relationship between these and blood alcohol levels will be the subject of a further paper. A fifth member of the research team acted as an unobtrusive observer of subjects' subsequent driving behaviour. Patrons were eligible to take part in the study if they were leaving the hotel for the evening and had consumed at least one drink, alcoholic or nonalcoholic, on the premises. For those who refused, estimates of patrons' intoxication levels were recorded by the assistants.
Procedure Premises were approached on the day prior to data collection and asked for their consent. Both licensees and the interview team were blind as to the risk score ranking of the pubs selected for study. Each of the 15 pubs were visited once, on either a Friday or Saturday night. High and Low Risk premises were balanced for night of the week. A random sample of
Table 1. A comparison of low (n = 8) and high (a = 7} risk premises on annual purchases of alcohol, distance from city centre, traffic accidents and drink-driving offences involving patrons during year prior to study Low risk Variable Annual purchases ($10,000) Distance (km) Accident DDO's Other DDO's
High risk
X
SD
3c
SD
34.0 6.0 0.0 4.6
12.5 3.4 0.0 1.8
39.8 6.9 3.0 29.1
12.1 2.7 0.8* 8.4*
* Difference between Low and High Risk significant at />< 0.0001, Student (-test. Differences not significant for other variables.
Licensed premises risk levels
The interview was designed to take at least 10 minutes to allow sufficient time for any alcohol in the mouth remaining after the last drink to be absorbed before taking a blood alcohol level (BAL) reading. Interviewers and assistants were instructed to pace the interview with this in mind. Patrons were also asked not to smoke during the interview as this may also artificially inflate alcometer readings. As an additional precaution patrons were also given a mouthwash of mineral water. After completing the interview patrons were asked to guess their BAL and then undertook a breath analysis. The patron was then advised of their legal status with respect to driving, asked how they intended to get home and, if over the limit, advised not to drive. The observer recorded how all patrons over 0.05 left the venue. All procedures were piloted on a sample of 52 patrons leaving two premises in a locality adjacent to the study area.
The patron interview Patrons were asked questions concerning: selfperceived level of intoxication; fitness and intentions to drive; estimated BAL and number of drinks consumed on that evening. Basic sociodemographic information concerning age, gender, occupation and place of residence was also collected.
Analyses A series of comparisons were made between patrons from the High and Low Risk premises on variables of interest. The significance of all differences obtained was tested by two-sample i-tests for continuous data, employing logarithmic transformations where distributions were highly skewed, and by x^ tests where variables were categorical. Yates corrected x^ where one-tailed significance tests were justified. Non-parametric correlations were also calculated between average BAL's, proportion of persons with BAL's in excess of 0.15 mg/ml and Risk Ratios across individual premises. One-tailed tests were used for all comparisons involving levels of drinking and intoxication.
Results Comparison of patrons who refused to participate Significantly more patrons of High Risk establishments refused to participate in the study {x^ = 4.49, df = 1, p < 0.05) and a significantly larger proportion of these refusers were rated by the interviewers as being 'moderately' or 'extremely' intoxicated than were refusers from Low Risk premises (/^ = 9.93, df = I. p < 0.005; see Table 2). The proportion of men and women refusing to participate in the study was very similar in the two Risk groups. Table 2. A comparison of patrons refusing co participate in the study from High and Low risk premises Low risk
Measures
Estimates of patrons' BAL's were obtained using the Drager Alcotest 7410 (Drager Laboratories, Germany). At the time of testing, the model superceded those which were in use by the Western Australian Traffic Police and was considered the most accurate and reliable of the available hand-held breathalyser units. Available technical data indicate that using an ethanol standard, measurement accuracy for BAL's below 0.1 mg/ml is within 0.005 mg/ml and above 0.1 mg/ml is within 0.05 mg/ml. The units were calibrated before starting data collection and have a recommended recalibration interval of 6 months. Readings provided by the unit were in the form of a three-digit display in % BAL. Ratings of patron levels of intoxication were conducted by both interviewers and assistants using a four-point scale comprising 'not at all', 'slightly', 'moderately' and 'extremely' intoxicated and assessed on the basis of visual signs.
877
High risk
Variable
(«)
(%)
(«)
(%)
Sex Male Female
31 11
73.8 26.2
49 16
75.4 24.6
Total
42
100.0
65
100.0
8 34
19.0 81.0
32 33
49.2 50.8
42
100.0
65
100.0
Intoxication* Mod/extreme None/slight Total
/^ = 9.93, p < 0.005.
Demographic
data
Comparisons between the two groups of patrons are shown in Table 3 for sex, age and occupation. Only the latter variable distinguished between the two groups to any significant degree (/^ = 27.92, ^ < 0.001). Individual x^ tests confirm that there
878
Tim Stockwell et al.
Table 3. A comparison of low (a = 157) and high (a = ISO) risk patrons on demographic variables
Low risk Variable
High risk
(n)
(%)
(«)
(%)
F
115 40
74.2 25.8
117 33
78.0 22.0
Total
155
100.0
150
100.0
0 90 42 14 9
0.0 58.1 27.1 9.0 5.8
0 78 55 9 7
0.0 52.3 36.9 6.0 4.7
155
100.0
149
100.0
Sex
M
Age
Under 18 18-25 26-35 36-45 46 + Total Occupation' Prof & Admin White Collar Blue Collar Student Other Total
25 30 45 36 11
14.7 21.0 31.5 25.2 1.1
19 41 62 6 17
13.1 28.3 42.8 4.1 11.7
143
100.0
145
100.0
N.B. There were missing data dne to refusals and coding errors so totals are not the same as those for the full sample. * Overall y^ = 27.92, p < 0.001. ** y} = 3.93, p < 0.05; '** x^ = 25.56, p < 0.001. were significantly more persons with 'Blue Collar' occupations (i.e. manual and trades) and fewer students leaving the High Risk Premises. Blood alcohol levels (BAL 's) The distribution of BALs shown in Fig. 2 across the categories created by the two legally defined BALs for traffic offences in Western Australia (0.08 and 0.15) was also significantly different between the two groups (x^ = 7.27, df = 2, p < 0.05). An individual test of the threefold difference between the groups with regard to numbers of patrons over 0.15 was also significant (Yates corrected /^ = 6.03, The High Risk patrons had a slightly higher average BAL than did Low Risk patrons though this difference was not significant (see Table 4). BAL's and Risk Ratios Spearman rank correlation coefficients were calculated between the Risk Ratios of each of the 15 premises in the study and the'proportions of patrons over 0.08 and 0.15 of all those interviewed at those
0.08 to 0.149
0.15 plus
Blood alcohol levels
Figure 2. BAL's of Low and High Risk patrons, D low risk; D high risk. premises. A significant correlation emerged for the proportion of patrons over 0.15 (r = 0.63,p
