Drink driving and traffic accidents in young people

Drink driving and traffic accidents in young people

Accident Analysis and Prevention 32 (2000) 805 – 814 www.elsevier.com/locate/aap Drink driving and traffic accidents in young people L. John Horwood,...

90KB Sizes 0 Downloads 2 Views

Accident Analysis and Prevention 32 (2000) 805 – 814 www.elsevier.com/locate/aap

Drink driving and traffic accidents in young people L. John Horwood, David M. Fergusson * Department of Psychological Medicine, Christchurch Health and De6elopment Study, Christchurch School of Medicine, PO Box 4345, Christchurch, New Zealand Received 22 August 1999; received in revised form 22 November 1999; accepted 29 November 1999

Abstract The relationship between drink driving behaviours and rates of traffic accidents was analysed in a birth cohort of 907 New Zealand young people studied to the age of 21. Drink driving was significantly (P B0.0001) related to active traffic accidents in which the driver’s behaviour contributed to the accident but was not related to passive accidents in which driver behaviours did not contribute to the accident (P \0.15). Those engaging in high rates of drink driving had rates of active traffic accidents that were 2.6 times higher than those who did not drink and drive. Further analysis suggested that much of this association was explained by confounding factors (and notably driver behaviour) that were associated with both drink driving and accident rates. After adjustment for confounding factors, those engaging in high rates of drink driving had rates of active accidents that were 1.5 (PB 0.01) times higher than those who did not drink and drive. It is concluded that although the study findings support the view that the regulation of drink driving behaviour amongst young people is likely to contribute to a reduction in traffic accidents, to be fully effective attempts at regulation of drink driving also need to be accompanied by a similar level of investment in regulating other aspects of risky or illegal driving behaviour amongst young people. © 2000 Elsevier Science Ltd. All rights reserved. Keywords: Drink driving; Traffic accidents; Longitudinal study; Driver behaviours

1. Introduction It is widely accepted that drink driving behaviours make significant contributions to driver risks and are associated with increased rates of risky driving behaviour, motor vehicle accidents and the associated mortality and morbidity from these accidents (Evans, 1991; Baker et al., 1992). The role that alcohol use is believed to play in accident risk is encapsulated in the legislation of many societies that have imposed both legal restrictions on the amount of alcohol that may be present in the blood of drivers and applied heavy penalties for drinking driving behaviours. This legislation has been underwritten by research findings of motor vehicle accidents and fatalities that have revealed elevated blood alcohol levels amongst substantial proportions of those involved in serious traffic accidents (Bailey, 1987; US Department of Health and Human Services, 1988; Stoduto et al., 1993) and is further * Corresponding author. Tel.: +64-3-3720406; fax: +64-33720405. E-mail address: [email protected] (D.M. Fergusson).

supported by laboratory studies showing that alcohol consumption leads to decreased reaction time, poorer eye–hand co-ordination and impaired driver performance (Moskowitz et al., 1985; West et al., 1993; Deery and Love, 1996). Despite these lines of evidence and the strong beliefs that drink driving behaviours make a substantial contribution to driver risks, this issue remains a source of contention in the literature on driver behaviours. In particular, a number of authors have pointed to the fact that the apparent associations between drink driving behaviours and driver risk may reflect the fact that those individuals who engage in drink driving behaviours are characterised by other behavioural tendencies that quite independently of alcohol consumption, place them at increased risk of both motor vehicle accidents and unsafe driving practices (Hedlund, 1994; West, 1995). More specifically, it has been well documented that drink driving behaviours and accident risk are frequently associated with other tendencies to risk taking and antisocial behaviour patterns that may include: risky driving behaviours; involvement in crime; illicit substance abuse; antisocial and oppositional be-

0001-4575/00/$ - see front matter © 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 1 - 4 5 7 5 ( 0 0 ) 0 0 0 0 5 - 1

806

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

haviours; and associated behavioural patterns (Donovan et al., 1985; Stacy et al., 1991; Beirness, 1993; West, 1996). Given these associations, it could be argued that the elevated driver risks found amongst those who drink and drive do not reflect the effects of alcohol consumption on driver risk but rather arise because those who engage in drink driving behaviours are also predisposed to risk taking and deviant behaviour patterns that increase their driver risk. More specifically it may be suggested that drink driving is one of a constellation of risky driving behaviours that may include speeding, unsafe and careless driving and that the higher rate of accidents amongst drink drivers reflects their general tendency to risky driving rather than the specific effects of alcohol on driver performance. There is some evidence to support this view from two studies reported by West (1995) examining the linkages between deviant driving behaviours, drink driving and traffic accident risk. The first of these studies examined these associations in a sample of novice drivers who were aged under 20 years and had passed their driving test in the 6 months preceding the study. The second involved a sample of older drivers (mean age 41) who had held their driving licences for at least 3 years. In both studies, the frequency of deviant driving behaviour was found to be a predictor of accident risk but alcohol use was found to be related to accident risk in the novice driver group only. On the basis of these findings West concludes this study has supported the view that in novice drivers the statistical link between drink driving and accidents is not attributable to a more general recklessness or carelessness on the part of drivers who happen to drive after drinking. It also suggests that the link between drink driving and accidents is weaker in the case of experienced drivers than with novice drivers (p. 461). One of the best methods for investigating the linkages between drink driving behaviours and driver risk is through longitudinal research designs which examine the ways in which driver risks vary with changing patterns of drink driving behaviour and other potentially confounding factors. Such a design opens the way to estimate the associations between drink driving behaviours and driver risk, taking into account factors that may be associated with both the development of drink driving and the development of other aspects of risky driving behaviours. In this paper, we report on a longitudinal study of the relationships between drink driving behaviour and traffic accident risk in a birth cohort of New Zealand young people who have been studied to the age of 21. The aims of this study were to address the following issues:

1. To what extent was participation in drink driving behaviour amongst cohort members associated with increased rates of traffic accidents? 2. Did any association between drink driving behaviours and traffic accident risk persist when due allowance was made for potentially confounding factors (and notably deviant driving behaviours) that may be correlated with both drink driving behaviours and traffic accident risk? 3. Did the association between drink driving behaviours and traffic accidents after control for confounding factors vary with gender, age and duration of driving experience? More generally, this study seeks to use data gathered over the course of a longitudinal study to examine the extent to which drink driving behaviours made independent contributions to traffic accident risks and the extent to which these associations varied with driver characteristics.

2. Methods The data described in this paper were gathered during the course of the Christchurch Health and Development Study (CHDS). In this project an unselected birth cohort of 1265 children born in the Christchurch (New Zealand) urban region during a 4 month period in mid-1977 has been studied at birth, 4 months, 1 year and annual intervals to age 16 years, 18 and 21 years, using information gathered from a combination of sources including: parental interviews; child interview; psychometric testing; teacher report; medical, police and other records. At age 21, 1011 of the original 1265 cohort members were assessed. This sample represented 80% of the original cohort and 90% of the cohort members who were still alive and resident in New Zealand at age 21. The present analysis is based on the 907 sample members interviewed at age 21 who reported driving a motor vehicle during the period from age 18–21 years.

2.1. Accident risk, drink dri6ing and related factors As part of the assessment at age 21, sample members were questioned in detail concerning their driving experience, attitudes and behaviours, accident involvement and driving violations over the period from age 18 to 21 years. This questioning formed the basis of the following measures used in the present analysis.

2.1.1. Drink dri6ing beha6iours The extent of the young person’s involvement in drink driving behaviour was assessed using a six item scale in which the young person reported the frequency with which they had engaged in the following drink

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

driving behaviours: driving a motor vehicle within 4 h of drinking alcohol; driving when drunk or over the legal alcohol limit; driving when seriously intoxicated; being stopped or arrested for driving while over the legal alcohol limit; continuing to drive after drinking alcohol despite previous arrest(s) for drink driving offences; driving after drinking alcohol even though friends or relatives tried to prevent this. Each item was scored on a six point scale ranging from 0= never to 5= 21+occasions. Subjects were asked to report the frequency of each behaviour in each of the 3 years 18 – 19, 19–20, 20–21, and the six items were summed to produce an overall measure of the extent of drink driving behaviour in each year. The resulting scale scores were of moderate reliability, with alpha coefficients ranging from 0.70 to 0.78. However, the correlations between the drink driving scores over time ranged between 0.69 and 0.84, suggesting substantial stability in reported drink driving behaviours from 18 to 21 years.

2.1.2. Accident in6ol6ement Subjects were questioned, at age 21, concerning their involvement in motor vehicle accidents in each year from age 18–21 years. Accidents were defined to include all incidents where a motor vehicle being driven by the subject was involved in a collision with another vehicle, object, person or animal or where the individual seriously lost control of the vehicle, irrespective of damage or injury. For each accident reported, the young person was asked for a detailed description of the incident including: the nature of the accident; how the accident occurred; what damage or injury resulted from the accident; whether the young person had been drinking alcohol or using illicit drugs within 12 h of the accident and how much; whether the police became involved. Based on the subject’s description of each incident, accidents were classified as either ‘active’ or ‘passive’, using a procedure similar to that described by West (1993). Active accidents were defined to be those which resulted primarily from the subject’s driving behaviours and for which the subject could be held responsible in law. Passive accidents were those which resulted primarily from other drivers’ behaviours or from totally unexpected circumstances (e.g. animal ran in front of car). Ratings of accident type were made by two raters on the basis of the narrative material provided by subjects. There was better than 95% agreement between these raters in their assignment of incidents to active or passive accidents. The 907 subjects reported a total of 624 accidents (477 active, 147 passive) over the 3 year period. In 62 (13.0%) of the active accidents, the subject reported having drunk alcohol within 12 h of the accident and in at least half of these cases, the reported level of alcohol consumed would have resulted in the subject’s breath alcohol level being over the legal

807

limit at the time of the accident. In only four cases (2.3%) did the subject report drinking prior to a passive accident: in all cases, the reported level of alcohol consumption was low. Overall, 52 subjects reported being arrested for a drink driving offence during the 3 year period, and in ten cases the arrest resulted from an active accident involvement. Consideration was given to further classifying accidents into injury and non injury accidents. However, of the 624 accidents reported, only 49 (7.9%) involved injury: one of these resulted in the death of a passenger and 16 involved hospitalisation of the subject or another party. These numbers were insufficient to enable detailed analysis of injury accident risk.

2.1.3. Dri6er beha6iours The extent of the young person’s involvement in risky or illegal driving behaviours in each year from age 18–21 was assessed, from reports made at age 21, using an instrument based on the violations subscale of the Driver Behaviour Questionnaire described by Reason et al. (1991), but modified to reflect New Zealand conditions. This instrument recorded the frequency with which young people reported committing a series of 12 driving violations including: exceeding the speed limit by more than 20 kph; driving without a seat belt; deliberately driving through red lights; street racing; driving without a licence; driving when the licence had been suspended; driving without a current vehicle registration; driving without a current vehicle warrant of fitness; changing lanes without signalling; overtaking without a clear view of the road ahead; overtaking illegally; and driving too close to other vehicles. Responses were graded on a four-point scale ranging from 0= never to 3=nearly every day. Subjects’ responses were summed across the 12 items to produce a total driving behaviour score reflecting the extent of involvement in risky or illegal driver behaviours in each of the 3 years. The reliabilities of these scales, assessed using coefficient a, ranged from 0.78 to 0.82. Reported driver behaviour scores were very stable over time, with across time correlations that ranged from 0.85 to 0.94. 2.1.4. Dri6er attitudes Attitudes to driving practices were assessed at age 21 using the Attitudes to Driving Violations Scale (West and Hall, 1997). This scale rates the extent to which subjects agree with a series of seven items regarding traffic violations (e.g. decreasing the speed limit on motorways is a good idea, penalties for speeding should be more severe). Ratings were made on a 5-point scale ranging from 1= strongly agree to 5= strongly disagree, and a total score was computed from a sum of the seven items. This score ranged from 5 to 35 with a high score indicating a laissez-faire attitude to driving violations. The reliability of the scale, assessed using coefficient a, was 0.60.

808

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

2.1.5. Dri6er experience Subjects were questioned, at age 21, concerning the types of motor vehicle they drove and the length of time that they had held a licence to drive each type of vehicle. The number of years that the young person had held a driver’s licence for any vehicle was used to provide an overall measure of driver experience.

over the period from age 14 to 18 years (Fergusson and Horwood, 1999; Fergusson and Horwood, in press).

2.3. Statistical analysis

2.1.6. Annual distance dri6en At age 21, subjects were questioned concerning the distances they had driven over the period from 18 to 21 years. This information was used to derive an estimate of the total distance driven in each year from age 18–21 years. Distances were graded on a 6-point scale ranging from 1= B 5000 km to 6= \25 000 km per annum.

Linkages between drink driving behaviours and accident rates were analysed using a General Estimating Equation (GEE) modelling approach (Liang and Zeger, 1986). The GEE approach enables the estimation of a range of linear models within a longitudinal framework. The approach also permits the use of differing distributional assumptions and differing assumptions about the across time correlations of model disturbances. In the present instance the general model fitted was:

2.2. Social, family and indi6idual factors

log (Yit )= B0 + B1 Xit + SBj Zij + Eit

To control for possible confounding of the association between drink driving behaviour and accident rates from sources other than driving related factors, a range of measures of social, family and individual characteristics were available from the data base of the study. These factors included: 1. Measures of socio-demographic characteristics including: maternal age at the birth of the subject; maternal education (no qualifications/high school qualifications/tertiary qualifications); family socioeconomic status assessed using the Elley and Irving (1976) scale of occupational status for New Zealand; family type (single parent/two parent). 2. Measures of family functioning including: parental change and conflict; parental history of alcohol problems or criminality; parental illicit substance use; parental use of physical punishment (Fergusson et al., 1992; Fergusson and Lynskey, 1997). 3. Measures of individual characteristics including: gender; child IQ assessed using the Wechsler Intelligence Scale for Children (WISC-R, Wechsler, 1974) at age 8 years; novelty seeking assessed using the Tridimensional Personality Questionnaire (Cloninger, 1987) administered at age 16 years; measures of conduct and attentional problems based on the Rutter (Rutter et al., 1970) and Conners (1969, 1970) parental and teacher behaviour rating scales, assessed at annual intervals from the point of school entry to adolescence (Fergusson et al., 1991). 4. Measures of adolescent lifestyle including: quantity and frequency of alcohol use at age 18; annual measures of the extent of alcohol abuse symptomatology over the period from 18 to 21 years, assessed using standardised (DSM-IV, American Psychiatric Association, 1994) diagnostic criteria for alcohol abuse; measures of the frequency of illicit drug use over the period from 18 to 21 years; measures of affiliations with delinquent or substance using peers

where Yit was the frequency of accidents reported by the ith subject at time t, Xit was the corresponding report of drink driving by subject i at time t, Zij were a set of covariate factors (e.g. gender, driver experience, driver behaviour), and Eit was the disturbance or error term of the model. The model disturbance was assumed to have a Poisson distribution and the fitted model assumed an unstructured correlation matrix between the disturbance terms (Eit ) across time. Model fitting was conducted using STATA (StataCorp, 1997). The regression parameters from the fitted GEE models have the interpretation of the predicted effect of a one unit change in the predictor variable on the log transformed accident rate over time. However, a more useful statistic is the rate ratio. This may be interpreted as the relative increase in the annual accident rate for a one unit increase in the predictor variable, when all other predictors in the model have been taken into account (the rate ratio may be estimated by raising the base of natural logarithms (e) to the power of the parameter of interest). To take account of missing data due to sample attrition, the methods described by Carlin et al. (1999) were used. These methods involved a two stage analysis process. In the first stage of the analysis, a sample selection model was constructed by using data gathered at birth to predict participation. The model fitted was: Logit (Oit )= B0 + SBj Xij where logit (Oit ) was the log odds that the ith subject would have been observed at time t and Xij were a set of variables describing this subject at the initial (birth) interview. These measures include: maternal age; maternal education; maternal smoking during pregnancy; ethnicity, family socio-economic status and family type (one parent or two parent family). On the basis of the fitted regression model, the sample was post-stratified into a series of groups and the probability of study participation estimated for each group. The observa-

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814 Table 1 Rate of active and passive traffic accidents (per 100 individuals per annum) by drink driving behaviour score Age

Drink driving behaviour score 0

1–3

4–6

7+

18–19 Years (N) Active accidents Passive accidents

(551) 13.8 3.6

(219) 16.9 9.1

(82) 29.3 7.3

(55) 41.8 9.1

19–20 Years (N) Active accidents Passive accidents

(476) 13.9 4.0

(274) 16.4 4.4

(101) 15.8 6.9

(56) 28.6 10.7

20–21 Years (N) Active accidents Passive accidents

(422) 13.7 5.0

(309) 19.7 6.8

(126) 29.4 5.6

(50) 36.0 6.0

Overall (N a) Active accidents Passive accidents

(1449) 13.8 4.1

(802) 17.8 6.6

(309) 24.9 6.5

(161) 35.4 8.7

a

Number of person years at risk.

tions for each individual were then weighted by the inverse of this probability in the fitted regression models.

3. Results

3.1. Drink dri6ing beha6iour and accident risk Table 1 shows the relationships between drink driving behaviour scores at ages 18 – 19; 19 – 20; 20 – 21 and corresponding rates of traffic accidents. In this table, the drink driving score has been classified into four class intervals that range from those who did not report drink driving in a given year, to those whose drink driving behaviours placed them in the most drink drive prone 5–6% of the cohort. As explained in the Section 2, accidents were classified as active (accidents which resulted from the driver’s behaviours) and passive (accidents that resulted from other drivers’ behaviours). The table also includes an overall summary analysis which shows rates of accidents over the period from 18 to 21, assessed on the basis of person years exposure to the drink driving categories. The table shows that, at all

809

ages, increasing drink driving behaviours were associated with corresponding increases in both active and passive traffic accidents. The person year analysis suggests that high levels of drink driving behaviour were associated with rates of active accidents that were 2.6 times the rate of accidents of those who did not drink and drive and rates of passive accidents that were 2.1 times the rate of those who did not drink and drive. To represent the results in Table 1, GEE models were fitted to the data (see Section 2). The models assumed that the accident outcome measures had a Poisson distribution and modelled the log of the accident rate in each year as a linear function of the individual’s reported extent of drink driving in that year. To take account of variations in exposure to risk resulting from variation in distance travelled and driver experience the analysis also included measures of the estimated distance driven by the respondent during each year and duration of time for which the individual had held a driving licence. The results of this analysis are reported in Table 2 which shows estimates of the model parameters, standard errors and tests of statistical significance for the drink driving variable for each accident outcome (active; passive). The table gives the parameters for the fitted models prior to and following statistical control for driver experience. This table shows: 1. Drink driving behaviours were significantly (PB 0.0001) related to rates of active traffic accidents both prior to and following control for measures of driver experience. The strength of this association is given by the rate ratio estimate which shows the proportionate increase in the rate of active traffic accidents for a one unit change in the drink driving score (when this measure was scored in the classes shown in Table 1). The adjusted rate ratio of 1.37 implies that those who had high (\ 6) drink driving scores in a given year, had rates of active traffic accidents that were 1.373 = 2.6 times higher than those who did not drink and drive in that year. 2. In contrast, the analysis suggested that drink driving scores were not significantly related to rates of passive traffic accidents after control for driver experience. Before adjustment for driver experience,

Table 2 Effects of drink driving behaviour on rates of traffic accidents before and after adjustment for distance travelled and driver experience Measure

Parameter

SE

P

Rate ratio (95% CI)

Acti6e accidents Effect of drink driving alone Effect of drink driving adjusted for distance travelled and driver experience

0.331 0.312

0.046 0.047

B0.0001 B0.0001

1.39 (1.27–1.52) 1.37 (1.25–1.50)

Passi6e accidents Effect of drink driving alone Effect of drink driving adjusted for distance travelled and driver experience

0.242 0.106

0.076 0.079

B0.001 \0.15

1.27 (1.10–1.48) 1.11 (0.95–1.30)

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

810

Table 3 Fitted GEE model of active accident rate and drink driving behaviour adjusted for distance driven, driver experience and driver behaviour Measure

Parameter SE

Drink driving 0.164 behaviour Driver 0.067 behaviour Annual 0.063 distance travelled Driver −0.062 experience Constant −2.328

P

Rate ratio (95% CI)

0.051 B0.001

1.18 (1.07–1.30)

0.009 B0.0001

1.07 (1.05–1.09)

0.027 B0.05

1.06 (1.01–1.12)

0.031 B0.05

0.94 (0.88–0.99)

0.111 B0.0001



the rate ratio was 1.27 (P B0.001) but after control for driver experience measures this ratio reduced to 1.11 (P\ 0.15). This analysis suggests that the association between drink driving and passive accident rates largely reflected driver experience variables and, notably, the distance the individual travelled per annum. In summary, the above results suggest that when due allowance was made for distance travelled and length of driving experience, drink driving behaviours were associated with clear and significant increases in rates of active traffic accidents but were not significantly related to corresponding increases in passive accidents.

dent risk. To address this possibility, the regression model for active traffic accidents reported in Table 2 was extended to include the driver behaviour score as a further covariate factor. The fitted model is shown in Table 3. This table shows that the introduction of the driver behaviour score as a covariate substantially reduced the association between drink driving and rates of active traffic accidents. Prior to control for driver behaviour the rate ratio associated with drink driving was 1.37 (95% CI 1.25–1.50; PB 0.0001)) whereas after control this variable the rate ratio reduced to 1.18 (95% CI 1.07–1.30; PB 0.001). The latter estimate implies that after adjustment for driver experience, distance travelled per annum and driver behaviours those with high (\6) drink driving scores had rates of accidents that were 1.183 = 1.6 times higher than those who did not drink and drive. Nonetheless, even after control for driver behaviour and driver experience measures, drink driving behaviours remained significantly (PB0.001) related to rates of active traffic accidents. In summary, the results in Table 3 suggest that, to a large extent, the association between drink driving behaviour and rates of active traffic accidents arose because drink driving behaviours were associated with more general tendencies to risky or illegal driver behaviours. Nonetheless, even after control for driver behaviours, there was evidence of a specific association between drink driving behaviours and rates of active traffic accidents.

3.3. The contribution of other co6ariates 3.2. Adjustment for dri6er beha6iours Although the results in Table 2 suggest a clear and significant relationship between drink driving behaviours and rates of active accidents, as noted earlier, it is possible that this association reflects the presence of confounding factors that were: (a) related to drink driving behaviours; and (b) associated with increased risks of active traffic accidents. Exploration of a large number of potentially confounding factors (see Section 2) revealed that the most important source of confounding arose from the associations between drink driving behaviours and the driver behaviour scale (the latter scale provided an assessment of the extent to which the respondent reported engaging in risky or illegal driving behaviours). In particular, reports of drink driving behaviours and driver behaviours were correlated in the region of 0.44 to 0.50 and driver behaviours were also related to risks of active traffic accidents (r = 0.15–0.22). These findings raised the possibility that the associations between drink driving and accident rates reported in Table 2 reflected a more general tendency for those who drank and drove, to engage in other risky or illegal driving behaviours rather than the specific effect of drink driving on acci-

Although driver behaviours proved to be the largest source of confounding, further exploration of the data identified additional confounding factors that were related to both drink driving behaviours and accident risks. These factors included gender and attitudes to driving violations. Table 4 shows the final fitted model including as covariate factors: driver behaviours; attitudes to driving violations; gender; distance travelled per annum and length of driving experience. The table shows that even after control for all factors, drink driving remained a significant (PB 0.01) predictor of rates of active traffic accidents. The rate ratio implies that those reporting high (\ 6) drink driving scores had rates of traffic accidents that were 1.5 times higher than those who did not drink and drive. The analysis also identifies high levels of risky or illegal driving behaviours (P B0.0001); male gender (PB 0.01); favourable attitudes to driving violations (PB 0.001); and declining duration of driver experience (PB 0.05) as significant predictors of active accidents rates. In general, these results suggest that the profile of the most at risk driver for active traffic accidents was a young person who: (a) engaged in frequent drink driving behaviours; (b) engaged in frequent risky or illegal

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

driving practices; (c) was male; (d) had a laissez-faire attitude to traffic violations; and (e) had limited driving experience. Conversely, those at least risk were those characterised by an absence of these risk factors.

811

effects of drink driving behaviours on accident rates did not vary with driver behaviour or experience.

4. Discussion

3.4. Model checking To examine whether the conclusions drawn above were altered by changes in model specification the following additional analyses were conducted.

3.4.1. Inclusion of further confounding factors To examine whether the conclusions and results drawn above were influenced by omitted confounding factors, the model in Table 4 was extended to include further confounding factors as described in Section 2. These factors included measures of family social background (maternal age, education, family SES, family type), measures of family functioning (parental change/ conflict, parental adjustment, child abuse), individual characteristics (IQ, novelty seeking, childhood behaviour problems) and measures of adolescent lifestyle (alcohol and illicit drug use, peer affiliations). None of these factors proved to be significantly related to rates of active traffic accidents after the factors summarised in Table 4 were taken into account. Furthermore, the association between drink driving behaviours and rates of active traffic accidents remained unaltered after the inclusion of these further factors. This analysis provides considerable re-assurance that the association between drink driving and active accident rates described in Table 4 does not arise from the effects of omitted confounding factors. 3.4.2. Tests of interaction To examine whether the main effects model described in Table 4 provided an adequate account of the data, this model was extended to include tests of interaction between gender and all risk factors, time of measurement and all risk factors and between drink driving and other driver characteristics. No significant interactions were found suggesting that the results held equally for males and females, for different ages, and that the

In this paper we have used data gathered over the course of a 21 year longitudinal study to examine the linkages between drink driving behaviours and traffic accident risk in a sample of over 900 respondents. The major findings and implications of these analyses are discussed below. First, in confirmation of the findings of previous research (Mayhew et al., 1986; West, 1995) there were clear and consistent linkages between self reported drink driving behaviour and accident rates. However, control for distance travelled and driver experience suggested the presence of a specific association in which drink driving behaviours were related to active traffic accidents, in which driver behaviour contributed to the accident, but not to passive accidents, in which the behaviour of the driver did not contribute to the accident. From the analysis, it was estimated that those who engaged in frequent drink driving behaviours had rates of active traffic accidents that were 2.5 times higher than those who did not drink and drive. On the basis of this evidence, it is clear that cohort members who engaged in drink driving behaviours were an at risk group for becoming involved in active traffic accidents. As pointed out earlier, it was possible that this association may not reflect the specific effects of drink driving on driver risk but may also reflect the presence of third or confounding factors that were: (a) correlated with drink driving behaviours; and (b) made independent contributions to risks of traffic accidents. The present analysis supports the view that the associations between drink driving and traffic accident risk were, in part, explained by such factors. In particular, drink driving behaviours proved to be quite strongly related to general driving behaviours, with those reporting drink driving behaviours also tending to report higher rates of other unsafe or illegal driving behaviours. Control for the correlated effects of driving behaviours

Table 4 Final GEE model of active accident rate and drink driving behaviour adjusted for all covariates Measure

Parameter

SE

P

Rate ratio (95% CI)

Drink driving behaviour Driver behaviour Attitudes to driving violations Male gender Annual distance travelled Driver experience Constant

0.135 0.051 0.052 0.310 0.047 −0.072 −3.160

0.051 0.009 0.015 0.116 0.027 0.032 0.336

B0.01 B0.0001 B0.001 B0.01 B0.10 B0.05 B0.0001

1.14 1.05 1.05 1.36 1.05 0.93 –

(1.04–1.27) (1.03–1.08) (1.02–1.08) (1.08–1.71) (0.99–1.11) (0.87–0.99)

812

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

reduced the association between drink driving and the active traffic accidents rate very substantially. Prior to control for driving behaviours, members of the most drink/drive prone 5% of the cohort had rates of traffic accidents that were two and a half times that of individuals who did not drink and drive, whereas after control for driver behaviours this rate ratio reduced to 1.6. However, it is possible that adjustment for the confounding effects of driver behaviours may have led to the ‘over control’ of the association because, in some cases, unsafe driving behaviours may have occurred as a consequence of alcohol consumption rather than being a correlate of this consumption. Control for a further set of confounders that included gender and driver attitudes, reduced the association further but even after all confounding factors had been controlled, significant associations remained between drink driving behaviours and traffic accident risk. From the fitted model, it was estimated that the 5% of cohort members who were most prone to drink driving behaviours had traffic accident rates that were 1.5 times higher those who did not drink and drive (after due allowance had been made for a series of confounding factors including driver behaviours, driver attitudes, gender and driver experience). Further exploration showed that this result remained unchanged when a wide range of additional covariate factors, including measures of social and family functioning, individual characteristics and lifestyle factors, were considered. The finding that adolescent life style variables did not predict accident rates when due allowance was made for driver related measures, is generally consistent with the findings of another New Zealand study (Begg et al., 1999) that concluded that adolescent life style factors made only a small contribution to accident risks. These findings, when taken in conjunction with the results of the present study, clearly suggest a need for traffic safety policies to target proximal driver behaviours including drink driving, driver behaviour and attitudes, rather than more general adolescent life style variables. In general, these findings provide substantial re-assurance that, independently of a wide range of potentially confounding factors, increasing participation in drink driving behaviour in this cohort was associated with a detectable increase in rates of traffic accidents. These findings bear a strong similarity to the results reported by West (1995) in his study of drink driving and traffic accident risk amongst novice drivers. In agreement with the findings of the present study, West found that the associations between drink driving and accident risks were, in part, explained by the correlated effects of driver behaviour but that even after driver behaviours had been taken into account, drink driving was associated with increased accident risk amongst drivers under the age of 20.

The analysis was also extended to examine possible interactions between drink driving behaviour and other risk factors for traffic accidents. This analysis failed to show the presence of interactive relationships and suggested that a main effects model adequately described the data. Specifically, there was no evidence to suggest that the effects of drink driving on accident risk varied with gender or distance driven, suggesting that drink driving increased accident risks in males and females in a similar way and that these effects did not vary with increasing driver experience (amongst those aged under 21). The present study has a number of limitations that should be recognised. These limitations centre around the fact that the assessments of drink driving behaviour and accident risks were based on self report data. It is likely that these data will not provide a completely accurate account of the drink driving and accident history of cohort members but may provide a more realistic assessment of the frequency of these events than an analysis of officially recorded drink driving and officially recorded accidents. This is because the analysis of officially recorded drink driving or traffic accidents is likely to reflect two processes: (a) the true but unobserved linkage between drink driving behaviours and accident risk; and (b) the legal and social processes by which drink driving comes to official attention. Although the direction of any bias resulting from reporting errors cannot be determined with certainty, in the present instance, it seems likely that errors of measurement in reports of drink driving and accident risk will lead to the association between drink driving and accident risk being under estimated rather than over estimated. For this reason, the present analysis probably provides a conservative estimate of the link between drink driving behaviours and traffic accident risk. A further limitation of the analysis, is that owing to the relatively low rate of traffic accidents involving injury, it was not feasible to analyse accidents involving injury and those not involving injury separately. In general, the results of this study support the view that the well established correlations between drink driving behaviours and traffic accident risk are likely to provide an over estimate of the effects of drink driving on these risks. In the present study, control for confounding factors reduced the size of association by up to 50%. These findings clearly highlight the need for the issue of drink driving to be placed in perspective in road safety programmes. In particular, the findings of this study suggest that drink driving behaviours are frequently part of more general tendencies to risky or unsafe driving that may also include speeding, unsafe overtaking and ill judged decision making on the road. These findings clearly suggest the need to embed issues

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

concerning drink driving into the context of more general road safety campaigns that focus on all aspects of unsafe driving practices rather the focussing specifically on drink driving as a major source of traffic accidents. This point has been clearly illustrated by a US study reported by Hingson et al. (1996) who found that a multi-strategy approach targeting all aspects of driver behaviour reduced traffic fatalities by up to 25% with this contribution being substantially greater than that of programmes which had targeted drink driving behaviours in isolation. In this respect, it is of interest to note that recent traffic safety planning for New Zealand has emphasised the role of such a multi strategy approach to accident reduction and the need for these interventions to be developed in the context of a comprehensive framework that examines the costs, benefits and consequences of different approaches to accident prevention and reduction (Land Transport Safety Authority, 1995, 1998). The findings of the present study clearly support the view that the regulation of drink driving behaviours amongst young people is likely to contribute to a reduction in traffic accidents amongst this group but to be fully effective, attempts at regulating drink driving also need to be accompanied by a similar level of investment in regulating other aspects of unsafe, risky or illegal driving behaviours amongst young people.

Acknowledgements This research was funded by grants from the Health Research Council of New Zealand, the National Child Health Research Foundation, the Canterbury Medical Research Foundation and the New Zealand Lottery Grants Board.

References American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders, fourth ed, DSM-IV. American Psychiatric Association, Washington DC. Bailey, J.P.M., 1987. The role of alcohol and drugs in young persons’ injury or fatal road accidents in New Zealand. In: Benjamin, T. (Ed.), Young Drivers Impaired by Alcohol and Other Drugs. Royal Society of Medicine Services, London, pp. 45–53. Baker, S.P., O’Neill, B., Ginsburg, M.J., Li, G., 1992. The Injury Fact Book. Oxford University Press, New York. Begg, D.J., Langley, J.D., Williams, S.M., 1999. A longitudinal study of lifestyle factors as predictors of injuries and crashes among young adults. Accident Analysis and Prevention 31, 1–11. Beirness, D., 1993. Do we really drive as we live? The role of personality factors in road crashes. Alcohol, Drugs and Driving 9, 129 – 143. Carlin, J.B., Wolfe, R., Coffey, C., Patton, G.C., 1999. Tutorial in biostatistics. Analysis of binary outcomes in longitudinal studies

813

using weighted estimating equations and discrete-time survival methods: prevalence and incidence of smoking in an adolescent cohort. Statistics in Medicine 18, 2655 – 2670. Cloninger, C.R., 1987. A systematic method for clinical description and classification of personality variants. Archives of General Psychiatry 44, 573 – 588. Conners, C.K., 1969. A teacher rating scale for use in drug studies with children. American Journal of Psychiatry 126, 884 –888. Conners, C.K., 1970. Symptom patterns in hyperkinetic, neurotic and normal children. Child Development 41, 667 – 682. Deery, H.A., Love, A.W., 1996. The effect of a moderate dose of alcohol on the traffic hazard perception profile of young drinkdrivers. Addiction 91, 815 – 827. Donovan, D., Queisser, H., Salzberg, P., Umlauf, R., 1985. Intoxicated and bad drivers: sub-groups within the same population of high-risk men drivers. Journal of Studies on Alcohol 46, 375–382. Elley, W.B., Irving, J.C., 1976. Revised socio-economic index for New Zealand. New Zealand Journal of Educational Studies 11, 25 – 36. Evans, L., 1991. Traffic Safety and the Driver. Reinhold, New York. Fergusson, D.M., Horwood, L.J., 1999. Prospective childhood predictors of deviant peer affiliations in adolescence. Journal of Child Psychology and Psychiatry 40, 581 – 592. Fergusson, D.M., Horwood, L.J. Does cannabis use encourage other forms of illicit drug use? Addiction (in press). Fergusson, D.M., Horwood, L.J., Lloyd, M., 1991. Confirmatory factor models of attention deficit and conduct disorder. Journal of Child Psychology and Psychiatry 32, 257 – 274. Fergusson, D.M., Horwood, L.J., Lynskey, M.T., 1992. Family change, parental discord and early offending. Journal of Child Psychology and Psychiatry 33, 1059 – 1075. Fergusson, D.M., Lynskey, M.T., 1997. Physical punishment/maltreatment during childhood and adjustment in young adulthood. Child Abuse and Neglect 21, 617 – 630. Hedlund, J., 1994. If they didn’t drink, would they crash anyway? The role of alcohol in traffic crashes. Alcohol, Drugs and Driving 10, 115 – 125. Hingson, R., McGovern, T., Howland, J., Herren, T., Winter, M., Zakocs, R., 1996. Reducing alcohol-impaired driving in Massachusetts: the saving lives program. American Journal of Public Health 86, 791 – 797. Land Transport Safety Authority, 1995. Safety Directions, second ed. Land Transport Safety Authority, Wellington. Land Transport Safety Authority, 1998. New Zealand Road Safety Programme 1999 – 2000. Land Transport Safety Authority, Wellington. Liang, K.Y., Zeger, S.L., 1986. Longitudinal data analysis using generalized linear models. Biometrika 73, 13 – 22. Mayhew, D.R., Donelson, A.C., Beirness, D.J., Simpson, H.M., 1986. Youth, alcohol and relative risk of crash involvement. Accident Analysis and Prevention 18, 273 – 287. Moskowitz, H., Burns, M.M., Williams, A.F., 1985. Skills performance at low blood alcohol levels. Journal of Studies on Alcohol 46, 482 – 485. Reason, J., Manstead, A., Stradling, S., Parker, D., Baxter, J., 1991. The Social and Cognitive Determinants of Aberrant Driving Behaviour, Contractors Report 253. Transport Research Laboratory, Crowthorne. Rutter, M., Tizard, J., Whitmore, K., 1970. Education, Health and Behaviour. Longmans, London. Stacy, A., Newcomb, M., Bentler, P., 1991. Personality, problem drinking and drunk driving: mediating, moderating and direct effect models. Journal of Personality and Social Psychology 60, 795 – 811. StataCorp, 1997. Stata Statistical Software: Release 5.0. Stata Corporation, College Station, Texas.

814

L.J. Horwood, D.M. Fergusson / Accident Analysis and Pre6ention 32 (2000) 805–814

Stoduto, G., Vingilis, E., Kapur, B.M., Sheu, W., McLellan, B.A., Liban, C.B., 1993. Alcohol and drug use among motor vehicle collision victims admitted to a regional trauma unit: demographic, injury and crash characteristics. Accident Analysis and Prevention 25, 411 – 420. United States Department of Health and Human Services, 1988. The Surgeon General’s Workshop on Drunk Driving: Proceedings, Office of the Surgeon General (PHS), Rockville, MD. Wechsler, D., 1974. Wechsler Intelligence Scale for Children — Revised. The Psychological Corporation, New York. West, R., 1993. Accident script analysis: Parts 1 and 2. Report to Department of Transport: contract no. N06100. Department of Transport, London.

.

West, R., 1995. Towards unravelling the confounding of deviant driving, drink driving and traffic accident liability. Criminal Behaviour and Mental Health 5, 452 – 462. West, R., 1996. Cross-cultural Generalisability of Relationship Between Anti-social Motivation and Accident Risk, Research report. Transport Research Laboratory, Crowthorne. West, R., Hall, J., 1997. The role of personality and attitudes in traffic accident risk. Applied Psychology: An International Review 46, 253 – 264. West, R., Wilding, J., French, D., Kemp, R., Irving, A., 1993. Effect of low and moderate doses of alcohol on driving hazard perception latency and driving speed. Addiction 88, 527 – 532.