Introduction

Despite a large body of research having investigated the antecedents of drink driving offending to inform deterrence initiatives (Freeman et al. 2021a, 2020, 2016), the behaviour remains an ongoing safety risk among all roads and communities. Among Australian drivers specifically, drink driving is attributed to approximately a quarter of all crashes and half of fatal crashes in Australia (Road Safety Commission 2021; Department of Transport and Main Roads 2021; Transport Accident Commission 2021). Within Australian jurisdictions, drink driving is a criminal offense and drivers must be under a blood alcohol concentration (BAC) of 0.05 g per 100 ml of blood. Nonetheless, recent statistics (Freeman et al. 2021b) have shown that a relatively large proportion of Queensland drivers have driven when they thought they may have been over the legal alcohol limit (22.1%) and when they were knowingly over the limit (15.9%). With consideration to the high proportion of self-reported drink driving offending and the increased fatality risk of alcohol-related crashes (ARCs), the issue continues to highlight a need for further research to help reduce offending frequency and crash severity.

Characteristics of drink drivers

Drink driving research has thus far focused on the perceptual and behavioural factors of individuals that are associated with drink driving intentions. Such research has generally concluded that factors such as: risk perceptions of harm and apprehension, perceptions of impairment, social influences, alcohol use frequency, past offending, experiences with punishment avoidance, exposure to sanctions, dimensions of personality (e.g., sensation seeking; psychopathy), and attitudes towards sanctions; in part, play a role in the intentions to offend (Fernandes et al. 2010; Freeman et al. 2020, 2021b; Szogi et al. 2017; Watson et al. 2017; Freeman and Watson 2009; Stringer 2021; Hatfield et al. 2014). However, while it is important to continue building an understanding of the precursors that influence drink driving behaviours, it is also crucial to identify the situational factors associated with an increased risk of crashing once the behaviour has been engaged. Such research may inform the procedures around policing the behaviours but may also highlight areas of focus for infrastructural change or community safety awareness campaigns.

As opposed to experiential and psychological risk factors, a number of studies have identified demographic (e.g., age) and situational (e.g., location) risk factors to drink driving. In particular, males have demonstrated to be over-represented within drink driving populations (Freeman et al. 2020), although the tendency for women’s drink driving behaviours has also been noted (Armstrong et al. 2014). Secondly, older drivers have been noted to be at a higher risk of offending (Goldenbeld et al. 2020), with the average drink driver age reportedly being between 36 and 40 years old (Davey et al. 2020; Freeman et al. 2021c); younger cohorts are proportionally more prevalent among fatal crashes, due to inexperience with both driving and alcohol (DTMR 2015). Such statistics suggest that the characteristics of drink driving populations may not be entirely translatable to the risk of crashing when under the influence of alcohol. Alternatively, reports of random breath testing (RBT) data have also identified the remoteness of a location and the time of day as situational factors of concern, in that drink driving is proportionally more frequent and done at higher BAC levels, in rural areas and at night between 6 p.m. and 6 a.m. (Armstrong et al. 2017).

How alcohol increases crash risk

Given the high proportion of ARCs within the crash data, investigators have also attempted to understand how alcohol may affect driving performance. To date, it has been established that alcohol consumption can greatly reduce divers’: ability to judge speed and distance, psychomotor functioning, co-ordination, attentional control, and reaction speed (Christoforou et al. 2013; Rakauskas et al. 2008; Zhao et al. 2014; Wester et al. 2010; Jongen et al. 2014). More specifically, studies have indicated that BAC has a negative relationship with the performance of driving-related factors (Martin et al. 2013). For example, several driving simulator studies have shown delayed reaction times when under the influence of alcohol, compared to non-alcohol conditions (Yadav and Velaga 2019). In fact, a 10% increase in BAC has been shown to lead to a 2% increase in reaction times (Christoforou et al. 2013). Such findings suggest that drink drivers may be more at risk when driving, particularly when in hazardous conditions, such as slippery roads or driving in poor lighting.

Another factor to consider is that alcohol has been evidenced to influence an increased tendency to take risks (Fromme et al. 1997; McMillen and Wells-Parker 1987). A study of coroner data by Freeman et al. (2021c) demonstrated that crashes involving alcohol and drugs also tend to have higher rates of other risky behaviours, such as speeding and reckless driving. This may mean that drink drivers are at significantly higher risk of crashing, not only because they could be more likely to engage in other risky driving behaviours (e.g., speeding, mobile phone use and fatigued driving), but also because they have a reduced ability to drive. One study looked at the impact that the combined effects of alcohol and distraction would have towards driving ability, and showed that alcohol and distraction produced amplified adverse effects on driving ability (Rakauskas et al. 2008). Regardless of the dangers associated with drink driving, no such studies have investigated the increased risk that the behaviour may have on safety, when accounting for the combination of risky driving behaviours and unfavourable road conditions.

The present study

In summary of the current drink driving literature, there has been strong support given by research studies in identifying the antecedents of intentional drink driving behaviour. However, very little research has been afforded to the investigation of crash data, which could highlight the increased risks that are associated with drink driving. Identifying the comparative risks of specific situational factors may help inform road authorities to designate finite resources more appropriately. Therefore, the primary purpose of this study was to explore the comparative risks associated with ARCs. Specifically, it first aimed to explore how specific situational variables (remoteness, speed zones, days of the week, hours of the day) and risk factors (risky behaviours and road-related conditions) might impact on the likelihood and crashing, compared to non-ARCs. Secondly, it was aimed to examine how the situational and risk factors might increase ARC injury severity and fatality risk. Thirdly, this study aimed to investigate the fatality risk of ARCs when engaging in other risky behaviours (i.e., speeding, fatigued driving, distraction, and being unrestrained), across specific situational conditions.

Method

Crash data

This study used data provided by the Department of Transport and Main Roads (DTMR), Queensland, which contained information related to road crash data (N = 63,226) that occurred during the years 2015 to 2019. All cases of data were included, and no missing coding was found. Five years of data were used to: (a) ensure an adequate sample of ARCs was collected; and (b) ensure the sample was sufficient in capturing the true impact of factors associated with drink driving crashes and reduce the effect of time specific anomalies that may have occurred between years. Overall, ARCs represented 5.7% (n = 3626) of all vehicle crashes in Queensland between 2015 and 2019. Five years of crash data allowed a comprehensive analysis to identify necessary trends and make comparisons between ARCs and non-ARCs, but also likely remained relevant to the current generational trends. Crash data are traditionally collected from an attending police officer, who records the crash characteristics and any potential causal factors that might have been present. More serious crashes (i.e., resulting in fatal or serious injuries), however, can be attended to by a forensic crash investigation unit, in which trained officers will use an in-depth investigation to inform their report of the crash characteristics and causal factors.

The variables of interest for this study were related to characteristics and causal factors that may be relevant to ARCs. Situational variables included: time of day, period of week (i.e., weekday; weekend), level of remoteness (i.e., rural, suburban, urban), and the speed zone (≤ 50 kmph, 60–70 kmph, 80–90 kmph, and 100–110 kmph). In addition, a number of dichotomously scored variables identifying whether certain factors were present in the crash were also utilised and included: drink driving, speeding behaviours, fatigue, distraction (e.g., inattention and mobile phone use), whether the road was unsealed, whether the road was slippery, whether visibility was low (i.e., rain and fog), and whether the road was poorly lit (e.g., at night). Finally, injury severity was also used as a means to determine the severity of the crash (1—minor injury, 2—injury requiring medical treatment, 3—injury requiring hospitalisation, and 4—injury leading to fatality).

Data analysis

The data were analysed using statistical analysis software SPSS (version 28). First, cross-tabulations and chi-square tests were used to examine and compare the differences in ARC and non-ARC proportions among the situational variables (i.e., where and when) and risk factors (i.e., behaviours and road conditions). A factor of comparative likelihood was then calculated by dividing the relative proportions of ARCs into non-ARCs for each variable, giving a value indicating the likelihood that the variables would relate to ARCs, compared to non-ARCs. Next, analysis of variance (ANOVA) tests were used to examine the differences in injury severity of ARCs amongst the situational variables and risk factors. Browns–Forsythe (BF) statistics were reported for variables violating homogeneity of variance assumptions, and effect sizes were interpreted based on Cohen’s (1988) for ANOVAs (small = 0.01, medium = 0.06, large = 0.14). Further, a logistic binomial regression was used to determine the impact that relevant situational variables and risk factors were having on the fatality rates of ARCs. The findings from the univariate ANOVAs were used to retrospectively inform which variables were statistically relevant for the analysis. Frequencies were also used to help determine an appropriate classification cut-off for the analysis. Finally, to identify the comparative fatality risk of each risky behaviour across the situational data, proportions and comparative likelihood statistics were calculated across ARCs.

Results

The comparative likelihood of ARCs and non-ARCs

As discussed, cross-tabulations and chi-square analysis (Table 1) were used to highlight the comparative risk of ARCs, compared to non-ARCs. When looking at the level of remoteness, the results revealed that despite the majority of ARCs occurring in urban (51.9%) and suburban (42.3%) areas, rural areas (5.8%) had a higher proportion of ARCs that ended in fatalities (17.5%), compared to suburban (7.6%) and urban (4.1%) areas. The spread of ARCs was also shown to be significantly different than non-ARCs (χ2 = 314.68, p < 0.001), particularly in rural areas, which were more than two times more likely to involve ARCs compared to non-ARCs. Concerning speed zones, frequencies showed that the highest frequency of ARCs occurred in 60–70 kmph speed zones (59.9%), followed by ≤ 50 kmph zones (22.2%), 100–110 kmph zones (9.7%), and then 80–90 kmph speed zones (8.9%). Chi-square tests indicated that there was some discrepancy in the proportions between ARCs and non-ARCs (χ2 = 163.36, p < 0.001), in that ARCs tended to be more common in higher speed zones than non-ARCs. Subsequently, ARCs were slightly less likely to occur in speed zones of 70 kmph and below, but slightly more likely to occur in higher speed zones.

Table 1 The proportions, comparative likelihood and statistical significance of ARCs compared to non-ARCs
Full size table

Next, comparative differences among time periods were examined, and it was shown that despite weekends having significantly fewer days compared to weekdays, a comparable proportion of ARCs were present (61.7%). Chi-square tests (χ2 = 1,001.55, p < 0.001) confirmed that compared to non-ARCs, ARCs were more than two times more likely to occur during the week and more than 1.5 less likely to occur on weekdays. The results also demonstrated that the majority of ARCs were occurring in the late evening (42.6%; 6 p.m. to 12 a.m.), followed by early morning (27.4%; 12am to 6am), early evening (20.9%) and late morning (9.0%). Comparative analysis indicated that there were large disparities in proportions between non-ARCs and ARCs (χ2 = 4,512.31, p < 0.001), in that ARCs were more than four times more likely to occur in the early morning, and nearly three times more likely in the late evening, and thus were subsequently less likely to occur in other time periods, compared to non-ARCs.

The crash proportions were also assessed among behavioural and road-related risks. Proportional statistics showed that firstly, all of the risk factors were identified as more common in ARCs than in non-ARCs. Most notably, speeding and poor lighting were found to be approximately five times more likely to be contributing factors, not wearing a seatbelt was four times more common, and fatigue and driving on an unsealed road were approximately two times more common among ARCs than in non-ARCs. Follow-up chi-square tests revealed that these differences were statistically significant in all cases (χ2 = 4.68 to 2,013.46, p < 0.001 to 0.031), except for distraction.

The risk of situational factors and behaviours towards ARC injury severity and fatality risk

ANOVAs were used to compare the risk of situational variance toward ARCs (Table 2). Firstly, preliminary analysis demonstrated that ARCs were significantly more dangerous than non-ARCs (BF (1, 4079) = 977.80, p < 0.001, η2 = 0.015, ΔM = 0.38). When investigating locational differences, it was shown that ARC injury severity (BF (2768) = 21.95, p < 0.001, η2 = 0.013) was significantly different between location types. Post-hoc Bonferroni tests revealed that rural locations contained ARCs that had higher injury severity, compared to suburban (ΔM = 0.14, p = 0.016) and urban (ΔM = 0.27, p < 0.001) locations; and suburban locations contained ARCs that were higher in injury severity compared to urban areas (ΔM = 0.13, p < 0.001). In addition, ARC injury severity marginally differed across speed zones (BF (3, 2077) = 2.62, p = 0.049, η2 = 0.002), with Bonferroni post-hoc comparisons showing the only difference to be between 100 kmph zones and 80–90 kmph zones (ΔM = 0.12, p = 0.043). However, further ANOVAs indicated that there was no significant difference in ARC injury severity between days of the week (F (1, 3624) = 0.00, p = 0.972, η2 = 0.000) and days of the day (BF (3, 2114) = 1.44, p = 0.230, η2 = 0.000).

Table 2 Descriptive data and ANOVA statistics of ARC injury severity across situational factors
Full size table

ANOVAs also indicated that there was significant higher injury severity when a number of risky conditions and behaviours were present, including: speeding (BF (1, 598) = 105.80, p < 0.001, η2 = 0.031, ΔM = 0.37), fatigue (BF (1, 373) = 23.84, p < 0.001, η2 = 0.006, ΔM = 0.20), no seatbelt (BF (1, 336) = 103.48, p < 0.001, η2 = 0.025, ΔM = 0.41), poor lighting (BF (1, 1680) = 93.71, p < 0.001, η2 = 0.023, ΔM = 0.24). In contrast, ARC severity was slightly less severe when distraction (BF (1, 741) = 19.15, p < 0.001, η2 = 0.005, ΔM = − 0.14), low visibility (BF (1, 177) = 5.56, p = 0.019, η2 = 0.002, ΔM = 0.10), and slippery roads (BF (1, 659) = 4.58, p = 0.033, η2 = 0.001, ΔM = − 0.08) were present factors. Proportional statistics of ARC severity types across the situational factors are also provided in Table 2 (right) to illustrate the spread of ARCs.

Finally, a binary logistic regression (Table 3) was used to investigate how the situational and risk factor variables might impact on the fatality rating of ARCs. Only those variables that showed to be statistically relevant to injury severity at the univariate level were included. Specifically, given the large sample size and capability to detect arbitrary effect sizes, only those variables with significance values of < 0.001 were included, as the associated effect sizes of variables outside of this were negligible (e.g., η2 ≤ 0.002). To first identify an appropriate cut-off level, frequencies were run on the fatality rates of ARCs and revealed that 6.3% (n = 230) of ARC crashes were fatal. The model was run accordingly and was found to be significant (χ2 = 346.10, Nagelkerke R2 = 0.242, p < 0.001) and a good fit to the data (Hosmer and Lemeshow = 3.66, p = 0.722). The model demonstrated to accurately classify a non-fatal crash 82.7% of the time, and correctly classify a positive result 66.6% of the time (overall = 81.7%). Individually, level of remoteness was found to decrease the risk of fatality by approximately half for each step up in population (B = − 0.59, p < 0.001, odds = 0.56), excessive speeding was found to increase the risk of ARCs being fatal by approximately nine times (B = 2.16, p < 0.001, odds = 8.69), fatigue by one and a half times (B = 0.51, p = 0.023, odds = 1.67), not wearing a seatbelt by approximately four times (B = 1.37, p < 0.001, odds = 3.95), and poor lighting by one and a half times (B = 0.45, p < 0.007, odds = 1.57). Conversely, distraction was found to reduce the risk of a fatality among ARCs by 84% (B = − 1.87, p < 0.001, odds = 0.16).

Table 3 A binary logistic regression with situational variables and risk factors predicting ARC fatalities
Full size table

The comparative risk of fatality among risky driving behaviours in specific situational conditions

Finally, proportions and the comparative likelihood were calculated to assess how risky behaviours may impact on the fatality rating of ARCs in certain situational conditions (Table 4). Firstly, the results showed that the increased risk of speeding was high throughout all situations, with comparative risk ratings between + 1.85 to + 5.41. However, speeding increased the fatality rating of ARCs the most in urban zones (+ 5.17), in low-speed zones (+ 5.07), on weekends (+ 5.41), in the late evening (+ 4.84) or late morning (+ 4.72), and on slippery roads (5.05). Next, fatigue was also shown to represent an increased ARC fatality risk, although the range between variables was smaller (+ 1.20 to + 3.58). Specifically, the highest fatality risk was in suburban areas (+ 1.82) in 60–70 kmph zones (+ 2.85), on weekends (+ 2.29), in the early morning (+ 2.11) or late evening (+ 2.00), and on slippery roads (+ 3.58).

Table 4 The fatality risk of ARCs when engaging in risky behaviours across different situational conditions
Full size table

Not wearing a seatbelt also dramatically increased the fatality risk of ARCs with a comparative risk range of − 1.35 to + 6.90. In particular, fatigue posed the biggest comparative fatality risks in suburban areas (+ 4.25), in lower speed areas (+ 4.88 to + 5.02), on weekdays (+ 4.61), in the early evening (+ 5.10), and in areas of low visibility (+ 6.90). Notably, there appeared to be a reduced risk of fatality in the late morning (− 1.35), although this is likely due to a low number of fatal ARCs under these conditions. Finally, as per previous results, distraction was found to reduce the fatality risk of ARC, with scores showing that distraction reduced the risk of fatalities by 2.10 to 10.92 times. Distraction had the largest reduced fatality risk in urban areas (− 10.92), both low (no fatal ARCs) and high-speed zones (− 4.27), on weekdays (− 5.00), in the late evening (− 6.67), and on unsealed roads (no fatal ARCs).

Discussion

The comparative risk of ARCs

The primary purpose of this study was to identify the increased situational risks involved with drink driving in Queensland, Australia; and to highlight what factors may increase the severity and fatality of ARCs. The findings indicated that the proportion of ARCs across locational, time-related and risk-related factors, differed significantly from non-ARCs. These differences were at times dramatic, as some situational factors were up to four times more likely to be present in ARCs than non-ARCs. Consistent with previous RBT data (Armstrong et al. 2017), the severity of crashes was significantly higher in rural areas, but far more common in populated areas. This highlights that greater police presence (and deterrence) may be required in all areas, but proportionally more in rural locations. Recent research suggests that the mere exposure to police roadside testing may be enough to create a deterrent effect among impaired drivers (Mills et al. 2022; Freeman et al. 2021b), and thus periodic high visibility operations in remote, which are not accustomed to heavy police enforcement, may be beneficial.

Alike non-ARCs, ARCs were observed to be more common in areas of lower speed zones (< 70 kmph). This trend is possibly because these areas contain a greater demand on the driving task (e.g., intersections; other road users; pedestrians) and drink drivers have an impaired cognitive (driving) ability (Christoforou et al. 2013; Rakauskas et al. 2008; Zhao et al. 2014; Wester et al. 2010; Jongen et al. 2014). However, another consideration might be related to the quantity of low-speed zones, compared to high-speed zones (i.e., a simple exposure effect). A final explanation was that speeding (a notable risk factor) was observably more common and fatal among ARCs in slower speed zones and urban areas. In fact, speeding was attributable to more than half of fatal ARCs in these areas. Also concerning was the high prevalence of drivers being unrestrained in fatal ARCs. Although the prevalence may have been higher in rural areas, the fatality rating of ARCs involving unrestrained drivers were again higher in more populated areas and lower speed zones.

Together, the findings highlight how specific situational variance impacts on different levels of risk. For example, while there is limited police presence (and thus proportionally higher offending rates) in more remote areas, there is also less environmental variables (e.g., pedestrians; intersections; stimuli) to consider and navigate when drink driving. Conversely, more populated areas may contain comparatively lower rates of offending but are more difficult to navigate when impaired and have significantly more environmental risks to consider.

In regard to time periods, the results showed that ARCs were more likely to occur on weekend days, and between the hours of 6 p.m. to 6 a.m., which is again consistent with previous findings from RBT data (Armstrong et al. 2017). One unexpected finding, however, is that unalike ARC likelihood, injury severity rarely deviated between the variables. This may be because ARCs were shown to be generally more severe and thus, less variation is found between the variables, or that underlying factors are introducing unforeseen risks. The results suggested that different risky behaviours were more prominent at different times of the day, and impacted the fatality risk of ARCs, which may in part explain the lack of variance. For example, fatigue was shown to be significantly more prevalent between the hours of 6 p.m. to 6 a.m., which may in part reduce risks during the day. However, there is also a significant increase in pedestrian and vehicle traffic volume, which may seemingly counterbalance the apparent risk, and confound broader variables, such as time periods. Nonetheless, further research on this area would help better understand the interactions between time, situational variance and crash risk.

Risky road-related conditions, such as lighting, and the road surface were shown to interact with the likelihood of ARCs (up to four and half times) and the associated injury severity. Such findings again suggest that drink drivers’ may be less likely to adequately meet situations requiring heightened attentional processing. This supports previous findings that alcohol can greatly reduce cognitive performance while driving (Christoforou et al. 2013; Rakauskas et al. 2008; Zhao et al. 2014; Wester et al. 2010; Jongen et al. 2014).

In conjunction, the results also showed that risky behaviours such as speeding and driving fatigued were significantly more common (two to five times more likely) among ARCs than for non-ARCs. This finding implies that drink drivers are at an exponential risk of crashing because they may be more likely to engage in several other risk-related behaviours, while also being cognitively impaired. As discussed, alcohol consumption has been linked to an increased tendency to take risks (Fromme et al. 1997; McMillen and Wells-Parker 1987), and crashes involving alcohol have been associated with other risky behaviours, such as speeding and reckless driving (Freeman et al. 2021c). This exponential risk was highlighted in the current findings, as risky driving behaviours were shown to increase the fatality risk (up to seven times) of ARCs when risky road conditions were present.

In contrast, road conditions such as slippery roads and low visibility (e.g., rain; fog) were shown to decrease the severity of ARCs. However, previous research has shown that drivers adopt safer driving behaviours in weather conditions (e.g., driving slower), due to increased risk perceptions related to crashing (Ahmed and Ghasemzadeh 2018; Cai et al. 2016). Therefore, these regulatory behaviours may also be present in drink drivers, who would be exponentially impaired in such conditions, despite their increased propensity for risk (Fromme et al. 1997; McMillen and Wells-Parker 1987). Previous research has shown that cannabis users adopt safer driving behaviours when under the influence in order to reduce the associated risks of intoxication (Arkell et al. 2020). While the evidence is sparse, this may be similar for alcohol users, who have noted the aversive effects of alcohol on their driving (Love et al. 2022).

Similarly but unexpectedly, distraction was also shown to reduce the odds of an ARC being fatal by as much as 94%, which is not supportive of previous research suggesting that distracted driving increases crash risk (Choudhary et al. 2020), particularly when paired with drink driving (Rakauskas et al. 2008). This discrepancy may have been due to underlying factors, such as that mobile phone use may be more likely to be combined with drink driving in lower speed areas or safer circumstances. As previously highlighted, mobile phone offenders regulate their driving and phone use to meet their perceived risk involved with the activity (Oviedo-Trespalacios et al. 2017). Alternatively, distraction is a difficult behaviour to determine in crashes, and it may be that attending officers were less likely to record distraction as a contributing factor in the circumstances where alcohol was already attributed to the crash. Nonetheless, further research is needed to identify both the prevalence rates of these combined behaviours and the increased risk that the collective effects might have toward driving ability.

Implications, limitations, and future suggestions

Overall, the findings of this study have highlighted particular situational variables, driving behaviours and road conditions that represent an increased crash and fatality risk for ARCs, and therefore may inform intervention strategists on a more targeted and effective countermeasure to drink driving offenders. While the crash data have indicated that drink driving represents an increased crash risk in general, it has also highlighted specific situations that represent higher crash quantities (e.g., urban areas), higher crash proportions (e.g., weekends), and higher risk for injury (e.g., rural areas) for drink drivers. There also appears to be other risky behaviours (e.g., speeding) that may be more common in specific situational driving conditions (e.g., remote areas), further mediating the level of risks involved. Therefore, to most efficiently reduce the fatality figures associated with drink driving, interventions should place particular focus towards these areas and populations (in addition more generalised approaches).

Specifically, it may be useful to delegate finite police resources to areas and populations of concern, at specific times. Alternatively, media-based interventions or educational training aimed at increasing the stigmatisation of drink driving can use the risk-based statistics outlined in this study to inform populations about the dangers of drink driving, and in particular, combining risky behaviours under specific circumstances. The findings demonstrated the potential for risk among the drink driving population, and by publicising the highlighted risk figures, both current and future drivers may develop more understanding of such risks. For example, a Japanese study showed that highly publicising ARCs and using media-based strategies may have reduced the number of ARCs, but also improved social norms and behaviour (Nakahara and Ichikawa 2011). In conjunction with this approach, designated driver programs, which aim to encourage designating a sober driver when drinking in a group, may be a useful strategy to help promote the use of safe driving practices surrounding alcohol consumption.

The statistics also indicated that the crash likelihood did not meaningfully increase when drink driving was a factor in some conditions. For example, lower-speed and more populated areas, on weekends, low visibility, slippery roads, and distraction were factors shown to have a lesser impact than others. Further research may wish to investigate whether there were little differences because these factors represent an increased risk to all drivers or whether other underlying factors (e.g., sober drivers are more likely to drive in populated areas; people drive safer when risk is present) are confounding on some effects. In addition, it would be beneficial to identify whether drink drivers adopt regulatory behaviours to reduce their perceptions of harm likelihood. While research has indicated that alcohol use is linked to risky behaviour, it may be enlightening to understand if there is some psychological variance between those who engage risk and those who negate it, after the drink driving has been initiated.

Despite the implications present, this study contained limitations and thus several future directions are suggested. Most notably, the data used contained only cases who have been involved in a traffic crash and is therefore not necessarily representative of the total drink driver population. Future research may therefore wish to make comparisons with other data sources from different jurisdictions. Similarly, it would be beneficial to investigate whether the characteristics of alcohol related infringement data matches the crash data to determine if those who drink drive and have crashed have similar dynamics to those who drink drive and get caught. Another limitation involved the reliance of police officers to report the apparent causes of the crash. Some factors may be likely to be reported as the primary crash indicator over some less observable (i.e., phone use) or objective (i.e., fatigue) factors. Self-report data may provide a contextual comparison, although this form of data is not without its own limitations.

Self-report methodologies involving the engagement of other risky behaviours while drink driving may further highlight how prevalent the magnified risk of combined offending is, specifically within the identified situational variables. Research focused on the deterrence of such offending behaviours may also shed further light on how this increased risk interacts with perceptions of impairment and risk among offenders. Finally, future research may benefit from investigations on how behavioural (e.g., speeding) and road-related (e.g., rain) risk factors may impact on specific cognitive performance indicators when under the influence of alcohol. Simulator-based studies may prove to be a beneficial avenue for this stream of research. In summary, this study has demonstrated that drink driving embodies a significant danger on the road, representing a relatively high portion of crashes (6.1%), due to an increased risk of crashing in specific conditions (up to five times). Further, the risk of being fatally injured is significantly amplified under certain circumstances (up two times) and when combined with other risky behaviours (up to nine times).