- Email: [email protected]
The Driver Behaviour Questionnaire for older drivers: Do errors, violations and lapses change over time?
Accident Analysis and Prevention 113 (2018) 171–178
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
The Driver Behaviour Questionnaire for older drivers: Do errors, violations and lapses change over time?
T
⁎
Koppel S.a, , Stephens A.N.a, Charlton J.L.a, Di Stefano M.b, Darzins P.c,d, Odell M.e, Marshall S.f a
Monash University Accident Research Centre, Monash University, Australia La Trobe University, Australia c Eastern Health, Australia d Monash University Eastern Health Clinical School, Australia e Victorian Institute of Forensic Medicine, Australia f Ottawa Hospital Research Institute, Canada b
A R T I C L E I N F O
A B S T R A C T
Keywords: Older drivers Aberrant driving behaviour Errors Violations Lapses Road safety
The aim of the current study was to examine how self-reported aberrant driving behaviours change across a three time-points in a group of older drivers. Two hundred and twenty-seven older drivers (males = 69.6%) from the Candrive/Ozcandrive longitudinal study completed the Driving Behaviour Questionnaire (DBQ) each yearacross three time-points (i.e., Year 1, Year 2, Year 3). At the third time-point, older drivers ranged in age from 77 to 96 years (M = 81.74 years; SD = 3.44 years). A longitudinal confirmatory factor analysis showed that a modified 21-item, 3-factor (errors, lapses and violations) DBQ was invariant across the time period, suggesting that the structure of the questionnaire was stable across each time-point. Further, multiple domain latent growth analysis on the resultant factors for errors, lapses and violations showed that the frequency of errors remained similar across the three-year period, while violations and lapses showed very marginal decreases in frequency. These changes were independent of the absolute number of these behaviours; Drivers with higher violations or lapses in Year one, showed similar decreases in frequency as those who self-reported lower frequencies of the behaviours. These results suggest that the DBQ is a reliable tool to measure older drivers’ self-reported aberrant driving behaviours, and that these behaviours do not show much change across time. Future research should validate the self-reported responses from the DBQ with more objective measures such as those collected through naturalistic driving study (NDS) methodology or on-road driving tasks.
1. Introduction It is important to understand how drivers’ behaviours may change over time, particularly for older drivers who are likely to form a larger proportion of the driving fleet as the population ages (Koppel and Berecki-Gisolf, 2015; Sivak and Schoettle, 2011). Older drivers are over-represented in fatal and serious injury crash statistics (Koppel et al., 2011; Langford and Koppel, 2006). While a large part of this over-representation is due to the frailty of older drivers, with around 60 to 90 percent of fatalities in this age category resulting from driver frailty (Li et al., 2003), age-related sensory, cognitive, and physical impairments also contribute (Marshall, 2008). However, previous research has shown that many older drivers become aware of such declines and change their driving patterns accordingly by self-regulating when, where and how they drive (Baldock et al., 2006; Blanchard et al., 2010; Charlton et al., 2006; Molnar and Eby, 2008; Molnar et al., 2014).
⁎
Appropriate self-regulation therefore requires recognition of age-related, and possibly driving, decline, particularly types of driving behaviours that may increase crash risk (Koppel and Charlton, 2013). One of the main threats to road safety is aberrant driving behaviour (Singh, 2015). Aberrant driving behaviours are often defined as driving errors, lapses and violations, with common examples including: exceeding the speed limit, red-light running, and heavy braking due to poor situation awareness (De Winter and Dodou, 2010; Gabaude et al., 2010). Associations have been found between self-reported aberrant driving behaviour and crash involvement (Gras et al., 2006; Parker et al., 1995; Rimmö and Åberg, 1999), with one recent study identifying some form of aberrant driving behaviours in approximately 74 percent of crashes (Singh, 2015). The Driving Behaviour Questionnaire (DBQ), initially developed by Reason et al. (1990), is one of the most widely used research tools to measure the frequency of self-reported aberrant driving behaviour.
Corresponding author. E-mail address: [email protected] (S. Koppel).
https://doi.org/10.1016/j.aap.2018.01.036 Received 17 October 2017; Received in revised form 26 January 2018; Accepted 28 January 2018 0001-4575/ © 2018 Elsevier Ltd. All rights reserved.
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Traditionally, the DBQ represents 50 individual behaviours that fit into three broad behaviour patterns: violations, errors and lapses. Violations are defined as deliberate behaviours that directly contravene road laws, for example exceeding the speed limit or failing to obey red traffic light signals. Errors are defined as behaviours that do not contravene road laws directly but, as with violations, are considered to increase a driver’s risk of crash. In contrast, lapses are defined as minor mistakes that are not considered to be associated with crash involvement (Reason, et al., 1990). While there are now a number of versions of the DBQ, which target specific groups of drivers, few studies have used this questionnaire specifically in an older population of drivers. Indeed, the original DBQ was validated on drivers ranging in age from 20 to 56 years (Reason, et al., 1990) and has mostly been used for drivers in that age range (De Winter and Dodou, 2010). Stephens and Fitzharris (2016) have shown that a four-factor version of the DBQ, based on the findings of Parker et al. (Parker et al., 1995) and that includes a factor for aggressive violations, is appropriate for a broad range of drivers, but less so when specifically examining older drivers. In line with this, Mattsson (2012) has shown that a number of DBQ items fit within different factors for older drivers. For example, behaviours that younger drivers may interpret as a violation are considered to be an error by older drivers (e.g., drink driving). Rimmö and Hakamies-Blomqvist (2002) have also studied the relationship between driving exposure, health, and four types of self-reported aberrant driving behaviours (e.g., inattention, inexperience errors, violations and mistakes) using a Swedish version of the DBQ with Swedish drivers aged between 55 and 92 years. They reported that, even after accounting for age and gender, self-reported inattention and inexperience errors, as well as impaired health, were related to self-imposed driving limitations, whereas self-reported violations and mistakes were not. Given the importance of understanding factors that increase crash risk in older drivers and identifying older drivers’ perceived level of violations, errors and lapses, it is necessary to establish an appropriate tool to measure perceived driving behaviour in this cohort. In doing this, the number and type of changes that may occur over time can then be examined. Some studies have focussed on deriving an appropriate factor structure for older drivers. Obriot-Claudel and Gabaude (2004) suggest that a three-factor, 42-item version of the DBQ is appropriate for use in drivers aged between 55 and 91 years. They suggest that these items, taken from the original 50-items, explain three main types of aberrant behaviours: inattention errors, dangerous errors and dangerous violations. Likewise, Martinussen et al. (Martinussen et al., 2013) also suggest a three-factor version of the DBQ is appropriate for drivers aged 50 to 85 years, with this version using 27 of the 50-items, and explaining factors for violations, errors and lapses. Indeed, Martinussen et al. reported that this version was the best fit for older drivers (i.e., 50–85 years) compared to middle-aged (i.e., 30–49 years) and younger drivers (i.e., 18–29 years). However, the item factor loadings of these two versions (i.e., Obriot-Claudel and Gabaude, 2004; Martinussen et al., 2013) are markedly different and therefore the degree to which each is appropriate in other samples is unclear. We need to establish an appropriate tool to measure behaviours so as to understand how these change over time, particularly in a group of vulnerable drivers who are likely to monitor their perceived driving behaviour, and possible declines, in order to self-regulate driving behaviour. To date, there is little research on how aberrant behaviours measured with the DBQ may change over time. Roman et al. (2015) investigated changes in DBQ scores for errors, slips, violations and aggressive violations across six monthly intervals in novice young drivers and found that scores for each increased across the time-points. However, despite the fact that changes in older drivers’ self-reported aberrant driving behaviour could be a crucial element in maintaining their safety, no similar research has been conducted with older drivers. Consequently, the aim of the current study was twofold; first to confirm
an appropriate structure of the DBQ for use in a sample of older drivers; and, second to assess the reliability (specifically, the test-retest reliability) of the newly structured DBQ. 2. Method 2.1. Participants Data from 227 older drivers (males = 69.6%) are presented in the current paper. Each driver was an Ozcandrive participant in the Candrive/Ozcandrive study (described in Section 2.2). Ozcandrive participants completed annual assessments across five time-points in Melbourne, Australia. All participants were required to meet the following inclusion criteria: a) aged 75 years or older; b) held a valid driver’s license; c) drove at least four times per week, and d) did not have an absolute contraindication to driving, as defined by the Austroads Fitness to Drive Guidelines (Austroads, 2010). Given drop-out rates across the fivetime-points (Year 1: n = 14; Year 2: n = 15; Year 3: n = 11; Year 4: n = 18; Year 5: n = 18)1, and to maintain an appropriate sample size (e.g., n > 200), data from participants who completed the first three assessments of study are presented in the current study. At the third time-point of the study, participants ranged in age from 77 to 96 years (M = 81.74 years; SD = 3.44 years). It should be noted that participants who remained in the study in Year 3 (n = 227) were not significantly different to participants who withdrew from the study (n = 30) in terms of their gender (Male: 69.6% vs. 80.0%, respectively, X2(1) = 1.386, p > 0.1), their Year 1 age (M = 79.66 years, SD = 3.45 vs. 80.33, SD = 3.97, respectively, t(225) = 0.992, p > 0.1), their Year 1 self-reported annual kilometres driven (< 5000 km: 15.0% vs. 20.0%; 5,001–15,000 km: 71.4% vs. 70.0%; > 15,0001 km: 13.7% vs. 10.0%, respectively, X2(2) = 0.704, p > 0.5) or their Year 1 self-reported total number of medical conditions (M = 10.52, SD = 4.13 vs. M = 11.67, SD = 4.58, respectively, t (255) = 1.405, p > 0.1). 2.2. The Candrive/Ozcandrive study The Candrive/Ozcandrive study is a longitudinal, multi-centre international research program with the core objective of identifying solutions to promote older drivers’ safe mobility (Marshall et al., 2013). The Candrive/Ozcandrive study involves 928 drivers aged 70 years and over in Canada and 302 drivers aged 75 years and older in Australia and New Zealand (Australia: n = 257; New Zealand: n = 45). Using a longitudinal study design, the project is tracking this cohort of older drivers for five years, assessing changes in their functional abilities, driving practices (e.g. exposure and patterns), as well as crashes and citations for violations of driving regulations. The primary purpose is to develop and validate a risk stratification tool to assist clinicians in identifying potentially at-risk drivers (Marshall et al., 2013). Participants’ usual (or naturalistic) driving practices (e.g., trip distance, duration, type of road, speed) are recorded through in-car recording devices (ICRD) installed in participants’ own vehicles, and measures of participants’ functional ability, medical conditions and self-reported driving-related abilities and practices are documented annually. 2.3. Procedure and materials Ethical Approval was obtained from the Monash University Human Research Ethics Committee (MUHREC), and all participants provided written informed consent. All participants underwent a baseline (Year 1) and two annual assessments (Years 2 and 3). The time between annual assessments was 1 Note some participants withdrew from the study after completing their annual assessment.
172
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
the factor loadings are consistent across the three models and this is achieved by constraining the loadings to be equal across the three models. Metric invariance determines whether similar patterns of responses are apparent across the time-points. As noted by Byrne (2013) and Cheung and Rensvold (2002), invariance is supported by a nonsignificant change in Chi-Square (from the configural model) and a change in the CFI no greater than 0.01. Once the equivalence of factor loadings is established, structural invariance is examined (Phase 3). Structural invariance requires the intercepts for each item on its factor to be constrained across models and demonstrates that the interpretations of items remain similar across the three time-points. After establishing the longitudinal invariance of the DBQ, a multiple domain latent growth curve (LGC) analysis was conducted. While the MGCFA determines reliability of the DBQ and associated factors, the LGC assesses any mean differences on factor scores across each time point. LGC analysis shows both group change and inter-individual change across the three time-points. Multiple domains consider change across the three factors in the same model. LGC examines the average intercepts for each factor (e.g., the average score at baseline/Year 1) and the degree of change represented by the regression slope over the three models. Inter-individual change is examined by the variance of the factors. To interpret the change, the LGC requires good model fit, determined using the same criteria listed in above and change is indicated through significance levels of p < 0.05.
approximately 12 months for each participant. Each assessment incorporated a range of demographic and driving history questions as well as a range of functional ability measures, reporting on medical conditions, and self-reported abilities and practices related to driving (Marshall et al., 2013). The DBQ (Reason, et al., 1990) was administered across the three time-points. The DBQ contains 50 items measuring the frequency of driving behaviours that are likely to increase crash risk. However, given that the study protocol was designed for older drivers in Canada (Candrive) as well as older drivers in Australia (Ozcandrive), two items relating to behaviour at roundabouts were removed as road infrastructure in Canada does not include roundabouts. In addition, given that most participants drove an automatic vehicle, the item relating to driving off in third gear was also removed. The final version of the DBQ contained 47-items. For each item, frequency was reported across a six-point scale of engagement (where: 1 = Never, 6 = Nearly all the time). ObriotClaudel and Gabaude (2004) report that the three factors appropriate for older drivers demonstrate good internal consistency, with Cronbach alphas of 0.85 (inattention errors), 0.84 (dangerous errors), 0.70 (dangerous violations). Martinussen et al. (Martinussen, et al., 2013) also report good internal consistency on their three-factor version (Cronbach alphas ranging from 0.64 to 0.85). 2.4. Data handling
3. Results
Missing DBQ responses were replaced with a 5.0% trimmed mean and this occurred in only four instances. Analyses were conducted using SPSS v.23 and AMOS v.22. An initial Confirmatory Factor Analysis (CFA) was performed on the baseline data to determine the most appropriate fitting model. Because an existing factor structure was being tested, CFA was chosen over Exploratory Factor Analysis (EFA) as the former allows a strict examination of the previously proposed model fit, taking into account measurement error within each factor loading. CFA also allows nested models which are imperative for longitudinal analyses. Goodness of fit indices to determine acceptable fit included: ChiSquared, χ2, acceptable value: p > 0.05, however significance is common with large sample sizes (Byrne, 2013); Comparative Fit Index, CFI, acceptable value: ≥0.90 where ≥0.95 indicates exceptional fit (Hu and Bentler, 1999); Tucker-Lewis Index, TLI, acceptable value: ≥0.90 where ≥0.95 indicates exceptional fit, (Hu and Bentler, 1999); Root Mean Square Error of Approximation, RMSEA, acceptable value: ≤0.06 (Browne and Cudeck, 1993); Significance value of the RMSEA, pclose, acceptable value: > 0.05 (Hu and Bentler, 1999); Confidence Interval around the RMSEA, 90% CI, acceptable value: maximum value not exceeding 0.06 (Browne and Cudeck, 1993). Goodness of fit change indices were used in the CFA models to determine acceptable model change (Change in CFI, Δ CFI, acceptable change that indicated invariance: < 0.01 [Cheung & Rensvold, 2002]; Change in Chi-Squared, Δ χ2, acceptable change that indicated invariance: non-significant p > 0.05 [Byrne, 2013]; Change in RMSEA, Δ RMSEA, acceptable change that indicated invariance: ≤0.01 [Little, 1997]). Once the appropriate structure was confirmed, a multi-group CFA (MGCFA) was conducted to test the invariance of the DBQ across the three time-points to ensure that the DBQ was a reliable measure of the same behaviours at each time-point. Invariance is determined across three phases of analysis. As noted by Byrne (2013), initial models are conducted individually for each time-point and then three further analyses are undertaken, each imposing more restrictions on the model and each reliant upon invariance being confirmed in the previous phase. Phase 1, examines whether the structure of the DBQ is equivalent across each time-point (configural invariance). In this phase, baseline/Year 1, Year 2 and Year 3 models are nested and tested simultaneously, producing one set of goodness of fit values. Once the appropriate model fit is determined, the Phase 2 analysis measuring metric invariance is conducted. Metric invariance examines whether
3.1. Confirmatory factor analysis on the baseline DBQ Initial CFAs were conducted on the 42-item, three-factor model proposed by Obriot-Claudel and Gabaude (2004) (χ2 (776) = 1144.19, p < 0.001; CFI = 0.80, TLI = 0.79, RMSEA = 0.05 [90% CI: 0.04, 0.05, pclose > 0.05]) and the 27-item, three-factor model used by Martinussen et al. (Martinussen et al., 2013) (χ2 (273) = 371.40, p < 0.001; CFI = 0.88, TLI = 0.87, RMSEA = 0.04 (90% CI: 0.03, 0.05, pclose > 0.05). The model proposed by Martinussen et al. showed the best fit to the data although modifications were required to achieve appropriate fit. Items “Disregard red lights when driving late at night along empty roads” and “Take a chance and drive through a red light” were removed from the model as they did not significantly load onto the violation factor. Items “Check your speedometer and discover that you are unknowingly travelling faster than the posted speed limit” and “Angered by another driver’s behaviour, you give chase with the intention of giving him/her a piece of your mind” and “Get involved in unofficial races with other drivers” were also removed from the model for unacceptably low loadings on the violations factor (< 0.20). The final model contained 21-items loading onto three-factors (errors, violations and lapses) and showed acceptable fit (χ2 (186) = 264.39, p < 0.001; CFI = 0.90, TLI = 0.89, RMSEA = 0.04 (90% CI: 0.03, 0.06, pclose > 0.05). Table 1 displays the factor structure, loadings and items’ mean scores across the three time-points. The majority of the item loadings were above 0.40 and all were significant in their models. Lapses were the most frequently reported type of behaviours, where mean scores across the time-points ranged from 1.62 to 1.68 out of a possible 5, suggesting these behaviours happened relatively rarely. The most frequent lapses were: forgetting where the car was parked in the car park (means scores between 1.89 and 1.97) and missing an exit on the highway (mean scores between 1.81 and 1.90). Violations were the second most frequent type of behaviour (mean scores ranged from 1.39 to 1.43) with the most frequent violations being: passing in the righthand lane (i.e., the passing lane in Australia; mean scores between 2.00 to 2.08). Errors were the least frequent type of behaviour reported by older drivers (mean scores ranged between 1.34 to 1.38) with the most frequent error being: misjudging the speed of an oncoming vehicle when passing (means from 1.49 to 1.58) and the least frequent being: 173
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Table 1 Item means (standard deviations) across the three-years. Means (SD)
Factor loadings
DBQ item
Baseline
Year 2
Year 3
Baseline
Year 2
Year 3
Lapses
1.68 (0.40)
CR: 0.67
CR: 0.64
1.97 1.90 1.89 1.56 1.48 1.26
1.62 (0.37) 1.89 (0.72) 1.81 (0.56) 1.74 (0.68) 1.57 (0.66) 1.49 (0.62) 1.22 (0.46)
CR: 0.60
8. Forget where you left your car in a parking lot 14. Miss your exit on a highway and have to make a lengthy detour 38. Get into the wrong lane to make a left or right hand turn at an intersection 17. intending to drive to destination A you wake up and realise you are driving to destination B 10. intend to switch the windshield wipers, but switch on the lights instead, or vice versa 15. Forget which gear you are currently in and you have to check
1.64 (0.39) 1.90 (0.75) 1.85 (0.62) 1.74 (0.64) 1.56 (0.62) 1.53 (0.67) 1.26 (0.49)
0.43 0.51 0.49 0.37 0.51 0.33
0.54 0.47 0.53 0.41 0.48 0.58
0.51 0.52 0.51 0.38 0.50 0.44
Violations
1.43 (0.34)
CR: 0.50
CR: 0.67
2.08 (0.88) 1.87 (1.06) 1.31 (0.57)
1.39 (0.33) 2.00 (0.85) 1.84 (1.02) 1.19 (0.46)
CR: 0.62
4. Become impatient with a slow driver in the right lane and pass in the right lane 5. Driving along country roads, you drive as fast with lights on low beam as you would on high beam 45. Drive with only half-an-eye on the road while looking at a road map, dialling/text messaging on a cell phone, changing a cassette/CD or radio channel 21. Deliberately disregard the speed limits late at night or very early in the morning 7. Drive especially close or flash the car in front as a signal to drive faster or get out of your way 16. Stuck behind a slow moving vehicle on a two-lane highway you are driven to frustration to try to pass under risky circumstances 48. Race oncoming vehicles for a one-car gap on a narrow or obstructed road
1.39 (0.30) 2.11 (0.88) 1.75 (0.98) 1.25 (0.45)
0.42 0.27 0.34
0.50 0.25 0.35
0.54 0.34 0.31
1.24 (0.51) 1.19 (0.48) 1.19 (0.42)
1.20 (0.48) 1.19 (0.45) 1.18 (0.46)
1.17 (0.46) 1.18 (0.45) 1.24 (0.48)
0.48 0.52 0.63
0.38 0.28 0.51
0.42 0.66 0.56
(0.71) (0.64) (0.68) (0.66) (0.65) (0.49)
1.10 (0.35)
1.06 (0.24)
1.08 (0.27)
0.36
0.19
0.47
Errors
1.40 (0.35)
CR: 0.81
CR: 0.82
1.58 (0.62) 1.52 (0.56)
1.38 (0.34) 1.51 (0.56) 1.45 (0.57)
CR 0.79
30. Misjudge the speed of an oncoming vehicle when passing 20. Try to pass a car without first checking in your mirror, and then get honked at by the car behind you which has already begun to pass 11. Turn left on to a main road into the path of an oncoming vehicle that you hadn’t seen or who’s speed you had misjudged 46. Fail to notice pedestrians crossing when turning into a side-street from a main road 41. Fail to check your mirror before pulling out, changing lanes, turning etc. 28. Lost in thought or distracted, you fail to notice someone waiting at a crosswalk light that has just turned red 32. Fail to notice someone stepping out from behind a bus or parked vehicle until it is nearly too late 42. Attempt to pass a vehicle that you hadn’t noticed was signalling its intention to turn right
1.36 (0.34) 1.49 (0.55) 1.43 (0.55)
0.59 0.50
0.56 0.50
0.64 0.57
1.40 (0.54)
1.40 (0.53)
1.39 (0.52)
0.48
0.66
0.67
1.39 (0.54) 1.35 (0.55) 1.33 (0.52)
1.38 (0.51) 1.33 (0.53) 1.33 (0.51)
1.43 (0.52) 1.38 (0.53) 1.35 (0.50)
0.70 0.53 0.74
0.65 0.53 0.70
0.65 0.58 0.52
1.33 (0.52) 1.27 (0.49)
1.28 (0.47) 1.25 (0.45)
1.28 (0.48) 1.21 (0.44)
0.56 0.56
0.60 0.53
0.54 0.53
CR = Composite reliability.
3.3. Latent growth model of changes in the DBQ across time
stuck behind a slow-moving vehicle on a two-lane highway you are driven to frustration to try to pass under risky circumstances (mean scores ranged between 1.19 to 1.24). Fig. 1 shows the factor means plotted across the three time-points. There was very little change in the reported frequency across the three time-points. The observed pattern of mean scores suggests a slight linear decline.
Once measurement invariance was confirmed, a multiple-domain latent growth curve was conducted to explore the degree of change over time, if any, across each DBQ factor (see Fig. 2). The final model showed good fit to the data (χ2 (24) = 73.26, p < 0.001; CFI = 0.97, TLI = 0.96, RMSEA = 0.08 (90% CI: 0.05, 0.10). The means for the intercepts of violations (1.42), errors (1.39) and lapses (1.68) were all significant with lapses being the most frequent behaviour and errors being the least frequent behaviour. The mean values reported in the LCG account for error, which is why the values are marginally different to those reported above. When the slopes were examined, which represent the average change across models, violations (−0.02) and lapses (−0.03) showed a significant average decline, however the degree of change on these factors was almost negligible. Self-reported errors remained stable across the three-time-points. The covariances between intercepts and slopes within each factor were also examined and none of them were statistically significant. This suggests that the change in lapses and violations were independent of the baseline score. In other words, older drivers who reported more frequent lapses at baseline (Year 1) were no more likely to report larger decreases in Years 2 or 3 than those with lower baseline lapses. While no within-domain covariances were observed, the model showed between-domain covariance, with significant and strong relationships between the intercepts of violations and errors (0.63), errors and lapses (0.73) and violations and lapses (0.57), showing that those older drivers who reported more frequent violations at Year 1 also reported more frequent lapses and errors in Years 2 and 3. In addition, the variances for the slopes for each factor were not significant, indicating no interindividual differences in change within each factor.
3.2. Longitudinal confirmatory factor analysis on the DBQ A MGCFA was conducted to confirm whether the DBQ as a measurement tool was invariant across the three time points (see Table 2). Initial models were conducted separately on mean scores for each timepoint, with the 21-item, three-factor model showing good fit to data from each time-point. The configural model, where the three models were nested but not constrained, also demonstrated acceptable fit (see Table 2), showing that the structure of the DBQ remained similar across the three timepoints. The metric model, where the factor loadings were constrained across models, also showed acceptable fit and non-significant Δ χ2 showed that the factor loadings were invariant across the three models. This indicates that the pattern of responses remained similar across the three time-points. Finally, a scalar model was conducted, where the intercepts were also constrained across models. As can be seen in Table 2, strong measurement invariance was met with a non-significant Δ χ2. Therefore, the DBQ remained invariant across the three timepoints.
174
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Fig. 1. DBQ factor mean scores across the three-year period.
4. Discussion
violations factor. Conventionally, violations are defined as deliberate behaviours that increase crash risk (Reason et al., 1990) and therefore, the wording of this item is in contrast to the other deliberate behaviour items included in this factor. Once these modifications were made, the resulting 21-item, three-factor model showed good fit across all three time-points. These findings contribute to the existing body of literature by suggesting an appropriate structure for use of the DBQ in an older population of drivers which will become increasingly important as the population of drivers ages. The current findings are also important because the original DBQ was validated on younger and middle-aged drivers (e.g., aged 20–56 years; Reason et al., 1990) and was not designed specifically for older drivers. This may explain why some researchers have found a less than ideal fit with older driver cohorts (e.g., Mattsen, 2012; Stephens & Fitzharris, 2016). Consistent with Mattsson (2012) who suggested that a number of DBQ items fit within different factors for older drivers, the factor loadings from the current study have demonstrated that some items may be interpreted differently by older drivers. For example, drink driving was interpreted as an ‘error’ rather than a ‘violation’ (according to the pattern of loadings) which suggests that they may drink drive accidently (i.e., the morning after) rather than deliberately or intentionally driving when drinking. Although the frequency of self-reported violations and lapses reported in the current study decreased across the three time-points, this
The current study examined the suitability of the DBQ for use with older drivers and, for the first time, assessed the reliability (specifically, the test-retest reliability) of this newly structured DBQ. This question is important because many older drivers are aware of age-related declines in functional capacities, and may be able to adapt their driving behaviours and/or patterns to match these changes by self-regulating when, where and how they drive, which is a crucial element in maintaining their safety. Our results showed that not only did older drivers’ responses to the DBQ remain stable across the three-time-points, there was relatively little change in the frequency of behaviours reported by the older drivers. These results suggest that the DBQ is a reliable tool to measure older drivers’ behaviour, and highlight that self-perceived driving behaviours remain relatively stable across a relatively short period of time (e.g., 2 years). This is the first study to report on a longitudinal examination of the DBQ in a group of older drivers. However, the structure of the DBQ required a number of modifications before acceptable fit was found and five non-relevant items were removed. These items were related to violation behaviours, and included disregarding red traffic light signals (2 items) and racing other drivers (2 items). One additional item about unintentionally speeding was also removed for not loading onto the
Table 2 Multigroup analysis of measurement invariance across the three-year period.
Baseline (Year 1) Year 2 Year 3 Configural (unconstrained) Metric (factor loadings; weak Measurement Invariance) Scalar (intercepts; strong Measurement Invariance)
χ2
df
CFI
TLI
RMSEA
90% CI
Δ CFI
Δ χ2
Δ df
264.39 217.10 285.71 767.15 801.35 822.57
186 186 186 558 594 606
0.90 0.96 0.90 0.92 0.92 0.92
0.89 0.96 0.88 0.91 0.92 0.91
0.043 0.027 0.049 0.024 0.023 0.023
0.031; 0.001; 0.037; 0.019; 0.018; 0.019;
– – – – < 0.01 < 0.01
– – – – 34.20 55.42
– – – – 36 48
All models were significant (p < 0.001); pclose was > 0.05 in all instances.
175
0.055 0.041 0.060 0.027 0.027 0.027
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Fig. 2. Multiple domain latent growth curve (LGC) on the DBQ across a three-year period. NB: The LGC follows structural equation conventions whereby latent factors (i.e., violations) are represented by circles and observed values (e.g., scores for Year 1) are represented by rectangles.
Accordingly, road safety authorities implement mechanisms, such as cumulative demerit point penalty systems, that seek to alter problem drivers’ behaviour (Goodwin et al., 2015). Finally, motor vehicle insurers are most aware that some drivers are more at risk than other drivers, and adjust drivers’ insurance premiums to reflect their risk profiles. The finding from the current study suggest that older drivers in the Ozcandrive study self-reported fewer errors and fewer violations than what has been reported in previous studies with younger drivers. In terms of the practical implications of these findings, the DBQ was designed to understand motivations behind different self-reported aberrant driving behaviours in order to develop targeted interventions with potential to reduce these driving behaviours and potentially reduce crash risk (i.e., drivers with higher lapse errors could be taught compensatory strategies to support navigation tasks etc.). The different motivations behind different self-reported aberrant driving behaviours should be investigated in other groups of older drivers who have not agreed to participate in a longitudinal study.
decrease was almost negligible, showing these self-reported behaviours to be relatively stable across the period. Indeed, the frequency of selfreported errors remained similar across the same time-period. These findings contrast the results reported by Roman et al. (Roman, et al., 2015) who reported that errors, violations and lapses increased relatively quickly in young novice drivers; highlighting further the distinction between these groups of drivers and the need for age-appropriate measurement tools. By way of example, Roman et al. reported that there was a large amount of variation within the novice young driver group and that changes were also reliant on individual factors such as age and gender as well as the initial starting value. However, for older drivers from the Ozcandrive study, the change in DBQ scores was independent of the starting value (e.g., Year 1) and did not vary within the group, showing that older drivers who remain driving, and in the study, may be a homogenous group of drivers in terms of their selfreported aberrant driving behaviours (e.g., errors, lapses and violations). One possible explanation could be that older drivers actively monitored their self-reported aberrant driving behaviours (i.e., due to the fact that they were participating in a longitudinal study and having to complete annual assessments and questionnaires, and kept driving while they perceive these behaviours remained unchanged (and hence remain in this longitudinal study) or cease driving if they perceived that these behaviours change or increase (however these drivers would not remain in the longitudinal study). Indeed, we did not have a large enough sample to examine the DBQ growth trajectory in drivers who subsequently dropped out of the study due to cessation of driving. However, further research in this area is clearly warranted. In addition, future research will investigate the relationship between older drivers’ driving performance on an on-road driving task (see Koppel et al., 2016, 2017) and self-reported aberrant driving behaviour using the 21-item, three-factor model of the DBQ. The observation that older drivers who reported more frequent violations at Year 1 also reported more frequent lapses and errors in Years 2 and 3 is noteworthy. This lends concurrent validity to the DBQ, as it is consistent with the common observation that driving behaviour of experienced or older drivers is deeply entrenched and resistant to change. Anecdotally, this is appreciated at a common-sense level where acquaintances choose who to drive with and who to avoid based at least in part on their knowledge of the potential chauffeurs’ driving habits. It is also appreciated by various road safety authorities through the observation that drivers with citations for road rule violations are more likely to have higher crash or fatality risk (Elvik et al., 2009).
4.1. Limitations Some limitations should be noted. First, the older drivers from the Ozcandrive study were participating in a larger longitudinal study which was tracking the cohort for five years, including assessing changes in their functional abilities, driving practices (e.g., exposure and patterns), as well as official and self-reported crashes and citations. It is possible that only older drivers who perceive themselves to have low-levels of self-reported aberrant driving behaviour would agree to participate in such a study. Therefore, these findings may not be representative of the general older driver population. Indeed, although almost 70 percent of the participants in current sample were male, recent figures from VicRoads (the licensing authority for the state of Victoria, Australia) have shown that 54 percent of licenced drivers aged 75 years and older were male in 2016 (personal communication, Jonathon Fam, VicRoads January 16th 2018). This is important to note because previous research has shown gender differences in terms of drivers reporting and accepting changes in driving abilities (e.g., Charlton et al., 2006). Thus, the results of the current study that show no or very small changes in self-reported aberrant driving behaviours may be different if the sample contained more female driver. Possible gender differences should be explored in future research. The mean scores obtained during the CFA analysis for DBQ factors of errors, lapses 176
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
and violations were all very low, showing infrequent engagement in each type of behaviour. These infrequent self-reported aberrant driving behaviours are consistent with that reported by Roman et al. with younger drivers (Roman et al., 2015) and are consistent with the findings from older drivers (aged 65–70 years, n = 378) from a representative sample of drivers in Australia (Stephens and Fitzharris, 2016). That is, errors were the least frequently reported behaviours and lapses were the most frequently reported behaviours. Although the scales used by Stephens and Fitzharris and the one used in the current study have different configurations, the mean scores are relatively similar. For example, when the current study is compared to the representative sample, we find that the frequency of reported errors is higher (1.39 compared to 1.31), however, violations (1.42 compared to 1.51) and lapses (1.68 compared to 1.77) are lower. Further studies are needed to sample a range of older drivers, particularly those older drivers with age-related functional declines or medical conditions, to determine whether these older drivers also self-report infrequent aberrant driving behaviours, and also confirm these DBQ factor loadings and this reliability assessment in an independent sample. Secondly, the findings of this study are based on self-reported aberrant driving behaviour. Previous research suggests that participants in research about behaviours that are not socially acceptable tend to minimise the extent of their negative behaviours (Swann et al., 2005). Future research should validate the self-reported responses with more objective measures such as those collected through naturalistic driving study (NDS) methodology or on-road driving tasks.
Blanchard, R., Myers, A., Porter, M., 2010. Correspondence between self-reported and objective measures of driving exposure and patterns in older drivers. Accid. Anal. Prev. 42 (2), 523–529. Browne, M.W., Cudeck, R., 1993. Alternative ways of assessing model fit. Sage Focus Ed. 154, 136. Byrne, B.M., 2013. Structural Equation Modeling With AMOS: Basic Concepts, Applications, and Programming. Routledge. Charlton, J.L., Oxley, J., Fildes, B., Oxley, P., Newstead, S., Koppel, S., O’Hare, M., 2006. Characteristics of older drivers who adopt self-regulatory driving behaviours. Transp. Res. Part F Traffic Psychol. Behav. 9 (5), 363–373. Cheung, G.W., Rensvold, R.B., 2002. Evaluating goodness-of-fit indexes for testing measurement invariance. Struct. Equ. Model 9, 233–255. De Winter, J., Dodou, D., 2010. The driver behaviour questionnaire as a predictor of accidents: a meta-analysis. J. Saf. Res. 41, 463–470. Elvik, R., Vaa, T., Hoye, A., Sorensen, M. (Eds.), 2009. The Handbook of Road Safety Measures. Emerald Group Publishing. Hu, L., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct. Equ. Model 6, 1–55. Gabaude, C., Marquié, J.C., Obriot-Claudel, F., 2010. Self-regulatory driving behaviour in the elderly: relationships with aberrant driving behaviours and perceived abilities. Le travail humain 73 (1), 31–52. Goodwin, A., Kirley, B., Sandt, L., Hall, W., Thomas, L., O’Brien, N., Summerlin, D., 2015. Countermeasures That Work: A Highway Safety Countermeasure Guide for State Highway Safety Offices. National Highway Traffic Safety Administration, Washington, DC. Gras, M.E., Sullman, M.J.M., Cunill, M., Planes, M., Aymerich, M., Font-Mayolas, S., 2006. Spanish drivers and their aberrant driving behaviours. Transp. Res. Part F Traffic Psychol. Behav. 9, 129–137. Koppel, S., Berecki-Gisolf, J., 2015. Car licensing trends of the baby boomer cohort (b. 1946-1965) compared to earlier birth cohorts: effects on the driving population in the state of Victoria. Aust. Traffic Inj. Prev. 16, 657–663. Koppel, S., Charlton, J.L., 2013. Behavioural adaptation and older drivers. In: RudinBrown, C.M., Jamson, S.L. (Eds.), Behavioural Adaptation and Road Safety: Theory, Evidence, and Action. CRC Press. Koppel, S., Charlton, J.L., Langford, J., Di Stefano, M., Macdonald, W., Mazer, B., Gelinas, I., Vrkljan, B., Eliasz, K., Myers, A., Tuokko, H., Marshall, S., 2016. The relationship between older drivers' driving behavior, functional fitness to drive, and self-reported driving-related abilities and practices. Can. J. Aging 35 (S), 15–31. Koppel, S., Charlton, J.L., Richter, N., Di Stefano, M., Macdonald, W., Darzins, P., Newstead, S.V., D’Elia, A., Mazer, B., Gelinas, I., Vrkljan, B., Eliasz, K., Myers, A., Tuokko, H., Marshall, S., 2017. Are older drivers’ on-road driving error rates related to functional performance and/or self-reported driving experiences? Accid. Anal. Prev. 103, 1–9. Koppel, S., Bohensky, M., Langford, J., Taranto, D., 2011. Older drivers, crashes and injuries. Traffic Inj. Prev. 12 (5), 459–467. Langford, J., Koppel, S., 2006. Epidemiology of older driver crashes - identifying older driver risk factors and exposure patterns. Transp. Res. Part F Traffic Psychol. Behav. 9 (5), 309–321. Li, G., Braver, E.R., Chen, L., 2003. Fragility versus excessive crash involvement as determinants of high death rates per vehicle-mile of travel among older drivers. Accid. Anal. Prev. 35, 227–235. Little, T.D., 1997. Mean and covariance structures (MACS) analyses of cross-cultural data: practical and theoretical issues. Multivariate Behav. Res. 32 (1), 53–76. Marshall, S., Man-Son-Hing, M., Bédard, M., Charlton, J.L., Gagnon, S., Gélinas, I., Koppel, S., Korner-Bitensky, N., Langford, J., Mazer, B., Myers, A., Naglie, G., Polgar, J., Porter, M., Rapoport, M., Tuokko, H., Vrkljan, B., Woolnough, A., 2013. Protocol for Candrive II/Ozcandrive, a multicentre prospective older driver cohort study. Accid. Anal. Prev. 61, 245–252. Marshall, S., 2008. The role of reduced fitness to drive due to medical impairments in explaining crashes involving older drivers. Traffic Inj. Prev. 9 (4), 291–298. Martinussen, L., Hakamies-Blomqvist, L., Møller, M., Özkan, T., Lajunen, T., 2013. Age, gender, mileage and the DBQ: the validity of the driver behavior questionnaire in different driver groups. Accid. Anal. Prev. 52, 228–236. Mattsson, M., 2012. Investigating the factorial invariance of the 28-item DBQ across genders and age groups: an exploratory structural equation modeling study. Accid. Anal. Prev. 48, 379–396. Molnar, L.J., Charlton, J.L., Eby, D.W., Langford, J., Koppel, S., Kolenic, G.E., Marshall, S., 2014. Factors affecting self-regulatory driving practices among older adults. Traffic Inj. Prev. 15 (3), 262–272. Molnar, L., Eby, D.W., 2008. The relationship between self-regulation and driving-related abilities in older drivers: an exploratory study. Traffic Inj. Prev. 9 (4), 314–319. Obriot-Claudel, F., Gabaude, C., 2004. The driver behaviour questionnaire: a French study applied to elderly drivers, Nottingham, UK. Paper presented at the International Conference on Traffic and Transport Psychology, 5th – 9th September, 2004. Parker, D., Reason, J.T., Manstead, A.S., Stradling, S.G., 1995. Driving errors, driving violations and accident involvement. Ergonomics 38, 1036–1048. Reason, J., Manstead, A., Stradling, S., Baxter, J., Campbell, K., 1990. Errors and violations on the roads: a real distinction? Ergonomics 33 (10–11), 1315–1332. Rimmö, P.E., Åberg, L., 1999. On the distinction between violations and errors: sensation seeking associations. Transp. Res. Part F Traffic Psychol. Behav. 2, 151–166. Rimmö, P.A., Hakamies-Blomqvist, L., 2002. Older drivers' aberrant driving behaviour, impaired activity, and health as reasons for self-imposed driving limitations. Transp. Res. Part F Traffic Psychol. Behav. 5 (1), 47–62. Roman, G.D., Poulter, D., Barker, E., McKenna, F.P., Rowe, R., 2015. Novice drivers' individual trajectories of driver behavior over the first three years of driving. Accid.
4.2. Summary and practical implications Overall, the results have demonstrated that the DBQ is a reliable tool to measure self-reported aberrant driving behaviour across time, although a number of modifications were required to make it suitable for older drivers. Further, the frequency of self-reported errors remained similar across the period, with only marginal decreases in violations and lapses, suggesting these also remain relatively consistent over time. Future research should validate the self-reported responses with more objective measures such as crashes and citations, as well as data collected through NDS methodology. Acknowledgements The Candrive II study was funded by a Team Grant from the Canadian Institutes of Health Research (CIHR) entitled “The CIHR Team in Driving in Older Persons (Candrive II) Research Program” (grant 90429). Ozcandrive (including the Ozcandrive eDOS subproject) is funded by an Australian Research Council Linkage Grant (LP 100100078) to the Monash University in partnership with La Trobe University, VicRoads, Victorian Government Department of Justice and Victoria Police, the Transport Accident Commission, Road Safety Trust New Zealand and Eastern Health. The authors thank Lorraine Atkinson, Ozcandrive Program Manager, for her role in managing and operationalizing the study for the Australian and New Zealand sites. The authors gratefully acknowledge the invaluable contribution of the Ozcandrive team, including: Amy Allen, Louise Beasley, Russ Boag, Matthew Catchlove, Cara Dawson, Johan Davydov, Lei Gryffydd, YikXiang Hue, Elizabeth Jacobs, Duncan Joiner, Jason Manakis, Kevin Mascarenhas, Emma Owen and Jarrod Verity. The authors also thank the Ozcandrive cohort study participants for their dedication. Without their commitment, this publication would not have been possible. References Austroads, 2010. Assessing Fitness to Drive for Commerical and Private Vehicle Drivers. Austroads, Sydney, Australia. Baldock, M.R.J., Mathias, J., McLean, J., Berndt, A., 2006. Self-regulation of driving and older Drivers' functional abilities. Clin. Gerontol. 30 (1), 53–70.
177
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Countries. University of Michigan Transportation Research Institute, Ann Arbor, MI. Stephens, A.N., Fitzharris, M., 2016. Validation of the driver behaviour questionnaire in a representative sample of drivers in Australia. Accid. Anal. Prev. 86, 186–198. Swann, A.A., Matthews, C.E., Ebbeling, C.B., Moore, C.G., Cunningham, J.E., Fulton, J., Hebert, J.R., 2005. The effect of social desirability and social approval on self-reports of physical activity. Am. J. Epidemiol. 161, 389–398.
Anal. Prev. 82, 61–69. Singh, S., 2015. Critical Reasons for Crashes Investigated in the National motor Vehicle Crash Causation Survey (Traffic Safety Facts Crash Stats. Report No. DOT HS 812 115). National Highway Traffic Safety Administration, Washington, DC, United States. Sivak, M., Schoettle, B., 2011. Recent Changes in the Age Composition of Drivers in 15
178
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
The Driver Behaviour Questionnaire for older drivers: Do errors, violations and lapses change over time?
T
⁎
Koppel S.a, , Stephens A.N.a, Charlton J.L.a, Di Stefano M.b, Darzins P.c,d, Odell M.e, Marshall S.f a
Monash University Accident Research Centre, Monash University, Australia La Trobe University, Australia c Eastern Health, Australia d Monash University Eastern Health Clinical School, Australia e Victorian Institute of Forensic Medicine, Australia f Ottawa Hospital Research Institute, Canada b
A R T I C L E I N F O
A B S T R A C T
Keywords: Older drivers Aberrant driving behaviour Errors Violations Lapses Road safety
The aim of the current study was to examine how self-reported aberrant driving behaviours change across a three time-points in a group of older drivers. Two hundred and twenty-seven older drivers (males = 69.6%) from the Candrive/Ozcandrive longitudinal study completed the Driving Behaviour Questionnaire (DBQ) each yearacross three time-points (i.e., Year 1, Year 2, Year 3). At the third time-point, older drivers ranged in age from 77 to 96 years (M = 81.74 years; SD = 3.44 years). A longitudinal confirmatory factor analysis showed that a modified 21-item, 3-factor (errors, lapses and violations) DBQ was invariant across the time period, suggesting that the structure of the questionnaire was stable across each time-point. Further, multiple domain latent growth analysis on the resultant factors for errors, lapses and violations showed that the frequency of errors remained similar across the three-year period, while violations and lapses showed very marginal decreases in frequency. These changes were independent of the absolute number of these behaviours; Drivers with higher violations or lapses in Year one, showed similar decreases in frequency as those who self-reported lower frequencies of the behaviours. These results suggest that the DBQ is a reliable tool to measure older drivers’ self-reported aberrant driving behaviours, and that these behaviours do not show much change across time. Future research should validate the self-reported responses from the DBQ with more objective measures such as those collected through naturalistic driving study (NDS) methodology or on-road driving tasks.
1. Introduction It is important to understand how drivers’ behaviours may change over time, particularly for older drivers who are likely to form a larger proportion of the driving fleet as the population ages (Koppel and Berecki-Gisolf, 2015; Sivak and Schoettle, 2011). Older drivers are over-represented in fatal and serious injury crash statistics (Koppel et al., 2011; Langford and Koppel, 2006). While a large part of this over-representation is due to the frailty of older drivers, with around 60 to 90 percent of fatalities in this age category resulting from driver frailty (Li et al., 2003), age-related sensory, cognitive, and physical impairments also contribute (Marshall, 2008). However, previous research has shown that many older drivers become aware of such declines and change their driving patterns accordingly by self-regulating when, where and how they drive (Baldock et al., 2006; Blanchard et al., 2010; Charlton et al., 2006; Molnar and Eby, 2008; Molnar et al., 2014).
⁎
Appropriate self-regulation therefore requires recognition of age-related, and possibly driving, decline, particularly types of driving behaviours that may increase crash risk (Koppel and Charlton, 2013). One of the main threats to road safety is aberrant driving behaviour (Singh, 2015). Aberrant driving behaviours are often defined as driving errors, lapses and violations, with common examples including: exceeding the speed limit, red-light running, and heavy braking due to poor situation awareness (De Winter and Dodou, 2010; Gabaude et al., 2010). Associations have been found between self-reported aberrant driving behaviour and crash involvement (Gras et al., 2006; Parker et al., 1995; Rimmö and Åberg, 1999), with one recent study identifying some form of aberrant driving behaviours in approximately 74 percent of crashes (Singh, 2015). The Driving Behaviour Questionnaire (DBQ), initially developed by Reason et al. (1990), is one of the most widely used research tools to measure the frequency of self-reported aberrant driving behaviour.
Corresponding author. E-mail address: [email protected] (S. Koppel).
https://doi.org/10.1016/j.aap.2018.01.036 Received 17 October 2017; Received in revised form 26 January 2018; Accepted 28 January 2018 0001-4575/ © 2018 Elsevier Ltd. All rights reserved.
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Traditionally, the DBQ represents 50 individual behaviours that fit into three broad behaviour patterns: violations, errors and lapses. Violations are defined as deliberate behaviours that directly contravene road laws, for example exceeding the speed limit or failing to obey red traffic light signals. Errors are defined as behaviours that do not contravene road laws directly but, as with violations, are considered to increase a driver’s risk of crash. In contrast, lapses are defined as minor mistakes that are not considered to be associated with crash involvement (Reason, et al., 1990). While there are now a number of versions of the DBQ, which target specific groups of drivers, few studies have used this questionnaire specifically in an older population of drivers. Indeed, the original DBQ was validated on drivers ranging in age from 20 to 56 years (Reason, et al., 1990) and has mostly been used for drivers in that age range (De Winter and Dodou, 2010). Stephens and Fitzharris (2016) have shown that a four-factor version of the DBQ, based on the findings of Parker et al. (Parker et al., 1995) and that includes a factor for aggressive violations, is appropriate for a broad range of drivers, but less so when specifically examining older drivers. In line with this, Mattsson (2012) has shown that a number of DBQ items fit within different factors for older drivers. For example, behaviours that younger drivers may interpret as a violation are considered to be an error by older drivers (e.g., drink driving). Rimmö and Hakamies-Blomqvist (2002) have also studied the relationship between driving exposure, health, and four types of self-reported aberrant driving behaviours (e.g., inattention, inexperience errors, violations and mistakes) using a Swedish version of the DBQ with Swedish drivers aged between 55 and 92 years. They reported that, even after accounting for age and gender, self-reported inattention and inexperience errors, as well as impaired health, were related to self-imposed driving limitations, whereas self-reported violations and mistakes were not. Given the importance of understanding factors that increase crash risk in older drivers and identifying older drivers’ perceived level of violations, errors and lapses, it is necessary to establish an appropriate tool to measure perceived driving behaviour in this cohort. In doing this, the number and type of changes that may occur over time can then be examined. Some studies have focussed on deriving an appropriate factor structure for older drivers. Obriot-Claudel and Gabaude (2004) suggest that a three-factor, 42-item version of the DBQ is appropriate for use in drivers aged between 55 and 91 years. They suggest that these items, taken from the original 50-items, explain three main types of aberrant behaviours: inattention errors, dangerous errors and dangerous violations. Likewise, Martinussen et al. (Martinussen et al., 2013) also suggest a three-factor version of the DBQ is appropriate for drivers aged 50 to 85 years, with this version using 27 of the 50-items, and explaining factors for violations, errors and lapses. Indeed, Martinussen et al. reported that this version was the best fit for older drivers (i.e., 50–85 years) compared to middle-aged (i.e., 30–49 years) and younger drivers (i.e., 18–29 years). However, the item factor loadings of these two versions (i.e., Obriot-Claudel and Gabaude, 2004; Martinussen et al., 2013) are markedly different and therefore the degree to which each is appropriate in other samples is unclear. We need to establish an appropriate tool to measure behaviours so as to understand how these change over time, particularly in a group of vulnerable drivers who are likely to monitor their perceived driving behaviour, and possible declines, in order to self-regulate driving behaviour. To date, there is little research on how aberrant behaviours measured with the DBQ may change over time. Roman et al. (2015) investigated changes in DBQ scores for errors, slips, violations and aggressive violations across six monthly intervals in novice young drivers and found that scores for each increased across the time-points. However, despite the fact that changes in older drivers’ self-reported aberrant driving behaviour could be a crucial element in maintaining their safety, no similar research has been conducted with older drivers. Consequently, the aim of the current study was twofold; first to confirm
an appropriate structure of the DBQ for use in a sample of older drivers; and, second to assess the reliability (specifically, the test-retest reliability) of the newly structured DBQ. 2. Method 2.1. Participants Data from 227 older drivers (males = 69.6%) are presented in the current paper. Each driver was an Ozcandrive participant in the Candrive/Ozcandrive study (described in Section 2.2). Ozcandrive participants completed annual assessments across five time-points in Melbourne, Australia. All participants were required to meet the following inclusion criteria: a) aged 75 years or older; b) held a valid driver’s license; c) drove at least four times per week, and d) did not have an absolute contraindication to driving, as defined by the Austroads Fitness to Drive Guidelines (Austroads, 2010). Given drop-out rates across the fivetime-points (Year 1: n = 14; Year 2: n = 15; Year 3: n = 11; Year 4: n = 18; Year 5: n = 18)1, and to maintain an appropriate sample size (e.g., n > 200), data from participants who completed the first three assessments of study are presented in the current study. At the third time-point of the study, participants ranged in age from 77 to 96 years (M = 81.74 years; SD = 3.44 years). It should be noted that participants who remained in the study in Year 3 (n = 227) were not significantly different to participants who withdrew from the study (n = 30) in terms of their gender (Male: 69.6% vs. 80.0%, respectively, X2(1) = 1.386, p > 0.1), their Year 1 age (M = 79.66 years, SD = 3.45 vs. 80.33, SD = 3.97, respectively, t(225) = 0.992, p > 0.1), their Year 1 self-reported annual kilometres driven (< 5000 km: 15.0% vs. 20.0%; 5,001–15,000 km: 71.4% vs. 70.0%; > 15,0001 km: 13.7% vs. 10.0%, respectively, X2(2) = 0.704, p > 0.5) or their Year 1 self-reported total number of medical conditions (M = 10.52, SD = 4.13 vs. M = 11.67, SD = 4.58, respectively, t (255) = 1.405, p > 0.1). 2.2. The Candrive/Ozcandrive study The Candrive/Ozcandrive study is a longitudinal, multi-centre international research program with the core objective of identifying solutions to promote older drivers’ safe mobility (Marshall et al., 2013). The Candrive/Ozcandrive study involves 928 drivers aged 70 years and over in Canada and 302 drivers aged 75 years and older in Australia and New Zealand (Australia: n = 257; New Zealand: n = 45). Using a longitudinal study design, the project is tracking this cohort of older drivers for five years, assessing changes in their functional abilities, driving practices (e.g. exposure and patterns), as well as crashes and citations for violations of driving regulations. The primary purpose is to develop and validate a risk stratification tool to assist clinicians in identifying potentially at-risk drivers (Marshall et al., 2013). Participants’ usual (or naturalistic) driving practices (e.g., trip distance, duration, type of road, speed) are recorded through in-car recording devices (ICRD) installed in participants’ own vehicles, and measures of participants’ functional ability, medical conditions and self-reported driving-related abilities and practices are documented annually. 2.3. Procedure and materials Ethical Approval was obtained from the Monash University Human Research Ethics Committee (MUHREC), and all participants provided written informed consent. All participants underwent a baseline (Year 1) and two annual assessments (Years 2 and 3). The time between annual assessments was 1 Note some participants withdrew from the study after completing their annual assessment.
172
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
the factor loadings are consistent across the three models and this is achieved by constraining the loadings to be equal across the three models. Metric invariance determines whether similar patterns of responses are apparent across the time-points. As noted by Byrne (2013) and Cheung and Rensvold (2002), invariance is supported by a nonsignificant change in Chi-Square (from the configural model) and a change in the CFI no greater than 0.01. Once the equivalence of factor loadings is established, structural invariance is examined (Phase 3). Structural invariance requires the intercepts for each item on its factor to be constrained across models and demonstrates that the interpretations of items remain similar across the three time-points. After establishing the longitudinal invariance of the DBQ, a multiple domain latent growth curve (LGC) analysis was conducted. While the MGCFA determines reliability of the DBQ and associated factors, the LGC assesses any mean differences on factor scores across each time point. LGC analysis shows both group change and inter-individual change across the three time-points. Multiple domains consider change across the three factors in the same model. LGC examines the average intercepts for each factor (e.g., the average score at baseline/Year 1) and the degree of change represented by the regression slope over the three models. Inter-individual change is examined by the variance of the factors. To interpret the change, the LGC requires good model fit, determined using the same criteria listed in above and change is indicated through significance levels of p < 0.05.
approximately 12 months for each participant. Each assessment incorporated a range of demographic and driving history questions as well as a range of functional ability measures, reporting on medical conditions, and self-reported abilities and practices related to driving (Marshall et al., 2013). The DBQ (Reason, et al., 1990) was administered across the three time-points. The DBQ contains 50 items measuring the frequency of driving behaviours that are likely to increase crash risk. However, given that the study protocol was designed for older drivers in Canada (Candrive) as well as older drivers in Australia (Ozcandrive), two items relating to behaviour at roundabouts were removed as road infrastructure in Canada does not include roundabouts. In addition, given that most participants drove an automatic vehicle, the item relating to driving off in third gear was also removed. The final version of the DBQ contained 47-items. For each item, frequency was reported across a six-point scale of engagement (where: 1 = Never, 6 = Nearly all the time). ObriotClaudel and Gabaude (2004) report that the three factors appropriate for older drivers demonstrate good internal consistency, with Cronbach alphas of 0.85 (inattention errors), 0.84 (dangerous errors), 0.70 (dangerous violations). Martinussen et al. (Martinussen, et al., 2013) also report good internal consistency on their three-factor version (Cronbach alphas ranging from 0.64 to 0.85). 2.4. Data handling
3. Results
Missing DBQ responses were replaced with a 5.0% trimmed mean and this occurred in only four instances. Analyses were conducted using SPSS v.23 and AMOS v.22. An initial Confirmatory Factor Analysis (CFA) was performed on the baseline data to determine the most appropriate fitting model. Because an existing factor structure was being tested, CFA was chosen over Exploratory Factor Analysis (EFA) as the former allows a strict examination of the previously proposed model fit, taking into account measurement error within each factor loading. CFA also allows nested models which are imperative for longitudinal analyses. Goodness of fit indices to determine acceptable fit included: ChiSquared, χ2, acceptable value: p > 0.05, however significance is common with large sample sizes (Byrne, 2013); Comparative Fit Index, CFI, acceptable value: ≥0.90 where ≥0.95 indicates exceptional fit (Hu and Bentler, 1999); Tucker-Lewis Index, TLI, acceptable value: ≥0.90 where ≥0.95 indicates exceptional fit, (Hu and Bentler, 1999); Root Mean Square Error of Approximation, RMSEA, acceptable value: ≤0.06 (Browne and Cudeck, 1993); Significance value of the RMSEA, pclose, acceptable value: > 0.05 (Hu and Bentler, 1999); Confidence Interval around the RMSEA, 90% CI, acceptable value: maximum value not exceeding 0.06 (Browne and Cudeck, 1993). Goodness of fit change indices were used in the CFA models to determine acceptable model change (Change in CFI, Δ CFI, acceptable change that indicated invariance: < 0.01 [Cheung & Rensvold, 2002]; Change in Chi-Squared, Δ χ2, acceptable change that indicated invariance: non-significant p > 0.05 [Byrne, 2013]; Change in RMSEA, Δ RMSEA, acceptable change that indicated invariance: ≤0.01 [Little, 1997]). Once the appropriate structure was confirmed, a multi-group CFA (MGCFA) was conducted to test the invariance of the DBQ across the three time-points to ensure that the DBQ was a reliable measure of the same behaviours at each time-point. Invariance is determined across three phases of analysis. As noted by Byrne (2013), initial models are conducted individually for each time-point and then three further analyses are undertaken, each imposing more restrictions on the model and each reliant upon invariance being confirmed in the previous phase. Phase 1, examines whether the structure of the DBQ is equivalent across each time-point (configural invariance). In this phase, baseline/Year 1, Year 2 and Year 3 models are nested and tested simultaneously, producing one set of goodness of fit values. Once the appropriate model fit is determined, the Phase 2 analysis measuring metric invariance is conducted. Metric invariance examines whether
3.1. Confirmatory factor analysis on the baseline DBQ Initial CFAs were conducted on the 42-item, three-factor model proposed by Obriot-Claudel and Gabaude (2004) (χ2 (776) = 1144.19, p < 0.001; CFI = 0.80, TLI = 0.79, RMSEA = 0.05 [90% CI: 0.04, 0.05, pclose > 0.05]) and the 27-item, three-factor model used by Martinussen et al. (Martinussen et al., 2013) (χ2 (273) = 371.40, p < 0.001; CFI = 0.88, TLI = 0.87, RMSEA = 0.04 (90% CI: 0.03, 0.05, pclose > 0.05). The model proposed by Martinussen et al. showed the best fit to the data although modifications were required to achieve appropriate fit. Items “Disregard red lights when driving late at night along empty roads” and “Take a chance and drive through a red light” were removed from the model as they did not significantly load onto the violation factor. Items “Check your speedometer and discover that you are unknowingly travelling faster than the posted speed limit” and “Angered by another driver’s behaviour, you give chase with the intention of giving him/her a piece of your mind” and “Get involved in unofficial races with other drivers” were also removed from the model for unacceptably low loadings on the violations factor (< 0.20). The final model contained 21-items loading onto three-factors (errors, violations and lapses) and showed acceptable fit (χ2 (186) = 264.39, p < 0.001; CFI = 0.90, TLI = 0.89, RMSEA = 0.04 (90% CI: 0.03, 0.06, pclose > 0.05). Table 1 displays the factor structure, loadings and items’ mean scores across the three time-points. The majority of the item loadings were above 0.40 and all were significant in their models. Lapses were the most frequently reported type of behaviours, where mean scores across the time-points ranged from 1.62 to 1.68 out of a possible 5, suggesting these behaviours happened relatively rarely. The most frequent lapses were: forgetting where the car was parked in the car park (means scores between 1.89 and 1.97) and missing an exit on the highway (mean scores between 1.81 and 1.90). Violations were the second most frequent type of behaviour (mean scores ranged from 1.39 to 1.43) with the most frequent violations being: passing in the righthand lane (i.e., the passing lane in Australia; mean scores between 2.00 to 2.08). Errors were the least frequent type of behaviour reported by older drivers (mean scores ranged between 1.34 to 1.38) with the most frequent error being: misjudging the speed of an oncoming vehicle when passing (means from 1.49 to 1.58) and the least frequent being: 173
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Table 1 Item means (standard deviations) across the three-years. Means (SD)
Factor loadings
DBQ item
Baseline
Year 2
Year 3
Baseline
Year 2
Year 3
Lapses
1.68 (0.40)
CR: 0.67
CR: 0.64
1.97 1.90 1.89 1.56 1.48 1.26
1.62 (0.37) 1.89 (0.72) 1.81 (0.56) 1.74 (0.68) 1.57 (0.66) 1.49 (0.62) 1.22 (0.46)
CR: 0.60
8. Forget where you left your car in a parking lot 14. Miss your exit on a highway and have to make a lengthy detour 38. Get into the wrong lane to make a left or right hand turn at an intersection 17. intending to drive to destination A you wake up and realise you are driving to destination B 10. intend to switch the windshield wipers, but switch on the lights instead, or vice versa 15. Forget which gear you are currently in and you have to check
1.64 (0.39) 1.90 (0.75) 1.85 (0.62) 1.74 (0.64) 1.56 (0.62) 1.53 (0.67) 1.26 (0.49)
0.43 0.51 0.49 0.37 0.51 0.33
0.54 0.47 0.53 0.41 0.48 0.58
0.51 0.52 0.51 0.38 0.50 0.44
Violations
1.43 (0.34)
CR: 0.50
CR: 0.67
2.08 (0.88) 1.87 (1.06) 1.31 (0.57)
1.39 (0.33) 2.00 (0.85) 1.84 (1.02) 1.19 (0.46)
CR: 0.62
4. Become impatient with a slow driver in the right lane and pass in the right lane 5. Driving along country roads, you drive as fast with lights on low beam as you would on high beam 45. Drive with only half-an-eye on the road while looking at a road map, dialling/text messaging on a cell phone, changing a cassette/CD or radio channel 21. Deliberately disregard the speed limits late at night or very early in the morning 7. Drive especially close or flash the car in front as a signal to drive faster or get out of your way 16. Stuck behind a slow moving vehicle on a two-lane highway you are driven to frustration to try to pass under risky circumstances 48. Race oncoming vehicles for a one-car gap on a narrow or obstructed road
1.39 (0.30) 2.11 (0.88) 1.75 (0.98) 1.25 (0.45)
0.42 0.27 0.34
0.50 0.25 0.35
0.54 0.34 0.31
1.24 (0.51) 1.19 (0.48) 1.19 (0.42)
1.20 (0.48) 1.19 (0.45) 1.18 (0.46)
1.17 (0.46) 1.18 (0.45) 1.24 (0.48)
0.48 0.52 0.63
0.38 0.28 0.51
0.42 0.66 0.56
(0.71) (0.64) (0.68) (0.66) (0.65) (0.49)
1.10 (0.35)
1.06 (0.24)
1.08 (0.27)
0.36
0.19
0.47
Errors
1.40 (0.35)
CR: 0.81
CR: 0.82
1.58 (0.62) 1.52 (0.56)
1.38 (0.34) 1.51 (0.56) 1.45 (0.57)
CR 0.79
30. Misjudge the speed of an oncoming vehicle when passing 20. Try to pass a car without first checking in your mirror, and then get honked at by the car behind you which has already begun to pass 11. Turn left on to a main road into the path of an oncoming vehicle that you hadn’t seen or who’s speed you had misjudged 46. Fail to notice pedestrians crossing when turning into a side-street from a main road 41. Fail to check your mirror before pulling out, changing lanes, turning etc. 28. Lost in thought or distracted, you fail to notice someone waiting at a crosswalk light that has just turned red 32. Fail to notice someone stepping out from behind a bus or parked vehicle until it is nearly too late 42. Attempt to pass a vehicle that you hadn’t noticed was signalling its intention to turn right
1.36 (0.34) 1.49 (0.55) 1.43 (0.55)
0.59 0.50
0.56 0.50
0.64 0.57
1.40 (0.54)
1.40 (0.53)
1.39 (0.52)
0.48
0.66
0.67
1.39 (0.54) 1.35 (0.55) 1.33 (0.52)
1.38 (0.51) 1.33 (0.53) 1.33 (0.51)
1.43 (0.52) 1.38 (0.53) 1.35 (0.50)
0.70 0.53 0.74
0.65 0.53 0.70
0.65 0.58 0.52
1.33 (0.52) 1.27 (0.49)
1.28 (0.47) 1.25 (0.45)
1.28 (0.48) 1.21 (0.44)
0.56 0.56
0.60 0.53
0.54 0.53
CR = Composite reliability.
3.3. Latent growth model of changes in the DBQ across time
stuck behind a slow-moving vehicle on a two-lane highway you are driven to frustration to try to pass under risky circumstances (mean scores ranged between 1.19 to 1.24). Fig. 1 shows the factor means plotted across the three time-points. There was very little change in the reported frequency across the three time-points. The observed pattern of mean scores suggests a slight linear decline.
Once measurement invariance was confirmed, a multiple-domain latent growth curve was conducted to explore the degree of change over time, if any, across each DBQ factor (see Fig. 2). The final model showed good fit to the data (χ2 (24) = 73.26, p < 0.001; CFI = 0.97, TLI = 0.96, RMSEA = 0.08 (90% CI: 0.05, 0.10). The means for the intercepts of violations (1.42), errors (1.39) and lapses (1.68) were all significant with lapses being the most frequent behaviour and errors being the least frequent behaviour. The mean values reported in the LCG account for error, which is why the values are marginally different to those reported above. When the slopes were examined, which represent the average change across models, violations (−0.02) and lapses (−0.03) showed a significant average decline, however the degree of change on these factors was almost negligible. Self-reported errors remained stable across the three-time-points. The covariances between intercepts and slopes within each factor were also examined and none of them were statistically significant. This suggests that the change in lapses and violations were independent of the baseline score. In other words, older drivers who reported more frequent lapses at baseline (Year 1) were no more likely to report larger decreases in Years 2 or 3 than those with lower baseline lapses. While no within-domain covariances were observed, the model showed between-domain covariance, with significant and strong relationships between the intercepts of violations and errors (0.63), errors and lapses (0.73) and violations and lapses (0.57), showing that those older drivers who reported more frequent violations at Year 1 also reported more frequent lapses and errors in Years 2 and 3. In addition, the variances for the slopes for each factor were not significant, indicating no interindividual differences in change within each factor.
3.2. Longitudinal confirmatory factor analysis on the DBQ A MGCFA was conducted to confirm whether the DBQ as a measurement tool was invariant across the three time points (see Table 2). Initial models were conducted separately on mean scores for each timepoint, with the 21-item, three-factor model showing good fit to data from each time-point. The configural model, where the three models were nested but not constrained, also demonstrated acceptable fit (see Table 2), showing that the structure of the DBQ remained similar across the three timepoints. The metric model, where the factor loadings were constrained across models, also showed acceptable fit and non-significant Δ χ2 showed that the factor loadings were invariant across the three models. This indicates that the pattern of responses remained similar across the three time-points. Finally, a scalar model was conducted, where the intercepts were also constrained across models. As can be seen in Table 2, strong measurement invariance was met with a non-significant Δ χ2. Therefore, the DBQ remained invariant across the three timepoints.
174
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Fig. 1. DBQ factor mean scores across the three-year period.
4. Discussion
violations factor. Conventionally, violations are defined as deliberate behaviours that increase crash risk (Reason et al., 1990) and therefore, the wording of this item is in contrast to the other deliberate behaviour items included in this factor. Once these modifications were made, the resulting 21-item, three-factor model showed good fit across all three time-points. These findings contribute to the existing body of literature by suggesting an appropriate structure for use of the DBQ in an older population of drivers which will become increasingly important as the population of drivers ages. The current findings are also important because the original DBQ was validated on younger and middle-aged drivers (e.g., aged 20–56 years; Reason et al., 1990) and was not designed specifically for older drivers. This may explain why some researchers have found a less than ideal fit with older driver cohorts (e.g., Mattsen, 2012; Stephens & Fitzharris, 2016). Consistent with Mattsson (2012) who suggested that a number of DBQ items fit within different factors for older drivers, the factor loadings from the current study have demonstrated that some items may be interpreted differently by older drivers. For example, drink driving was interpreted as an ‘error’ rather than a ‘violation’ (according to the pattern of loadings) which suggests that they may drink drive accidently (i.e., the morning after) rather than deliberately or intentionally driving when drinking. Although the frequency of self-reported violations and lapses reported in the current study decreased across the three time-points, this
The current study examined the suitability of the DBQ for use with older drivers and, for the first time, assessed the reliability (specifically, the test-retest reliability) of this newly structured DBQ. This question is important because many older drivers are aware of age-related declines in functional capacities, and may be able to adapt their driving behaviours and/or patterns to match these changes by self-regulating when, where and how they drive, which is a crucial element in maintaining their safety. Our results showed that not only did older drivers’ responses to the DBQ remain stable across the three-time-points, there was relatively little change in the frequency of behaviours reported by the older drivers. These results suggest that the DBQ is a reliable tool to measure older drivers’ behaviour, and highlight that self-perceived driving behaviours remain relatively stable across a relatively short period of time (e.g., 2 years). This is the first study to report on a longitudinal examination of the DBQ in a group of older drivers. However, the structure of the DBQ required a number of modifications before acceptable fit was found and five non-relevant items were removed. These items were related to violation behaviours, and included disregarding red traffic light signals (2 items) and racing other drivers (2 items). One additional item about unintentionally speeding was also removed for not loading onto the
Table 2 Multigroup analysis of measurement invariance across the three-year period.
Baseline (Year 1) Year 2 Year 3 Configural (unconstrained) Metric (factor loadings; weak Measurement Invariance) Scalar (intercepts; strong Measurement Invariance)
χ2
df
CFI
TLI
RMSEA
90% CI
Δ CFI
Δ χ2
Δ df
264.39 217.10 285.71 767.15 801.35 822.57
186 186 186 558 594 606
0.90 0.96 0.90 0.92 0.92 0.92
0.89 0.96 0.88 0.91 0.92 0.91
0.043 0.027 0.049 0.024 0.023 0.023
0.031; 0.001; 0.037; 0.019; 0.018; 0.019;
– – – – < 0.01 < 0.01
– – – – 34.20 55.42
– – – – 36 48
All models were significant (p < 0.001); pclose was > 0.05 in all instances.
175
0.055 0.041 0.060 0.027 0.027 0.027
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Fig. 2. Multiple domain latent growth curve (LGC) on the DBQ across a three-year period. NB: The LGC follows structural equation conventions whereby latent factors (i.e., violations) are represented by circles and observed values (e.g., scores for Year 1) are represented by rectangles.
Accordingly, road safety authorities implement mechanisms, such as cumulative demerit point penalty systems, that seek to alter problem drivers’ behaviour (Goodwin et al., 2015). Finally, motor vehicle insurers are most aware that some drivers are more at risk than other drivers, and adjust drivers’ insurance premiums to reflect their risk profiles. The finding from the current study suggest that older drivers in the Ozcandrive study self-reported fewer errors and fewer violations than what has been reported in previous studies with younger drivers. In terms of the practical implications of these findings, the DBQ was designed to understand motivations behind different self-reported aberrant driving behaviours in order to develop targeted interventions with potential to reduce these driving behaviours and potentially reduce crash risk (i.e., drivers with higher lapse errors could be taught compensatory strategies to support navigation tasks etc.). The different motivations behind different self-reported aberrant driving behaviours should be investigated in other groups of older drivers who have not agreed to participate in a longitudinal study.
decrease was almost negligible, showing these self-reported behaviours to be relatively stable across the period. Indeed, the frequency of selfreported errors remained similar across the same time-period. These findings contrast the results reported by Roman et al. (Roman, et al., 2015) who reported that errors, violations and lapses increased relatively quickly in young novice drivers; highlighting further the distinction between these groups of drivers and the need for age-appropriate measurement tools. By way of example, Roman et al. reported that there was a large amount of variation within the novice young driver group and that changes were also reliant on individual factors such as age and gender as well as the initial starting value. However, for older drivers from the Ozcandrive study, the change in DBQ scores was independent of the starting value (e.g., Year 1) and did not vary within the group, showing that older drivers who remain driving, and in the study, may be a homogenous group of drivers in terms of their selfreported aberrant driving behaviours (e.g., errors, lapses and violations). One possible explanation could be that older drivers actively monitored their self-reported aberrant driving behaviours (i.e., due to the fact that they were participating in a longitudinal study and having to complete annual assessments and questionnaires, and kept driving while they perceive these behaviours remained unchanged (and hence remain in this longitudinal study) or cease driving if they perceived that these behaviours change or increase (however these drivers would not remain in the longitudinal study). Indeed, we did not have a large enough sample to examine the DBQ growth trajectory in drivers who subsequently dropped out of the study due to cessation of driving. However, further research in this area is clearly warranted. In addition, future research will investigate the relationship between older drivers’ driving performance on an on-road driving task (see Koppel et al., 2016, 2017) and self-reported aberrant driving behaviour using the 21-item, three-factor model of the DBQ. The observation that older drivers who reported more frequent violations at Year 1 also reported more frequent lapses and errors in Years 2 and 3 is noteworthy. This lends concurrent validity to the DBQ, as it is consistent with the common observation that driving behaviour of experienced or older drivers is deeply entrenched and resistant to change. Anecdotally, this is appreciated at a common-sense level where acquaintances choose who to drive with and who to avoid based at least in part on their knowledge of the potential chauffeurs’ driving habits. It is also appreciated by various road safety authorities through the observation that drivers with citations for road rule violations are more likely to have higher crash or fatality risk (Elvik et al., 2009).
4.1. Limitations Some limitations should be noted. First, the older drivers from the Ozcandrive study were participating in a larger longitudinal study which was tracking the cohort for five years, including assessing changes in their functional abilities, driving practices (e.g., exposure and patterns), as well as official and self-reported crashes and citations. It is possible that only older drivers who perceive themselves to have low-levels of self-reported aberrant driving behaviour would agree to participate in such a study. Therefore, these findings may not be representative of the general older driver population. Indeed, although almost 70 percent of the participants in current sample were male, recent figures from VicRoads (the licensing authority for the state of Victoria, Australia) have shown that 54 percent of licenced drivers aged 75 years and older were male in 2016 (personal communication, Jonathon Fam, VicRoads January 16th 2018). This is important to note because previous research has shown gender differences in terms of drivers reporting and accepting changes in driving abilities (e.g., Charlton et al., 2006). Thus, the results of the current study that show no or very small changes in self-reported aberrant driving behaviours may be different if the sample contained more female driver. Possible gender differences should be explored in future research. The mean scores obtained during the CFA analysis for DBQ factors of errors, lapses 176
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
and violations were all very low, showing infrequent engagement in each type of behaviour. These infrequent self-reported aberrant driving behaviours are consistent with that reported by Roman et al. with younger drivers (Roman et al., 2015) and are consistent with the findings from older drivers (aged 65–70 years, n = 378) from a representative sample of drivers in Australia (Stephens and Fitzharris, 2016). That is, errors were the least frequently reported behaviours and lapses were the most frequently reported behaviours. Although the scales used by Stephens and Fitzharris and the one used in the current study have different configurations, the mean scores are relatively similar. For example, when the current study is compared to the representative sample, we find that the frequency of reported errors is higher (1.39 compared to 1.31), however, violations (1.42 compared to 1.51) and lapses (1.68 compared to 1.77) are lower. Further studies are needed to sample a range of older drivers, particularly those older drivers with age-related functional declines or medical conditions, to determine whether these older drivers also self-report infrequent aberrant driving behaviours, and also confirm these DBQ factor loadings and this reliability assessment in an independent sample. Secondly, the findings of this study are based on self-reported aberrant driving behaviour. Previous research suggests that participants in research about behaviours that are not socially acceptable tend to minimise the extent of their negative behaviours (Swann et al., 2005). Future research should validate the self-reported responses with more objective measures such as those collected through naturalistic driving study (NDS) methodology or on-road driving tasks.
Blanchard, R., Myers, A., Porter, M., 2010. Correspondence between self-reported and objective measures of driving exposure and patterns in older drivers. Accid. Anal. Prev. 42 (2), 523–529. Browne, M.W., Cudeck, R., 1993. Alternative ways of assessing model fit. Sage Focus Ed. 154, 136. Byrne, B.M., 2013. Structural Equation Modeling With AMOS: Basic Concepts, Applications, and Programming. Routledge. Charlton, J.L., Oxley, J., Fildes, B., Oxley, P., Newstead, S., Koppel, S., O’Hare, M., 2006. Characteristics of older drivers who adopt self-regulatory driving behaviours. Transp. Res. Part F Traffic Psychol. Behav. 9 (5), 363–373. Cheung, G.W., Rensvold, R.B., 2002. Evaluating goodness-of-fit indexes for testing measurement invariance. Struct. Equ. Model 9, 233–255. De Winter, J., Dodou, D., 2010. The driver behaviour questionnaire as a predictor of accidents: a meta-analysis. J. Saf. Res. 41, 463–470. Elvik, R., Vaa, T., Hoye, A., Sorensen, M. (Eds.), 2009. The Handbook of Road Safety Measures. Emerald Group Publishing. Hu, L., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct. Equ. Model 6, 1–55. Gabaude, C., Marquié, J.C., Obriot-Claudel, F., 2010. Self-regulatory driving behaviour in the elderly: relationships with aberrant driving behaviours and perceived abilities. Le travail humain 73 (1), 31–52. Goodwin, A., Kirley, B., Sandt, L., Hall, W., Thomas, L., O’Brien, N., Summerlin, D., 2015. Countermeasures That Work: A Highway Safety Countermeasure Guide for State Highway Safety Offices. National Highway Traffic Safety Administration, Washington, DC. Gras, M.E., Sullman, M.J.M., Cunill, M., Planes, M., Aymerich, M., Font-Mayolas, S., 2006. Spanish drivers and their aberrant driving behaviours. Transp. Res. Part F Traffic Psychol. Behav. 9, 129–137. Koppel, S., Berecki-Gisolf, J., 2015. Car licensing trends of the baby boomer cohort (b. 1946-1965) compared to earlier birth cohorts: effects on the driving population in the state of Victoria. Aust. Traffic Inj. Prev. 16, 657–663. Koppel, S., Charlton, J.L., 2013. Behavioural adaptation and older drivers. In: RudinBrown, C.M., Jamson, S.L. (Eds.), Behavioural Adaptation and Road Safety: Theory, Evidence, and Action. CRC Press. Koppel, S., Charlton, J.L., Langford, J., Di Stefano, M., Macdonald, W., Mazer, B., Gelinas, I., Vrkljan, B., Eliasz, K., Myers, A., Tuokko, H., Marshall, S., 2016. The relationship between older drivers' driving behavior, functional fitness to drive, and self-reported driving-related abilities and practices. Can. J. Aging 35 (S), 15–31. Koppel, S., Charlton, J.L., Richter, N., Di Stefano, M., Macdonald, W., Darzins, P., Newstead, S.V., D’Elia, A., Mazer, B., Gelinas, I., Vrkljan, B., Eliasz, K., Myers, A., Tuokko, H., Marshall, S., 2017. Are older drivers’ on-road driving error rates related to functional performance and/or self-reported driving experiences? Accid. Anal. Prev. 103, 1–9. Koppel, S., Bohensky, M., Langford, J., Taranto, D., 2011. Older drivers, crashes and injuries. Traffic Inj. Prev. 12 (5), 459–467. Langford, J., Koppel, S., 2006. Epidemiology of older driver crashes - identifying older driver risk factors and exposure patterns. Transp. Res. Part F Traffic Psychol. Behav. 9 (5), 309–321. Li, G., Braver, E.R., Chen, L., 2003. Fragility versus excessive crash involvement as determinants of high death rates per vehicle-mile of travel among older drivers. Accid. Anal. Prev. 35, 227–235. Little, T.D., 1997. Mean and covariance structures (MACS) analyses of cross-cultural data: practical and theoretical issues. Multivariate Behav. Res. 32 (1), 53–76. Marshall, S., Man-Son-Hing, M., Bédard, M., Charlton, J.L., Gagnon, S., Gélinas, I., Koppel, S., Korner-Bitensky, N., Langford, J., Mazer, B., Myers, A., Naglie, G., Polgar, J., Porter, M., Rapoport, M., Tuokko, H., Vrkljan, B., Woolnough, A., 2013. Protocol for Candrive II/Ozcandrive, a multicentre prospective older driver cohort study. Accid. Anal. Prev. 61, 245–252. Marshall, S., 2008. The role of reduced fitness to drive due to medical impairments in explaining crashes involving older drivers. Traffic Inj. Prev. 9 (4), 291–298. Martinussen, L., Hakamies-Blomqvist, L., Møller, M., Özkan, T., Lajunen, T., 2013. Age, gender, mileage and the DBQ: the validity of the driver behavior questionnaire in different driver groups. Accid. Anal. Prev. 52, 228–236. Mattsson, M., 2012. Investigating the factorial invariance of the 28-item DBQ across genders and age groups: an exploratory structural equation modeling study. Accid. Anal. Prev. 48, 379–396. Molnar, L.J., Charlton, J.L., Eby, D.W., Langford, J., Koppel, S., Kolenic, G.E., Marshall, S., 2014. Factors affecting self-regulatory driving practices among older adults. Traffic Inj. Prev. 15 (3), 262–272. Molnar, L., Eby, D.W., 2008. The relationship between self-regulation and driving-related abilities in older drivers: an exploratory study. Traffic Inj. Prev. 9 (4), 314–319. Obriot-Claudel, F., Gabaude, C., 2004. The driver behaviour questionnaire: a French study applied to elderly drivers, Nottingham, UK. Paper presented at the International Conference on Traffic and Transport Psychology, 5th – 9th September, 2004. Parker, D., Reason, J.T., Manstead, A.S., Stradling, S.G., 1995. Driving errors, driving violations and accident involvement. Ergonomics 38, 1036–1048. Reason, J., Manstead, A., Stradling, S., Baxter, J., Campbell, K., 1990. Errors and violations on the roads: a real distinction? Ergonomics 33 (10–11), 1315–1332. Rimmö, P.E., Åberg, L., 1999. On the distinction between violations and errors: sensation seeking associations. Transp. Res. Part F Traffic Psychol. Behav. 2, 151–166. Rimmö, P.A., Hakamies-Blomqvist, L., 2002. Older drivers' aberrant driving behaviour, impaired activity, and health as reasons for self-imposed driving limitations. Transp. Res. Part F Traffic Psychol. Behav. 5 (1), 47–62. Roman, G.D., Poulter, D., Barker, E., McKenna, F.P., Rowe, R., 2015. Novice drivers' individual trajectories of driver behavior over the first three years of driving. Accid.
4.2. Summary and practical implications Overall, the results have demonstrated that the DBQ is a reliable tool to measure self-reported aberrant driving behaviour across time, although a number of modifications were required to make it suitable for older drivers. Further, the frequency of self-reported errors remained similar across the period, with only marginal decreases in violations and lapses, suggesting these also remain relatively consistent over time. Future research should validate the self-reported responses with more objective measures such as crashes and citations, as well as data collected through NDS methodology. Acknowledgements The Candrive II study was funded by a Team Grant from the Canadian Institutes of Health Research (CIHR) entitled “The CIHR Team in Driving in Older Persons (Candrive II) Research Program” (grant 90429). Ozcandrive (including the Ozcandrive eDOS subproject) is funded by an Australian Research Council Linkage Grant (LP 100100078) to the Monash University in partnership with La Trobe University, VicRoads, Victorian Government Department of Justice and Victoria Police, the Transport Accident Commission, Road Safety Trust New Zealand and Eastern Health. The authors thank Lorraine Atkinson, Ozcandrive Program Manager, for her role in managing and operationalizing the study for the Australian and New Zealand sites. The authors gratefully acknowledge the invaluable contribution of the Ozcandrive team, including: Amy Allen, Louise Beasley, Russ Boag, Matthew Catchlove, Cara Dawson, Johan Davydov, Lei Gryffydd, YikXiang Hue, Elizabeth Jacobs, Duncan Joiner, Jason Manakis, Kevin Mascarenhas, Emma Owen and Jarrod Verity. The authors also thank the Ozcandrive cohort study participants for their dedication. Without their commitment, this publication would not have been possible. References Austroads, 2010. Assessing Fitness to Drive for Commerical and Private Vehicle Drivers. Austroads, Sydney, Australia. Baldock, M.R.J., Mathias, J., McLean, J., Berndt, A., 2006. Self-regulation of driving and older Drivers' functional abilities. Clin. Gerontol. 30 (1), 53–70.
177
Accident Analysis and Prevention 113 (2018) 171–178
S. Koppel et al.
Countries. University of Michigan Transportation Research Institute, Ann Arbor, MI. Stephens, A.N., Fitzharris, M., 2016. Validation of the driver behaviour questionnaire in a representative sample of drivers in Australia. Accid. Anal. Prev. 86, 186–198. Swann, A.A., Matthews, C.E., Ebbeling, C.B., Moore, C.G., Cunningham, J.E., Fulton, J., Hebert, J.R., 2005. The effect of social desirability and social approval on self-reports of physical activity. Am. J. Epidemiol. 161, 389–398.
Anal. Prev. 82, 61–69. Singh, S., 2015. Critical Reasons for Crashes Investigated in the National motor Vehicle Crash Causation Survey (Traffic Safety Facts Crash Stats. Report No. DOT HS 812 115). National Highway Traffic Safety Administration, Washington, DC, United States. Sivak, M., Schoettle, B., 2011. Recent Changes in the Age Composition of Drivers in 15
178