Inicio Explorar Ayuda
Iniciar sesión
ardo
/
Obsidian
1
0
Fork 0
Archivos Incidencias 0 Pull Requests 0 Wiki
Árbol: a0e0ef701f
Ramas Etiquetas
Obsidian
/
Zotero
/
001_artiklid
/
chevalierexplorationolderdriversspeeding2016.md

chevalierexplorationolderdriversspeeding2016.md 72 KB
Histórico Raw

category: literaturenote citekey: chevalierexplorationolderdriversspeeding2016 title: Exploration of older drivers’ speeding behaviour authors: "Chevalier, Anna; Coxon, Kristy; Chevalier, Aran John; Wall, John; Brown, Julie; Clarke, Elizabeth; Ivers, Rebecca; Keay, Lisa" year: 2016 date: 2016-00-00 2016 doi: 10.1016/j.trf.2016.01.012 publication: "Transportation Research Part F: Traffic Psychology and Behaviour" url: "https://doi.org/10.1016/j.trf.2016.01.012" zotero_key: DNWBGZ36 zotero_storage: 4P3JKAXA collections: transpordi_planeerimine folder: 001_artiklid firstAuthor: "Chevalier, Anna"

status: converted

Transportation Research Part F

journal homepage: www.elsevier.com/locate/trf

Exploration of older drivers' speeding behaviour

Anna Chevalier a,⇑ , Kristy Coxon a,b , Aran John Chevalier c , John Wall d , Julie Brown e , Elizabeth Clarke f , Rebecca Ivers a , Lisa Keay a

  • a The George Institute for Global Health, Sydney Medical School, The University of Sydney, PO Box M201, Missenden Rd, Sydney NSW 2050, Australia
  • b School of Science and Health, University of Western Sydney, Australia
  • c Safer Roads Consulting, 53 Lachlan St, Thirroul NSW 2515, Australia
  • d The Centre for Road Safety, Transport for NSW, Road Safety Technology, Level 3, 84 Crown St, Wollongong NSW 2500, Australia
  • eNeuroscience Research Australia (NeuRA), Margarete Ainsworth Building, Barker St, Randwick NSW 2031, Australia
  • f Kolling Institute of Medical Research, Sydney Medical School, The University of Sydney, Level 10, Kolling Building 6, Royal North Shore Hospital, St Leonards NSW 2065, Australia

article info

Article history: Received 30 November 2014 Received in revised form 3 April 2015 Accepted 25 January 2016 Available online 20 February 2016

Keywords: Older drivers Speed Road safety Naturalistic In-vehicle monitoring Device

abstract

Speeding increases crash risk and resulting injury severity. Older drivers are at increased risk of injury due to frailty, at increased risk of crashing due to slower reaction times, and have less agility judging time and distance compared to younger drivers. However, there is little objective evidence about older drivers' speeding behaviour. Cross-sectional data from older drivers living in the suburban outskirts of Sydney, Australia, were used to determine the proportion of drivers involved in speed events, and examine factors that may influence this behaviour. Driving speed was estimated in approximately one-second intervals using Global Positioning System (GPS) location. Speed events were defined as driving 1 km/h or more, with 3% tolerance, above a single speed limit, averaged over 30-s of travel. Driving data from one-week were recorded for 344 participants aged 75–94 years (median 80). The majority of participants (78%) were involved in a speed event. Speed events per participant ranged from zero to 186 (median 8). Younger participants, those living in rural areas, and those driving on familiar roads closer to home were more likely to be involved in speed events adjusted for distance driven. In addition, rural residents were more likely to be involved in speed events when they had not been a driver involved in a crash during the previous year compared to those involved in a crash. Measures of visual and cognitive function did not predict involvement with speed events per distance (p > 0.2). These findings are important for policy-makers and researchers addressing older drivers' speeding to reduce the incidence of crashes and resulting fatalities and injuries. As no evidence was found for speeding being associated with functional decline, countermeasures to address speeding for other drivers seem likely to be relevant to older drivers.

2016 Elsevier Ltd. All rights reserved.

1. Introduction

Speeding is one of the most common driver behaviours contributing to fatal crashes (Geneva, 2008). Increases in vehicle speed lengthen stopping distance, and reduce drivers' response time in dangerous situations (Kloeden, McLean, Moore, &

⇑ Corresponding author at: The George Institute for Global Health, Sydney Medical School, The University of Sydney, PO Box M201, Missenden Rd, Sydney NSW 2050, Australia. Tel.: +61 2 8052 4446.

E-mail addresses: achevalier@georgeinstitute.org.au (A. Chevalier), kcoxon@georgeinstitute.org.au (K. Coxon), aran.chevalier@saferroadsconsulting.com. au (A.J. Chevalier), john.wall@transport.nsw.gov.au (J. Wall), j.brown@neura.edu.au (J. Brown), elizabeth.clarke@sydney.edu.au (E. Clarke), rivers@georgeinstitute.org.au (R. Ivers), lkeay@georgeinstitute.org.au (L. Keay).

Ponte, 1997). Even small increases in speed substantially raise the risk of crash involvement and injury severity (Aarts & van Schagen, 2006; Kloeden, Ponte, & McLean, 2001; Kloeden et al., 1997; Williams, Kyrychenko, & Retting, 2006). A study identifying speeders (at least 15 mph above the speed limit) found these drivers had a 60% higher crash history than drivers of adjacent vehicles travelling at 65 mph above the speed limit (Williams et al., 2006). Speeding is involved in approximately 40% of road fatalities in NSW (Centre for Road Safety Transport for NSW).

In Australia, older licensed drivers are increasing at a rate greater than the ageing population (BITRE, 2014). Older drivers are healthier and continuing to drive longer than previous generations (BITRE, 2014), though at increased risk of crash injury due to fragility (Kent, Henary, & Matsuoka, 2005; Meuleners, Harding, Lee, & Legge, 2006). Despite an overall reduction in fatal crashes, those involving older drivers are increasing in Australia (BITRE, 2014; Rakotonirainy, Steinhardt, Delhomme, Darvell, & Schramm, 2012). Research has found drivers aged 80 years and over were more frequently responsible for causing casualty crashes than younger drivers (Rakotonirainy et al., 2012). Older drivers were involved in fatal and reported crashes at higher or similar rates per distance travelled compared to younger drivers (Cicchino & McCartt, 2014; Keall & Frith, 2004).

Despite the seriousness of these issues, there is a paucity of research examining the relationship between speeding behaviour and ageing. Perryman and Fitten (1996) found younger drivers undertaking a road test in an instrumented vehicle were more likely than cognitively-intact older drivers to speed, and suggested older drivers may reduce speeds to compensate for other age-related declines affecting driving ability. Older drivers may be less likely to be involved in crashes involving speed (Abdel-Aty, Chen, & Schott, 1998; Cook, Knight, Olson, Nechodom, & Dean, 2000). Using on-road driving tests of older volunteers, Broberg and Dukic Willstrand (2014) found participants drove too fast for the conditions. Another study found 21% failed an on-road test, predominately due to exceeding speed limits and failure to control speed (Selander, Lee, Johansson, & Falkmer, 2011). Interviews with participants indicated they did not regularly monitor their speedometer, and were surprised they had been speeding (Broberg & Dukic Willstrand, 2014). Compared to younger drivers, older drivers have slower reaction times (Stelmach & Nahom, 1992), delaying braking and increasing stopping distances. Older road users are less agile at judging speed and distance than younger road users (Braitman, Kirley, Ferguson, & Chaudhary, 2007; Dommes, Cavallo, Vienne, & Aillerie, 2012; Lobjois & Cavallo, 2007; Yan, Radwan, & Guo, 2007), particularly evident in the over-representation of older drivers in turning across traffic crashes (Braitman et al., 2007; Cook et al., 2000; Zhang, Lindsay, Clarke, Robbins, & Mao, 2000). However, despite this body of evidence, to our knowledge, there has been no reported research that has investigated speeding in older drivers using naturalistic driving measures.

Given the relationship between speed and crash risk, Moreno and Garcia (2013) argue, in the absence of crash data, speed can be used as a surrogate safety performance measure. While other researchers have measured the proportion of time speeding, as well as mean, maximum and standard deviation of speed (af Wahlberg, 2006; Greaves & Ellison, 2011), we elected to count speeding events, based on the concept of point-to-point speeding. This method is closer to current enforcement methods, while catering for any potential inaccuracies in high readings. The current study was undertaken to determine the proportion of older drivers involved in speed events, and to assess factors that may influence this behaviour, including demographic, extent of driving, behavioural, attitude, cognitive and visual factors.

2. Methods

2.1. Study design

Participants were involved in a randomised control trial (RCT) investigating the effectiveness of an education program encouraging safe transport (Keay et al., 2013). Driving data were collected for approximately 12 months. We investigated speeding practices in a cross-sectional analysis of a week-long period of in-vehicle monitoring before the intervention. The driving monitoring system was hard-wired into the participants' vehicle ignition by a mobile technician.

2.2. Participants

The sample size was determined to be 380 for the RCT (Keay et al., 2013), but also served the purpose of investigating speed events. Assuming an average rate of involvement in speed events between 5 and 15 per 100 km travel and an over-dispersion parameter of 2, a sample size of 380 provides 80% power to detect a 20% increase in the rate of speed events for a binary predictor with a prevalence of about 50%.

A sample of 380 participants was recruited primarily through letters sent by a local motoring association to members meeting eligibility criteria, followed by promotion through local media, community groups and medical services. Volunteers undertook a brief telephone interview to confirm eligibility. Recruitment was undertaken between July 2012 and October 2013.

Eligibility criteria included being 75 years or over, holding a current driver's licence, owning a vehicle, and residing in north-west Sydney. To ensure driving data was largely from the individual, participants were required to be the main vehicle driver (undertaking greater than 80% of driving). Volunteers were excluded who received >2 errors on the Short Portable Mental Status Questionnaire cognitive assessment (Pfeiffer, 1975).

2.3. Outcome and exposure measures

The outcome was the number of speed events with distance driven applied as an offset. Participants who did not drive during the week were excluded.

2.3.1. Definition of a speed event

Driving speed data to the nearest km/h were recorded in approximately one-second intervals by the in-vehicle device. Greaves, Fifer, Ellison, and Germanos (2010) validated the device as capturing highly reliable data, with only 3% of trips determined as missing, affecting one or two of the 30 study participants (Greaves et al., 2010). Speed limit data was based on a service-provider database developed through on-road mapping of the road network. Speed events were identified from a comparison of average driving speed with mapped speed limits. To minimise identification of false positive events, events were defined as driving at an average 1 km/h or more above the speed limit plus 3% tolerance, and were 30-s in duration. This methodology was based on the notion of speeding over time, similar to point-to-point speed camera technology. The tolerance was applied to account for inaccuracies in the measuring equipment or participant's speedometer and therefore, the driver's perception of speeding. The 30-s duration was applied as average speed does not reduce sufficiently quickly to define the end of an event based on driving at or below the speed limit. An event was required to occur in a single speed zone (minimum 40 km/h). Events drew upon a minimum three satellites per record to triangulate speed. Events contained a minimum of 25 GPS records in 30 s, to account for missing records. If no event was identified, the search re-commenced at the next record.

Validity tests of speed events included assessing the definition parameters were met with respect to speed limits, and minimum numbers of satellites and GPS records. To ensure data were captured accurately, speed events/100 km and peak speeds were assessed.

2.4. Influential factors

Demographics, extent of driving, behavioural, attitude, and visual and cognitive functional measures were examined as factors that may influence speeding. Demographics consisted of age, gender and residential location (dichotomised as rural or urban, by postcode according to Council boundaries for community transport services).

The extent of driving measures included furthest radius from home, proportion of night-time driving, longest trip (distance while the ignition was turned on, and reset when the vehicle was turned off or stalled), and total trips. There were some instances where device data were unreliable, mostly due to intermittent power supply issues. These issues resulted in missing data, and we undertook validity tests to screen data for plausibility. Data that did not meet pre-specified criteria were individually examined to determine validity. Extent of driving tests included: furthest radius from home <4000 km, based on approximate road distances between Sydney and the most distant major Australian city (Perth); longest trip <385 km, based on a participant driving for a maximum 3.5 h at the highest NSW speed limit of 110 km/h; trips/week <70 trips, based on a maximum average of 10 trips/24 h; and average km/trip < 385 km/trip.

Six self-reported measures were used to consider participants' behaviour and attitudes. These included answers to ''Do you always use a seatbelt?" Answers to: ''How many accidents have you been involved in over the past year when you were the driver?"; ''How many times in the past year have you received a traffic fine where you were found to be guilty, regardless of whether or not you think you were at fault?"; and ''How fast do you generally drive compared to the general flow of traffic? Would you say: much slower, somewhat slower, about the same, somewhat faster, or much faster?", were taken from the Driving Habits Questionnaire (DHQ) (Owsley, Stalvey, Wells, & Sloane, 1999), with minor adjustments for the Australian context. Research has shown the DHQ to be reliable on test–retest (Owsley et al., 1999).

The last measure assessed participant's confidence using the published Driving Confidence Questionnaire (Baldock, Mathias, McLean, & Berndt, 2006; Owsley et al., 1999; Stalvey, Owsley, Stalvey, & Owsley, 2000). This instrument assessed confidence in nine difficult situations: when reverse parallel parking, turning across traffic, travelling on freeways, and driving in the rain, alone, on high traffic roads, during peak hour, at night, and at night in the rain. Based on further research (Sullivan, Smith, Horswill, & Lurie-Beck, 2011), three additional situations were added: driving with sun in eyes, in school zones during operation, and long distances. Participants rated their confidence in each of the 12 situations on a five-point Likert scale ranging from ''Not confident at all" to ''Completely confident". To address the different levels of situational difficulty, a Rasch model was applied to the data to produce a log-odds interval scale (Bond & Fox, 2007). The mean questionnaire item score was set at zero and each item was ranked. A positive score for an item indicated an activity required a higher level of confidence than the mean of all driving situations. Items that did not conform to the Rasch Model were removed (Pesudovs, Wright, & Gothwal, 2010). The log-odds scores were re-scaled onto a 0–100 scale. Higher scores indicated greater driving confidence.

We administered a series of tests to characterise driving-related vision and cognitive function (Keay et al., 2013). Vision was assessed by examining binocular contrast sensitivity using the Mars Perceptrix letter chart test, where higher scores indicate greater contrast sensitivity (Thayaparan, Crossland, & Rubin, 2007). Contrast sensitivity (the ability to distinguish an object from its background) has been shown to be significantly associated with crash involvement (Owsley, Stalvey, Wells, Sloane, & McGwin, 2001). Cognitive function was assessed using the Australian DriveSafe/DriveAware instrument (Kay, Bundy, & Clemson, 2009) and Trail Making Test (TMT) Parts A and B (Betz & Fisher, 2009; Bowie & Harvey, 2006).

The DriveSafe/DriveAware clinic-based driving assessment measured participant's visual attention to the driving environment and awareness of their driving ability. DriveSafe/DriveAware has been found to predict on-road driving performance, and have high sensitivity and specificity compared to other clinic-based driving tests (Kay, Bundy, Clemson, Cheal, & Glendenning, 2012; Kay et al., 2009). DriveSafe assessed attention to the driving environment through viewing and recalling components from 11 driving scenes. DriveAware assessed participant's awareness of driving ability and functional limitations through seven driving performance questions. Combined DriveSafe/DriveAware scores classified drivers as Safe, Requires further assessment, or Unsafe.

The TMT Parts A and B assessed visual scanning and executive function. Results were measured in seconds with greater completion time indicating lower function. Research indicates TMT B is predictive of crash involvement (Emerson et al., 2012; Staplin, Gish, & Sifrit, 2014; Stutts, Stewart, & Martell, 1998) and driving performance (Classen, Wang, Crizzle, Winter, & Lanford, 2013).

2.5. Analysis method

Multiple regression modelling of speed events, with distance driven during the week applied as an exposure measure, was used to examine the association between speed events and possible predictive factors. Categorical variables were collapsed where cells had low frequency (<2). Stratification was considered for significant factors. The count data model (Poisson model or negative binomial model) was established once over-dispersion was measured in the outcome variable, and significance tests at the p < 0.05 level assessed the appropriateness of the model.

Backward elimination model building utilised all available data, with significance levels set at p < 0.05. Forward testing with a significance level of p < 0.01 was undertaken to assess interaction terms for effect modification. Variables were tested for collinearity to meet criteria of variance inflation factor (VIF) < 10 (Armitage, Berry, & Matthews, 2002). Within the final model, the linearity relationship was assessed between continuous variables and the outcome. Unusual and influential data were examined by a review of Cook's Distance scores, with a criteria threshold of >1.

3. Results

3.1. Participants

Of the 380 drivers aged 75–94 years, 95% (362/380) had their vehicle instrumented, and 91% (344/380) were recorded as driving at least one-metre during the week. This study is based on these 344 participants. Characteristics of participants included in the analyses are compared with those excluded (Table 1). The only significantly different characteristic was confidence level, with excluded participants having higher mean confidence. This may be because the 18 participants who refused a device tended to own newer or more expensive vehicles. This non-instrumented group had more years of education (t378 = 3.29, p = 0.001), fewer comorbidities (t378 = 3.04, p = 0.003), and less prescription medications (t378 = 2.51, p = 0.01).

3.1.1. Model testing

A test of collinearity found the highest VIF for any single predictor or dummy variable was 1.67, and a mean 1.35 for combined predictors. Therefore, high levels of correlation were not considered an issue. A test of linear relationship between speed events and each of the continuous predictors adjusted for covariates within the non-stratified multivariate model, found both age and radius driven from home to be most suitably treated as linear. No interaction terms were found to be significant at the p < 0.01 level.

To assess the fit of each of the final models, a likelihood-ratio test of the over-dispersion of the outcome indicated a negative binomial model was suitable, and the model fitted the data well. Tests were run to compare final multivariate models with the dispersion equal to the mean or constant. The original models with the dispersion equal to the mean had the absolute log likelihood closer to zero, indicating they were a better fit to the data. Examination of unusual and influential data points in the initial multivariate model indicated there was no need to remove data.

3.2. Outcome and exposure measures: Speed events adjusted for distance driven

During the week, 5340 speed events were identified and 52,979 km driven. All event data met validity test criteria. The majority of events (81%, 4337/5340) occurred in areas with a 50–60 km/h speed limit, which are the speed limits in populated areas, where most driving occurs (Fig. 1). The majority of speed events (92%, 4900/5340) involved driving 1–9 km/h over the speed limit (Fig. 2), and a few events involved high range speeding.

The majority of participants (78%, 269/344) were involved in at least one event. Speed events per participant ranged from zero to 186 (median 8). The total distance driven ranged from <1 km to 1458 km (median 111 km). The rate of speed events/100 km ranged from 0 to 67 (median 7). The highest number of events/100 km (67), equates to speeding, according to the event definition, for at least 33.5 min of a 100 km trip.

Table 1 Older driver participants' characteristics by inclusion or exclusion from the speed events analyses.

Factors Included (N = 344) Excluded (N = 36*, p-valuey
Median or N** IQR or %** Median or N** IQR or %**
Demographics
Age (yrs) 80 77–83 78 76–81 0.12
Sex 0.13
– Male 204 59% 26 72%
– Female 140 41% 10 28%
Residential location 0.35
– Urban 296 92% 33 92%
– Rural 48 8% 3 8%
Extent of driving
Distance from home (radius km) 12 6–20 na na na
Percentage km driven at night (%) 0 0–12 na na na
Longest trip (km) 18 10–32 na na na
Total trips Mean: 22.2 SD: 11.49 na na na
Behaviour and attitude
Seat-belt usage 0.43
– Yes 338 98% 35 97%
– No 6 2% 1 3%
Crash involvement 0.87
– None 318 92% 33 92%
– 1–2 26 8% 3 8%
Citations 0.28
– None 333 97% 36 100%
– 1–2 11 3% 0 0%
Travel speed 0.63
– Somewhat slower 45 13% 6 17%
– About the same 278 81% 29 81%
– Somewhat faster 21 6% 1 3%
Driving confidence (Rasch score/10) Mean: 4.6 SD: 2.28 Mean: 5.8 SD: 2.60 0.004
Function
Contrast sensitivity score (letter) 42 41–43 42 41–43 0.24
DriveSafe/DriveAware 0.22
– Safe 76 22% 11 31%
– Require further testing/Unsafe 267 78% 24 67%
Trails A (s) 37 30–45 34 30–50 0.69
Trails B (s) 91 71–119 77 63–108 0.22

Bold values are significant at the p < 0.05 level.

3.3. Influential factor measures: Behavioural, attitude and extent of driving measures

When constructing the driver confidence scale Rasch model, responses to the items 'driving long distances' and 'reverse parallel parking' had larger than expected variance, indicating these items were measuring something different to the overall scale. When these items were removed to improve uni-dimensionality (infit and outfit mean squares 0.7–1.3), the scale conformed to the Rasch model. Final scores were based on the 10 remaining items.

Extent of driving validity tests for longest trip distance and average km/trip were plausible. One participant's figure for distance from home was inaccurate, because the latitude and longitude coordinates for the participant's residence had not been set. One participant's total trips were implausibly high, perhaps because of power supply issues. Both these single figures were removed from the analysis dataset.

The number of missing or removed figures for any single variable constituted <2.0% of data. Four variables each had a single missing or removed value, equivalent to 0.3% (1/343 or 1/344) data in each instance. In addition to the removed data, one participant did not complete the DriveSafe/DriveAware assessments, and the score was not recorded for one participant's TMT Part A.

* 18 participants vehicles were not fitted with a device, and 18 participants were not recorded as driving during the week.

** Median and interquartile range (IQR) provided for skewed continuous data. Mean and standard deviation (SD) provided and stated for normally distributed continuous data. Frequency (N) and proportion (%) provided for categorical data. Contrast sensitivity was the only skewed variable that was negatively skewed (the rest were positively skewed).

y Pearson chi-squared significance test to compare proportions. Two-sample t-test to compare means. Wilcoxon rank sum test to compare continuous nonnormally distributed data (medians). Statistical significance set at p < 0.05.

One excluded participant withdrew consent prior to completing the driving confidence, contrast sensitivity, DriveSafe/DriveAware, and Trails A and B assessments.

Fig. 1. Frequency distribution of speed events by speed limit, indicating the majority of speed events occurred in 50–60 km/h speed limits.

Fig. 2. Frequency distribution of speed events by km/h over the speed limit, indicating the majority of events involve driving 1–9 km above the speed limit.

3.4. Regression analyses

3.4.1. Model testing

Speed events ranged from zero to 186 per participant, with a mean of 16 and a standard deviation (SD) of 22.8. The variance being much larger than the mean indicates over-dispersion. This was confirmed by the likelihood ratio test of alpha, being significant at the p < 0.05 level. Therefore, it was appropriate to use a negative binomial rather than Poisson distribution. Multivariate regression modelling found the incidence of participant speed events adjusted for distance driven increased by approximately 57% for rural compared to urban residents, decreased by approximately 3% with every year increase in age, and decreased by 22% for every 100 km increase in radius driven from home, when other significant predictors were constant (Table 2). Univariate analysis found participants were significantly more likely to be involved in events adjusted for distance when they reported driving about the same or somewhat faster than traffic flow, however, this was not independently predictive when other factors were considered. Measures of behaviour, attitude and function were not predictive.

3.4.2. Stratified analysis by residential location

While a small proportion of participants lived in rural areas (14% or 48/344), they were over-represented in speed events 26% (1383/5340), without considering distance driven. Speed events per 100 km for rural residents ranged from 0 to 56 (median 13), while for urban residents the range was 0–67 (median 6) (Fig. 3). Rural residents have a higher rate of speed events per distance driven compared to urban residents (Table 2). Therefore, we conducted a stratified analysis by residential location.

Table 2 Results of regression of factors predicting older driver participants' involvement with speed events adjusted for distance driven.

| Factors | Unadjusted | | | Adjusted (N = 343) | | | | |------------------------------------------------------|------------|------------|------|--------------------|-----------|--|--| | | IRR** | p-value*** | IRR | 95% CI | p-value** | | | | Demographics | | | | | | | | | Age (yrs) | 0.97 | 0.04 | 0.97 | 0.94–1.00 | 0.04 | | | | Gender: Female | 0.93 | 0.61 | | | | | | | Residential location: Rural | 1.64 | 0.007 | 1.57 | 1.10–2.24 | 0.01 | | | | Extent of driving | | | | | | | | | Distance from home/100 km (radius) | 0.78 | 0.005 | 0.78 | 0.66–0.93 | 0.005 | | | | Proportion of night-time driving (%) | 1.01 | 0.19 | | | | | | | Longest trip/100 km | 0.82 | 0.25 | | | | | | | Total trips | 1.00 | 1.00 | | | | | | | Behaviour and attitude | | | | | | | | | Seat-belt usage: No | 1.24 | 0.66 | | | | | | | Crash involvement: 1–2 | 0.83 | 0.46 | | | | | | | Citations: 1–2 | 0.96 | 0.91 | | | | | | | Travel speed | | 0.047 | | | | | | | – Somewhat slower* | 1.00* | | | | | | | | – About the same | 1.67 | | | | | | | | – Somewhat faster | 1.63 | | | | | | | | Driving confidence (Rasch score/10) | 1.05 | 0.08 | | | | | | | Function | | | | | | | | | Contrast sensitivity score (letter) | 1.01 | 0.59 | | | | | | | DriveSafe/DriveAware: Require further testing/Unsafe | 0.88 | 0.44 | | | | | | | TMT Part A (/10 s) | 0.98 | 0.66 | | | | | | | TMT Part B (/10 s) | 0.99 | 0.69 | | | | | |

Bold values are significant at the p < 0.05 level.

For rural residents, the likelihood of involvement in speed events decreased for drivers involved in 1–2 crashes in the past year, and this was the only significant predictor (Table 3).

For urban residents, the incidents of speed events adjusted for distance driven decreased for drivers driving further from home and with every year increase in age, when the other significant predictor was held constant (Table 3). Age only became significant when included in the multivariate model (Table 3).

4. Discussion

The majority of older drivers (78%) were involved in speeding events. However, as hypothesised, speeding was less common with advancing age. The finding older drivers were less likely to speed, complements other literature indicating drivers

Fig. 3. Frequency distribution of speed events/100 km for urban residents (median 6 events/100 km) and rural residents (median 13/100 km). Participants residing in urban areas account for 86% of participants.

* Reference category.

** Incident rate ratio.

*** Statistical significance set at p < 0.05.

Table 3 Results of regression of factors predicting involvement with older driver participants' speed events adjusted for distance driven stratified by residential location.

Residential location Urban Rural
Unadjusted Adjusted (N = 295) Unadjusted
Factors IRR p-value IRR 95% CI p-value IRR p-value
Demographics
Age (yrs) 0.97 0.053 0.97 0.94–0.999 0.04 0.99 0.68
Gender: Female 0.89 0.42 0.98 0.96
Extent of driving
Distance from home/100 km (radius) 0.80 0.02 0.79 0.66–0.94 0.04 0.23 0.16
Proportion of night-time driving (%) 1.006 0.18 1.01 0.47
Longest trip/100 km 0.84 0.30 0.81 0.77
Total trips 0.9999 0.99 0.996 0.78
Behaviour and attitude
Seat-belt usage: No 1.36 0.54 Omitted**
Crash involvement: 1–2 0.99 0.96 0.08 (95% CI: 0.02–0.40) 0.002
Citations: 1–2 0.63 0.28 1.78 0.45
Travel speed 0.10 0.64
– Somewhat slower* 1.00* 1.00*
– About the same 1.60 1.77
– Somewhat faster 1.54 1.82
Driving confidence (Rasch score/10) 1.05 0.47 1.05 0.13
Function
Contrast sensitivity score (letter) 1.005 0.82 1.06 0.44
DriveSafe/DriveAware: Require further testing/Unsafe 0.83 0.29 1.23 0.75
TMT Part A (/10 s) 0.9999 1.00 0.87 0.30
TMT Part B (/10 s) 1.001 0.95 0.97 0.37

Bold values are significant at the p < 0.05 level.

exhibit more caution, self-restriction or avoidance with increasing age (Blanchard & Myers, 2010; Charlton et al., 2006; Molnar & Eby, 2008; Molnar et al., 2013). However, we did not find evidence of lower cognitive or visual function or confidence explaining a change in speeding events.

As speed events were defined as having 30-s duration, the extent of speeding is under-represented as shorter duration speeding (less than 30-s) was not detected. While Fig. 4 shows several instances when a participant exceeded the speed limit, only one met the criteria for a speed event. Previous studies using different naturalistic driving measures have also found speeding was common, although to our knowledge no previous work has been reported about older drivers (Greaves & Ellison, 2011; Stigson, Hagberg, Kullgren, & Krafft, 2014). Greaves' and Ellison's (2011) study was based on Doppler speeds, and measured distance speeding within the speed zone, when travelling at or greater than 75% of the speed limit. Every driver exceeded the speed limit during the five-week study, and overall 19% of distance driven was above the speed limit (Greaves & Ellison, 2011).

Participants driving closer to home were significantly more likely to be involved in speed events per distance driven, particularly those drivers living in urban areas. This may indicate a propensity to speed when driving in familiar areas, where speed enforcement measures may be less prevalent. Driving further from home may reflect recreational trips with reduced pressure to reach a destination (Broberg & Dukic Willstrand, 2014).

Crash analyses show high levels of crashes occur close to home, where the majority of driving occurs. There is evidence those travelling fewer kilometres have reduced fitness-to-drive (Langford et al., 2013), and the same pattern was observed in this study cohort (Coxon et al., 2015). Research indicates, independent of age, drivers travelling further annually have lower crash rates per kilometre, compared to those driving fewer kilometres (low-mileage bias) (Langford, Koppel, McCarthy, & Srinivasan, 2008; Langford, Methorst, & Hakamies-Blomqvist, 2006; Langford et al., 2013). While this study applied kilometres driven as an exposure measure, it was based on weekly mileage, which may not represent all typical driving patterns. Although, a longer data capture period may have increased the significance of this predictor.

We found rural residents were more likely to be involved in speed events, and this complements research showing speedrelated crashes are more common on rural roads. Research has found speeding was responsible for a larger proportion of rural crashes (Kloeden et al., 2001) than urban crashes (Kloeden et al., 1997). While these two studies were not confined to older drivers, it confirms the importance of speeding on rural roads. Fatal crashes were more prevalent in rural areas (Clark, 2001; Zwerling et al., 2005). Crashes involving older rural drivers were more than twice as likely to result in a serious injury or fatality as those involving urban drivers (Thompson, Baldock, Mathias, & Wundersitz, 2013b). Rural road environments may allow more opportunity for involvement in speed events. With lower traffic density, driving speeds may not be constrained by traffic flow. Fewer intersections and lower pedestrian activity may mean drivers are less aware or cautious.

* Reference category.

** All rural residents reported always wearing a seatbelt.

Fig. 4. Two participant trips depicting driving above the speed limit in orange/mid-grey and a speed event in red/dark-grey which met the criteria of being on average at least 1 km/h above the speed limit for 30-s. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Research indicates rural drivers' perceive a decrease in enforcement of speeds (Rakauskas, Ward, & Gerberich, 2009). Similarly, a study found older rural residents were less likely to stop at stop-signs, and suggested this may be related to the road environment, including reduced traffic and improved visibility, increasing driver's willingness to take the calculated-risk of not stopping (Keay et al., 2009).

The stratified analysis served to isolate the influence of rural roads; however, we did not find any further explanatory factors for speeding in the urban sample. When examined separately, rural drivers were significantly less likely to be involved in speed events per distance driven when they had been involved as a driver in a crash during the previous year. A driver's crash involvement may cause higher propensity to adhere to the speed limit for a period. Previous research found male drivers aged 65 years and older, involved in three crashes in five years, reported safer driving behaviour including reducing speed compared to a control group (Daigneault, Joly, & Frigon, 2002). The extent of this speed reduction may be influenced by the number of crashes experienced, crash type and severity. This was not explored in our study where 8% of all participants were involved in crashes in the last year. Self-reported crashes are likely to include low-level crashes that would not be reflected in police-reported crashes, and may be under-reported by participants due to recall and socialdesirability bias issues.

As the rural cohort sample was small, a sample with a larger rural cohort or from more remote regions would be useful to revisit this finding. Significant predictors of involvement with speed events may be different for older drivers living in rural areas, due to higher driving reliance caused by lower availability of inexpensive alternative transport options (Siskind, Steinhardt, Sheehan, O'Connor, & Hanks, 2011; Thompson, Baldock, Mathias, & Wundersitz, 2013a).

Though we measured some behavioural and attitude factors, a validated measure of personality or risk-taking was not available. This could be addressed in future studies. Beanland, Sellbom, and Johnson (2014) have provided a review of studies examining the relationship between personality traits and aberrant driving behaviours. Measures of personality traits and driving attitudes in older drivers have been associated with self-reported risky driving, including violations, errors and lapses (Beanland et al., 2014; Lucidi, Mallia, Lazuras, & Violani, 2014), which have been associated with older drivers' crash involvement (Lucidi et al., 2014).

Unlike other studies that found an association between visual and cognitive decline and a decrease in safe driving behaviours (including lane-change errors, and failure to stop at red lights) (Munro et al., 2010; West et al., 2010), in this study function was not a predictor of involvement in speed events adjusted for distance driven. This may be related to research volunteers having higher levels of health and function (Molnar & Eby, 2008; Molnar et al., 2013; Thompson et al., 2013a), though we did observe a range in function. Speeding may be related to a calculated risk many drivers make, or alternatively inattention to driving speed.

According to our inclusion criteria, due to the length of the study, participants were the driver of the instrumented vehicle greater than 80% of the time. Participants were encouraged to inform the study if someone else was driving the vehicle, and these weeks of data were manually removed. When asked about use by other drivers, participant's reported this to be infrequent. Therefore, a minority of travel captured may reflect other drivers. Future research would benefit from utilising driver recognition technology.

As no evidence was found for speeding per distance being associated with cognitive decline, the types of countermeasures to address speeding for other drivers seem likely to be relevant to older drivers. Potential countermeasures include targeted enforcement and education campaigns aimed at reducing speeding. Education campaigns could include the promotion of speed alerting tools based on Intelligent Speed Adaption (ISA) technology. ISA technology has been found to lower offender's speed (van der Pas, Kessels, Veroude, & van Wee, 2014). Cost-benefit modelling of ISA impact, which at certain levels of market saturation lowers the general speed, indicates substantial reductions in fatal and serious crashes overtime (Lai, Carsten, & Tate, 2012). In semi-rural areas, countermeasures could include speed limit reduction (Thompson et al., 2013b) and increased alternative transport options (Siskind et al., 2011). Research comparing speeding behaviour across a broader range of age groups would provide useful information to policy makers considering resource allocation.

5. Conclusion

This study defined speed events based on objective GPS data and found a large proportion of drivers aged 75–94 years (mean 80 years) were involved in speed events. To our knowledge, this is the first time objective speeding behaviour of community-based older drivers has been examined. Participants were significantly more likely to be involved in speed events adjusted for distance driven when younger (closer to 75 years); their observed driving was limited to local, familiar roads; and rural residents. Predictive factors were found to be different for drivers living in rural compared to urban areas. We found no evidence that speeding adjusted for distance driven was explained by deficits in visual or cognitive function. These findings are important for road safety policy makers and researchers wishing to address older drivers speeding to reduce the incidence of crashes involving older drivers and resulting casualties.

Other information

Ethics: The University of Sydney Human Ethics Committee approved the study, and written consent was provided by all participants. This research was undertaken in accordance with the STROBE Statement for observational studies (von Elm et al., 2014).

Hardware: The in-vehicle device consisted of a C4D Data Recorder, external GPS receiver, tri-axial accelerometer, tachograph, real-time clock, 128 MB memory and internal battery (1300 mA). The devices had low-power usage, similar to devices and accessories supplied with a 12v plug for mobile use. Data were transmitted over the mobile telecommunications network. To address confidentiality and validity, there were a number of protocols in place to protect the transfer, storage, accuracy and use of data.

Software: Device data was processed using MatLab (MathsWorks, R2012a). Data was stored within REDCap (Research Electronic Data Capture) (Harris et al., 2009). WinSteps was used to produce the driver confidence Rasch scores (Linacre, 2013). Analyses were undertaken using StataC 13 (StataCorp, 2013).

Acknowledgements

The authors thank the study participants, without whom this study would not have been possible, and NRMA Motoring for their assistance with participant recruitment. Claire Allan, Laura Peattie, Freya Saich and RachelleMason were also involved in data collection. This research was funded by an Australian Research Council Discovery Project, University of Sydney Equipment Grant (DP110101740), IRT Foundation and the Centre for Road Safety, Transport for NSW. AC is the recipient of an NRMA-ACT Trust scholarship. However, the publication content has not been endorsed by, is not guaranteed by, and does not necessarily reflect the views of the NRMA-ACT Trust. KC holds an Australian Postgraduate Award. EC, RI and JB are supported by NHMRC Research Fellowships.

References

Aarts, L., & van Schagen, I. (2006). Driving speed and the risk of road crashes: A review. Accident Analysis & Prevention, 38(2), 215–224. http://dx.doi.org/ 10.1016/j.aap.2005.07.004.

Abdel-Aty, M. A., Chen, C. L., & Schott, J. R. (1998). An assessment of the effect of driver age on traffic accident involvement using log-linear models. Accident Analysis & Prevention, 30(6), 851–861.

af Wahlberg, A. E. (2006). Speed choice versus celeration behavior as traffic accident predictor. Journal of Safety Research, 37(1), 43–51. http://dx.doi.org/

Armitage, P., Berry, G., & Matthews, J. N. S. (2002). Statistical methods in medical research (4th ed.). Oxford: Blackwell Scientific Publications, p. 359.

Baldock, M. R., Mathias, J. L., McLean, A. J., & Berndt, A. (2006). Self-regulation of driving and its relationship to driving ability among older adults. Accident Analysis & Prevention, 38(5), 1038–1045. http://dx.doi.org/10.1016/j.aap.2006.04.016.

Beanland, V., Sellbom, M., & Johnson, A. K. (2014). Personality domains and traits that predict self-reported aberrant driving behaviours in a southeastern US university sample. Accident Analysis & Prevention, 72C, 184–192. http://dx.doi.org/10.1016/j.aap.2014.06.023.

Betz, M. E., & Fisher, J. (2009). The trail-making Test B and driver screening in the emergency department. Traffic Injury Prevention, 10(5), 415–420. http://dx. doi.org/10.1080/15389580903132819.

BITRE (2014). Road safety of older Australians: Recent statistics.

Blanchard, R. A., & Myers, A. M. (2010). Examination of driving comfort and self-regulatory practices in older adults using in-vehicle devices to assess natural driving patterns. Accident Analysis & Prevention, 42(4), 1213–1219. http://dx.doi.org/10.1016/j.aap.2010.01.013.

Bond, T., & Fox, C. M. (2007). Applying the Rasch model: Fundamental measurement in the human sciences.

Bowie, C. R., & Harvey, P. D. (2006). Administration and interpretation of the trail making test. Nature Protocols, 1(5), 2277–2281. http://dx.doi.org/10.1038/ nprot.2006.390.

Braitman, K. A., Kirley, B. B., Ferguson, S., & Chaudhary, N. K. (2007). Factors leading to older drivers' intersection crashes. Traffic Injury Prevention, 8(3), 267–274. http://dx.doi.org/10.1080/15389580701272346.

Broberg, T., & Dukic Willstrand, T. (2014). Safe mobility for elderly drivers–considerations based on expert and self-assessment. Accident Analysis & Prevention, 66, 104–113. http://dx.doi.org/10.1016/j.aap.2014.01.014.

Centre for Road Safety Transport for NSW. Crash behavioural factors. <http://roadsafety.transport.nsw.gov.au/statistics/crashbehaviouralfactors.html> Retrieved October, 2014.

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. Transportation Research Part F: Traffic Psychology and Behaviour, 9(5), 363–373. http://dx.doi.org/10.1016/j.trf.2006.06.006.

Cicchino, J. B., & McCartt, A. T. (2014). Trends in older driver crash involvement rates and survivability in the United States: An update. Accident Analysis & Prevention, 72, 44–54. http://dx.doi.org/10.1016/j.aap.2014.06.011.

Clark, D. E. (2001). Motor vehicle crash fatalities in the elderly: Rural versus Urban. Journal of Trauma-Injury Infection & Critical Care, 51(5), 896–900.

Classen, S., Wang, Y., Crizzle, A. M., Winter, S. M., & Lanford, D. N. (2013). Predicting older driver on-road performance by means of the useful field of view and trail making test part B. American Journal of Occupational Therapy, 67(5), 574–582. http://dx.doi.org/10.5014/ajot.2013.008136.

Cook, L. J., Knight, S., Olson, L. M., Nechodom, P. J., & Dean, J. M. (2000). Motor vehicle crash characteristics and medical outcomes among older drivers in Utah, 1992–1995. Annals of Emergency Medicine, 35(6), 585–591.

Coxon, K., Chevalier, A., Lo, S., Ivers, R., Brown, J., & Keay, L. (2015). Behind the wheel: Predictors of driving exposure in older drivers. Journal of the American Geriatrics Society, 63(6), 1137–1145. http://dx.doi.org/10.1111/jgs.13440.

Daigneault, G., Joly, P., & Frigon, J. Y. (2002). Executive functions in the evaluation of accident risk of older drivers. Journal of Clinical and Experimental Neuropsychology, 24(2), 221–238. http://dx.doi.org/10.1076/jcen.24.2.221.993.

Dommes, A., Cavallo, V., Vienne, F., & Aillerie, I. (2012). Age-related differences in street-crossing safety before and after training of older pedestrians. Accident Analysis & Prevention, 44(1), 42–47. http://dx.doi.org/10.1016/j.aap.2010.12.012.

Emerson, J. L., Johnson, A. M., Dawson, J. D., Uc, E. Y., Anderson, S. W., & Rizzo, M. (2012). Predictors of driving outcomes in advancing age. Psychology and Aging, 27(3), 550–559. http://dx.doi.org/10.1037/a0026359.

Geneva, G. R. S. P. (2008). Speed management: A road safety manual for decision-makers and practitioners.

Greaves, S., & Ellison, A. (2011). Personality, risk aversion and speeding: An empirical investigation. Accident Analysis & Prevention, 43(5), 1828–1836. http:// dx.doi.org/10.1016/j.aap.2011.04.018.

Greaves, S., Fifer, S., Ellison, R., & Germanos, G. (2010). Development of a global positioning system web-based prompted recall solution for longitudinal travel surveys. Transportation Research Record: Journal of the Transportation Research Board, 2183, 69–77. http://dx.doi.org/10.3141/2183-08.

Harris, P. A., Taylor, R., Thielke, R., Payne, J., Gonzalez, N., & Conde, J. G. (2009). Research electronic data capture (REDCap) – A metadata-driven methodology and workflow process for providing translational research informatics support. Journal of Biomedical Informatics, 42(2), 377–381.

Kay, L., Bundy, A., & Clemson, L. (2009). Predicting fitness to drive in people with cognitive impairments by using DriveSafe and DriveAware. Archives of Physical Medicine and Rehabilitation, 90(9), 1514–1522. http://dx.doi.org/10.1016/j.apmr.2009.03.011.

Kay, L., Bundy, A., Clemson, L., Cheal, B., & Glendenning, T. (2012). Contribution of off-road tests to predicting on-road performance: A critical review of tests. Australian Occupational Therapy Journal, 59(1), 89–97. http://dx.doi.org/10.1111/j.1440-1630.2011.00989.x.

Keall, M., & Frith, W. (2004). Adjusting for car occupant injury liability in relation to age, speed limit, and gender-specific driver crash involvement risk. Traffic Injury Prevention, 5(4), 336–342. http://dx.doi.org/10.1080/15389580490509473.

Keay, L., Coxon, K., Brown, J., Clarke, E., Boufous, S., Bundy, A., ... Ivers, R. (2013). A randomized trial to evaluate the effectiveness of an individual, educationbased safe transport program for drivers aged 75 years and older. BMC Public Health, 13, 106. http://dx.doi.org/10.1186/1471-2458-13-106.

Keay, L., Jasti, S., Munoz, B., Turano, K. A., Munro, C. A., Duncan, D. D., ... West, S. K. (2009). Urban and rural differences in older drivers' failure to stop at stop signs. Accident Analysis & Prevention, 41(5), 995–1000. http://dx.doi.org/10.1016/j.aap.2009.06.004.

Kent, R., Henary, B., & Matsuoka, F. (2005). On the fatal crash experience of older drivers. In Annual proceedings of the association for the advancement of automotive medicine (Vol. 49, pp. 371–391).

Kloeden, C. N., McLean, A. J., Moore, V. M., & Ponte, G. (1997). Travelling speed and the risk of crash involvement. 1 and 2 (CR204).

Kloeden, C. N., Ponte, G., & McLean, A. J. (2001). Travelling speed and the risk of crash involvement on rural roads. (CR204).

Lai, F., Carsten, O., & Tate, F. (2012). How much benefit does Intelligent Speed Adaptation deliver: An analysis of its potential contribution to safety and environment. Accident Analysis & Prevention, 48, 63–72. http://dx.doi.org/10.1016/j.aap.2011.04.011.

Langford, J., Charlton, J. L., Koppel, S., Myers, A., Tuokko, H., Marshall, S., ... Macdonald, W. (2013). Findings from the Candrive/Ozcandrive study: Low mileage older drivers, crash risk and reduced fitness to drive. Accident Analysis & Prevention, 61, 304–310. http://dx.doi.org/10.1016/j.aap.2013.02.006.

Langford, J., Koppel, S., McCarthy, D., & Srinivasan, S. (2008). In defence of the 'low-mileage bias'. Accident Analysis & Prevention, 40(6), 1996–1999. http://dx. doi.org/10.1016/j.aap.2008.08.027.

Langford, J., Methorst, R., & Hakamies-Blomqvist, L. (2006). Older drivers do not have a high crash risk – A replication of low mileage bias. Accident Analysis & Prevention, 38(3), 574–578. http://dx.doi.org/10.1016/j.aap.2005.12.002.

Linacre, J. M. (2013). Winsteps Rasch measurement (Version 3.80.1) (Computer software).

Lobjois, R., & Cavallo, V. (2007). Age-related differences in street-crossing decisions: The effects of vehicle speed and time constraints on gap selection in an estimation task. Accident Analysis & Prevention, 39(5), 934–943. http://dx.doi.org/10.1016/j.aap.2006.12.013.

Lucidi, F., Mallia, L., Lazuras, L., & Violani, C. (2014). Personality and attitudes as predictors of risky driving among older drivers. Accident Analysis & Prevention, 72C, 318–324. http://dx.doi.org/10.1016/j.aap.2014.07.022.

MathsWorks (R2012a). MatLab (matrix laboratory) Software.

Meuleners, L. B., Harding, A., Lee, A. H., & Legge, M. (2006). Fragility and crash over-representation among older drivers in Western Australia. Accident Analysis & Prevention, 38(5), 1006–1010. http://dx.doi.org/10.1016/j.aap.2006.04.005.

Molnar, L. J., & Eby, D. W. (2008). The relationship between self-regulation and driving-related abilities in older drivers: An exploratory study. Traffic Injury Prevention, 9(4), 314–319. http://dx.doi.org/10.1080/15389580801895319.

Molnar, L. J., Eby, D. W., Charlton, J. L., Langford, J., Koppel, S., Marshall, S., & Man-Son-Hing, M. (2013). Driving avoidance by older adults: Is it always selfregulation? Accident Analysis & Prevention, 57, 96–104. http://dx.doi.org/10.1016/j.aap.2013.04.010.

Moreno, A. T., & Garcia, A. (2013). Use of speed profile as surrogate measure: Effect of traffic calming devices on crosstown road safety performance. Accident Analysis & Prevention, 61, 23–32. http://dx.doi.org/10.1016/j.aap.2012.10.013.

Munro, C. A., Jefferys, J., Gower, E. W., Muñoz, B. E., Lyketsos, C. G., Keay, L., ... West, S. K. (2010). Predictors of lane-change errors in older drivers. Journal of the American Geriatrics Society, 58(3), 457–464. http://dx.doi.org/10.1111/j.1532-5415.2010.02729.x.

Owsley, C., Stalvey, B., Wells, J., & Sloane, M. E. (1999). Older drivers and cataract: Driving habits and crash risk. The Journals of Gerontology Series A: Biological

Sciences and Medical Sciences, 54(4), M203–M211. Owsley, C., Stalvey, B. T., Wells, J., Sloane, M. E., & McGwin, G. Jr., (2001). Visual risk factors for crash involvement in older drivers with cataract. Archives of

Ophthalmology, 119(6), 881–887. Perryman, K. M., & Fitten, L. J. (1996). Effects of normal aging on the performance of motor-vehicle operational skills. Journal of Geriatric Psychiatry and

Neurology, 9(3), 136–141.

Pesudovs, K., Wright, T. A., & Gothwal, V. K. (2010). Visual disability assessment: Valid measurement of activity limitation and mobility in cataract patients. British Journal of Ophthalmology, 94(6), 777–781. http://dx.doi.org/10.1136/bjo.2009.169490.

Pfeiffer, E. (1975). A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. Journal of the American Geriatrics Society, 23(10), 433–441.

Rakauskas, M. E., Ward, N. J., & Gerberich, S. G. (2009). Identification of differences between rural and urban safety cultures. Accident Analysis & Prevention, 41(5), 931–937. http://dx.doi.org/10.1016/j.aap.2009.05.008.

Rakotonirainy, A., Steinhardt, D., Delhomme, P., Darvell, M., & Schramm, A. (2012). Older drivers' crashes in Queensland, Australia. Accident Analysis & Prevention, 48, 423–429. http://dx.doi.org/10.1016/j.aap.2012.02.016.