Transportation Research Part F 21 (2013) 1–13

Contents lists available at ScienceDirect

Transportation Research Part F

journal homepage: www.elsevier .com/locate / t r f

The impact of motor vehicle injury on distress: Moderatorsand trajectories over time

1369-8478/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.trf.2013.08.004

⇑ Corresponding author at: Student Development Centre, University of Western Ontario, Room 4100, Western Student Services Building, 1151 RStreet, London, ON, N6A 3K7, Canada. Tel.: +1 519 661 2111x85938; fax: +1 519 850 2374.

E-mail addresses: nwiesent@uwo.ca (N. Wiesenthal), evingili@uwo.ca (E. Vingilis).1 Tel.: +1 519 858 5063x2; fax: +1 519 661 3878.

Naomi Wiesenthal a,b,⇑, Evelyn Vingilis c,1

a Student Development Centre, University of Western Ontario, Room 4100, Western Student Services Building, 1151 Richmond Street, London, ON, N6A 3K7, Canadab Consultation-Liaison Service, Mental Health Care Program, London Health Sciences Centre, A2-650, Victoria Hospital, 800 Commissioners Road East, POBox 5010, London, Ontario N6A 5W9, Canadac Population & Community Health Unit, Family Medicine, University of Western Ontario, Clinical Skills Building, Room 2711, London, Ontario N6A 5C1, Canada

a r t i c l e i n f o

Article history:Received 17 December 2012Received in revised form 11 April 2013Accepted 7 August 2013

Keywords:Motor vehicle injuryDistressAlcohol

a b s t r a c t

Research reveals that motor vehicle injuries (MVIs) can result in severe and debilitatingpsychological distress. Yet, not every person who has sustained a MVI suffers psychologi-cally. It appears that risk of distress varies by demographic and psychosocial characteris-tics. The present study aimed to explore the trajectories of post-MVI distress and theeffect of pre-MVI psychological functioning on post-MVI distress. Hierarchical linear mod-eling was used to explore the longitudinal dataset from the Canadian National PopulationHealth Survey. Participants were assessed up to nine years post-MVI. Post-MVI distressincreased over time. Men experienced greater overall distress than women and a greaterincrease in distress over time. Pre-MVI distress predicted post-MVI distress. This relation-ship was strongest for those with greater pre-MVI alcohol consumption. At low levels ofpre-MVI distress, greater pre-MVI alcohol consumption was related to lower post-MVI dis-tress, but at high levels of pre-MVI distress, greater pre-MVI alcohol consumption pre-dicted increased post-MVI distress. Those with partners experienced less distress thanthe unpartnered. This study supports the general findings of other post-MVI andpost-trauma studies, although the current study’s main and interaction effects reveal morecomplex and nuanced relationships among variables in their prediction of post-MVIpsychological distress.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

World-wide, 20–50 million people are injured in motor vehicle collisions each year (WHO, 2009). The WHO notes thatonly during this last decade has this issue ‘‘gained the prominence it deserves among the world’s most pressing internationalhealth and development concerns’’ (p. 2). Despite the high world-wide prevalence of motor vehicle injuries (MVIs) andrecent research demonstrating that psychological distress is a key predictor of post-injury quality of life (Brasel, deRoon-Cas-sini, & Bradley, 2010), research remains limited on the psychological sequelae of MVIs over time and on risk factors for post-MVI distress.

Previous clinical research reveals that MVIs can result in severe and debilitating psychological distress. The mostcommonly reported post-MVI distress reactions are mood disorders (e.g., depression) and anxiety disorders, including

ichmond

2 N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13

post-traumatic stress disorder (PTSD), acute stress disorder (ASD), and travel phobia (Blanchard et al., 1996; Ehlers,Hofmann, Herda, & Roth, 1994; Ehring, Ehlers, & Glucksman, 2008; Goldberg & Gara, 1990; Mayou, Bryant, & Duthie,1993; Taylor, Deane, & Podd, 2002; Taylor & Koch, 1995; Vingilis, Larkin, Stoduto, Parkinson-Heyes, & McLellan, 1996).Primary research indicates that MVI survivors presenting for medical care are diagnosed with PTSD at rates approaching46% (Blanchard, Hickling, Taylor, Loos, & Gerardi, 1994; Bryant & Harvey, 1995; Kupchik et al., 2007; Seethalakshmi, Dhavale,Gawande, & Dewan, 2006). Another 20–26% experience symptoms without meeting criteria for the full syndrome (Blanchardet al., 1994; Kupchik et al., 2007). A related disorder, ASD, is frequently diagnosed in individuals displaying PTSD-like symp-toms prior to meeting the one-month symptom duration criterion necessary to diagnose PTSD. Within this one-month time-frame post-MVI, 14–16% of hospital patients could be diagnosed with ASD, and an additional 14% experienced subsyndromalsymptoms (Harvey & Bryant, 1999). Depression is also common, with rates ranging between 21% and 67% (Beck & Coffey,2007; Blanchard et al., 2004; Blaszczynski et al., 1998; Vingilis et al., 1996). Estimates of anxiety range between 4% and87% (Blanchard et al., 1994; Blanchard et al., 2004; Blaszczynski et al., 1998; Kupchik et al., 2007; Vingilis et al., 1996),and up to 100% show some reluctance in driving (Blanchard et al., 1994). Mayou, Bryant, and Ehlers (2001) found that atone year post-MVI, 16% of individuals experienced phobic travel anxiety. Other research suggests phobias occur in anywherefrom 2% to 47% of those injured (Blaszczynski et al., 1998). Psychosomatic complaints, adjustment concerns, and irritabilityare also common sequelae (Blaszczynski et al., 1998; Vingilis et al., 1996).

Twenty percent of those injured still met criteria for diagnoses other than PTSD at six months post-MVI (Frommbergeret al., 1998). Of those injured who presented to an Emergency Department, 33% maintained psychological complaints atone year post-MVI: 17% with PTSD, 16% with phobic travel anxiety, 19% with general anxiety, and 6% with depression(Mayou et al., 2001). Moreover, different groups show different trajectories of distress over time (Blanchard et al., 1996).One five-year MVI follow-up study found that 10% of participants had PTSD at five years, but there was notable variabilityin outcomes. Indeed, four of five participants diagnosed with PTSD at year one had improved by year five, whereas eight ofnine diagnosed with PTSD at year five had not previously met criteria for the disorder (Mayou, Tyndel, & Bryant, 1997).Despite these intriguing findings, conclusions remain limited because this study had a small sample size and used non-para-metric tests and logistic regression to examine group differences, and thus, the authors were not able to identify which vari-ables were systematically predictive of different trajectories. In a different study, deRoon-Cassini, Mancini, Rusch, andBonanno (2010) studied a sample experiencing a range of different trauma presenting to a level 1 trauma center and exam-ined their trajectories of PTSD and depression at 1, 3 and 6 months follow-up, using latent growth curve modeling. Theirfindings confirmed different trajectories of distress over time. Similarly, other research with less specific patient populations(victims of any loss or trauma) has identified different trajectories of psychological sequelae (chronic, improved and delayedonset distress) (Bonanno, 2004; Bonanno, 2005), although very few studies have specifically examined post-MVI trajectoriesof psychological sequelae over the long term, or pre-MVI predictors of the post-MVI distress trajectories.

Despite the pervasiveness of psychological sequelae, not every person who has sustained a MVI suffers psychologically.Instead, it appears that risk of distress varies by pre-injury psychological and socio-demographic characteristics. Measures ofpre-MVI psychological functioning and alcohol use and abuse consistently have been associated with post-MVI distress(Blanchard et al., 1994; Blanchard et al., 1996; Blanchard et al., 2004; Brewin, Andrews, & Valentine, 2000; Bromet, Sonnega,& Kessler, 1998; Bryant & Harvey, 1995; Ehlers, Mayou, & Bryant, 1998; Fullerton et al., 2001; Mayou et al., 2001; Ursanoet al., 1999; Vingilis et al., 1996; Zatzick et al., 2002). Indeed, in their meta-analysis of risk for PTSD among adults facingany trauma, Brewin et al. (2000) stated that psychiatric problems/history had the most uniformity of effect, although theycautioned that effects varied by type of study and type of trauma.

Alcohol use and abuse also have been found to predict post-trauma distress, although few studies actually have examinedpre-MVI alcohol use (Blanchard et al., 1994; Vingilis et al., 1996; Zatzick et al., 2002). Heavy alcohol consumption is thoughtto be both a marker for distress and a maladaptive coping strategy. Alcohol-related disorders are frequently co-morbid withother substance use disorders, depression, anxiety, and antisocial personality; these disorders may precede, coexist with, orfollow alcohol difficulties (McCrady, 1993). ‘‘Prolonged alcohol intoxication’’ can contribute to depressive-like symptoms(Schuckit, 1994, p. 46) and can increase depressive-like symptoms (Schuckit, 1994, p. 46), while interfering with an individ-ual’s ability to recover from a major depressive disorder (Mueller et al., 1994). In sum, despite the clear importance of exam-ining the impact of pre-MVC alcohol consumption on post-MVC distress, very few studies actually have done so in aprospective manner.

Studies examining sociodemographic characteristics (i.e., sex, age, marital status, household income, education, and race)and post-MVI distress have found mixed results. Most suggest that women are at greater risk than men for post-MVI distress,especially PTSD (Blanchard et al., 2004; Ehlers et al., 1998; Frommberger et al., 1998; Tolin & Foa, 2006; Ursano et al., 1999;Wu & Cheung, 2006; Zatzick et al., 2002) – up to 4.7 times greater risk (Fullerton et al., 2001). Although women’s injuriesoften are less severe than men’s, they perceive collisions as more frightening and report greater distress (Blaszczynskiet al., 1998; Ehlers et al., 1998). Yet other studies have found no sex differences in measures of post-trauma distress (Chiu,deRoon-Cassini, & Brasel, 2011; Cieslak, Benight, Luszczynska, & Laudenslager, 2011; Norris, 1992). Norris (1992) found nosex difference for post-MVC traumatic stress or PTSD, in her sample drawn from four southeastern American cities, althougha race by sex interaction was found, such that post-traumatic stress following motor vehicle collisions was strongest for menfrom African–American backgrounds.

The impact of other demographic characteristics on distress also have been examined. Age has been shown to be nega-tively correlated with post-trauma distress (Brewin et al., 2000; Chiu et al., 2011; Norris, 1992), although some studies have

N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13 3

found no relationship (Delahanty et al., 1997; Frans, Rimmö, Åberg, & Fredrikson, 2005; Mayou et al., 2001). Similarly, minor-ity ethnic or racial identification, unmarried status, and lower SES have been found to be associated with greater post-traumadistress (Brewin et al., 2000; Chiu et al., 2011; Frans et al., 2005; Norris, 1992), although not consistently so (Delahanty et al.,1997). It is important to note that because most studies have examined PTSD following a wide range of trauma, it is unclearwhether similar patterns would exist among MVI survivors, or when examining stress reactions other than PTSD.

Patterns of risk for post-trauma distress appear complex. For example, in a large health maintenance organization sample,Breslau, Davis, Andreski, Peterson, and Schultz (1997) found complex patterns of psychological predictors for PTSD, for in-stance, pre-existing anxiety or major depressive disorders were associated with increased risk of PTSD in women but not inmen. Similarly, in their National Comorbidity Study, Bromet et al. (1998) found that history of affective disorder predictedPTSD in women, whereas a history of anxiety disorder and parental mental disorder predicted PTSD in men. Norris (1992)found a race by sex interaction for post-traumatic stress following motor vehicle collisions. Despite these complex risk pat-terns, interaction effects among variables in relation to post-MVI distress typically have not been explored.

Much of this body of research has relied primarily on a (medical or psychiatric) treatment-seeking population, typically inBritain, Australia, or the USA (e.g., Blanchard et al., 1994; Blanchard et al., 2004), in which the psychologically distressedlikely are overrepresented. Often the studies have had limited sample sizes ranging from 30 to 158 participants (e.g. Blan-chard et al., 1996; Brasel et al., 2010; Cieslak et al., 2011; Delahanty et al., 1997; Fullerton et al., 2001; Mayou et al., 1997;Zatzick et al., 2002). Further, many studies examining psychological sequelae have examined the relationship between arange of traumas and distress, with motor vehicle collisions grouped together with war trauma, terrorist attacks, hurricanes,illness or death of loved one, and violence (Breslau et al., 1997; Bromet et al., 1998; Norris, 1992), despite evidence suggest-ing that type of injury is related to substantive differences in both post-trauma sequelae and risk factors for distress (Brewinet al., 2000; deRoon-Cassini et al., 2010; Norris,1992; Zatzick et al., 2002).

Additionally, most research conducted on pre-MVI predictors and post-MVI distress has been either cross-sectional, orprospective from the time of trauma onward, making it difficult to assess pre-MVI psychological functioning (King, Vogt,& King, 2004). Indeed, pre-MVI psychological functioning commonly has been assessed using post-MVI retrospective self-re-ports. This is problematic, because recall is state dependent: people may unknowingly alter their reports of the past to beconsistent with their current psychological functioning (Barsky, 2002; King et al., 2004). Further, although some studies haveused standardized instruments to assess pre-MVI psychological functioning (e.g., PTSD module of the Composite Interna-tional Diagnostic Interview), other studies have relied simply on history of psychiatric treatment (Harvey & Bryant, 1999),despite the fact that the absence of psychiatric treatment does not imply the absence of distress. Finally, most studies haveused nonparametric or non-nested regression techniques to assess the impact of pre-MVI psychosocial factors: thus, theyhave been unable to assess trajectories of post-MVI distress over time or predict change over time. Only one recent studyhas used hierarchical linear modeling to examine trajectories of post-MVI distress and to predict changes in distress overtime, but their sample was broad (level 1 trauma center patients experiencing a wide range of traumatic injuries) and theirfindings limited by six month follow-up (deRoon-Cassini et al., 2010).

To address some of the limitations of, and mixed findings in, the current body of literature, the present study aimed toexplore the trajectories of post-MVI distress and the effect of pre-MVI psychological functioning (pre-MVI distress and alco-hol use) and demographic characteristics on post-MVI distress, using hierarchical linear modeling to investigate the longi-tudinal, prospective dataset from the Canadian National Population Health Survey (NPHS). Risk factors were selected forinclusion based on the literature and on the availability of the measures within the survey. It was predicted thatpost-MVI distress would decrease gradually over time. Based on the previous literature, those individuals with greaterpre-MVI distress were expected to experience greater post-MVI distress. Similarly, those individuals with greater pre-MVIalcohol use were expected to experience higher post-MVI distress. Demographic characteristics that were expected to berelated to increased distress over time, based on the literature, included female gender, unpartnered status, non-Whiteidentification, younger age, and lower income.

Three strengths of this study, which address limitations in the current literature, are the longitudinal, prospective natureof the database, the large sample size, and the representativeness and generalizability of the sample to the Canadian popu-lation. As Charlton et al. (2004) state, population-based studies ‘‘provide the strongest recruitment approaches with leastpotential for bias’’ (p. 11).

2. Materials and methods

Due to the archival nature of the current study, formal institutional ethics approval was not required by The University ofWestern Ontario. However, the researchers were granted permission to conduct this research study by the Canadian Social Sci-ences and Humanities Research Council (SSHRC). Further, a security screening process was completed, a research contract withStatistics Canada (including a confidentiality agreement) was signed, analyses were conducted in a secure location, and all sta-tistical output was screened by Statistics Canada staff for confidential data prior to its release for use in the current manuscript.

2.1. Participants

The National Population Health Survey (NPHS) is a repeated measures longitudinal study conducted by Statistics Canadato monitor Canadians’ health. Wave 1 of the NPHS was conducted in 1994/1995, and participants were reassessed every two

4 N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13

years (i.e., Wave 2 1996/1997, Wave 3 1998/1999, Wave 4 2000/2001, and Wave 5 2002/2003). The NPHS target populationwas individuals residing in households in the ten provinces in Wave 1 (1994/1995), and excluded those living on Indian Re-serves, Crown Lands, in health institutions, in remote areas in Quebec and Ontario, or working full-time on Canadian ForcesBases. In the first Wave, respondents, selected from across Canada using a two-staged, stratified, random sampling proce-dure, completed the general component of the survey. Strata were formed by dividing the data by province and further sub-dividing each province into three types of areas (major urban centers, urban towns, and rural areas). The original sample wasobtained by selecting households via clustering techniques, with one member from each household randomly selected tocomplete the survey. In most strata, six clusters, usually Census Enumeration Areas, were selected with probability propor-tional to size. A rejective method sampling technique was used to correct for underrepresentation of groups of people in thepopulation and to account for undersampling from large households, thus reducing bias for potentially oversampling young,single-member households and the elderly. Participants provided informed consent, and a high response rate was achievedin Wave 1, with 88.7% agreeingto participate in the survey. This high rate was maintained through Wave 5, in which 80.8% ofthe original respondents participated. (For further methodological information, please consult Statistics Canada, 2010.)

The entire NPHS sample in Wave 1 comprised 17, 276 people, who were, on average, 35.25 years old (SD = 21.51 years).The gender composition was 49.5% male, 51.5% female, 49.2% were married or partnered and the remaining 50.7% wereunmarried, separated, divorced, or widowed. Race was reported as 88.9% White and 9.7% non-White. With regard to educa-tion, 43.1% reported having completed at least some post-secondary education; 40.2% had completed high school educationor less. Just over 63% reported earning over $30,000, and 31.7% earning under $30,000 per year. Data were weighted, usingStatistics Canada’s Longitudinal Square Weight (i.e., WT64LS) to most accurately represent the age and sex composition ofthe population of the 10 Canadian provinces in Wave 1 (for further information on weighting, please see Statistics Canada.(2009)). For all subsequent analyses, these weights were adjusted to represent the sample size of the single-MVI subpopu-lation in Waves 1–5 of the NPHS data, rather than the NPHS sample as a whole.

To assess the validity and representativeness of self-reported injury rates in the NPHS, an initial study compared theNPHS self-reported MVI rates with police collision reports of MVI rates, based on Transport Canada’s Traffic Accident Infor-mation Database (TRAID) for Canadian provinces and territories (Roberts, Vingilis, Wilk, & Seeley, 2007). The age and sextrends for both datasets were highly similar and showed no significant differences. The current study’s sample contained536 individuals who reported sustaining a MVI during participation in Waves 1–5 (Wave 1 = 153 participants; Wave2 = 95 participants; Wave 3 = 104 participants; Wave 4 = 104; and Wave 5 = 80 participants). Seventeen participants whohad reported more than one MVI were excluded from the subsample. The MVI sample was 44.7% male and 55.3% female.In the wave prior to their MVI, participants’ average age was 32.71 years (SD = 15.42 years). Of those who reported race(2.3% did not), 89% were white and 11% were non-white. Pre-MVI marital status (when reported; 30.6% did not report maritalstatus) was 48.5% married or partnered and 51.5% not married or partnered. Although 36.6% of individuals did not reportincome, of those reporting, 33.1% reported total household income from all sources to be below $30,000, and 66.9% reportedincome above $30,000. Of those reporting (33.2% did not), 54.6% reported completing at least some post-secondary education(45.4% had not). These findings are representative of the Canadian population at the time of the 1996 Census (Statistics Can-ada, 2012), in which the population was 49.1% male and 50.9% female; 88.8% White and 11.2% non-White; 36% reported highschool education or less; and average family income was $54,584.

Because participants sustained MVIs at different waves of the NPHS, data were recoded, such that Time 0 corresponded todata collected in the wave prior to reporting the MVI (applicable for all participants except those who sustained MVI at Wave1), Time 1 corresponded to the time at which MVI in the past 12 months was reported (applicable to all participants). Time 2corresponded to the wave after which MVI was reported (applicable to all participants except those reporting injuries inWave 5); Time 3, two waves afterward (applicable to all participants except those reporting injuries at Time 4 or 5); andso forth, to Time 5 (applicable to all participants except those who reported injuries in Waves 2–5). As such, sample sizeis greatest at Time 1 and dwindles over time.

2.2. Measures

2.2.1. Motor vehicle injuriesThe MVI measure was derived from two questions: (1) ‘‘In the past 12 months, did you have any injuries that were serious

enough to limit your normal activities?’’ Participants who answered ‘‘yes’’ were then asked, (2) ‘‘What happened?’’ Partic-ipants who cited a ‘‘transportation accident’’ as the cause of their injuries were coded as 1 (indicating a MVI); all others werecoded as zero and excluded from analyses. The MVI variable was assessed from Wave 1 through Wave 5.

2.2.2. Distress scale – K6 (Kessler et al., 2002)The Kessler Psychological Distress Scale consists of six questions addressing ‘‘nonspecific psychological distress’’ during

the past month (Statistics Canada, 2009), using a 5-point, Likert-style scale. Higher scores are indicative of greater psycho-logical distress. Items are a subset of questions from the US National Health Interview Survey (Kessler et al., 2002; Robinset al., 1988). Sample questions include: ‘‘How often did you feel so sad that nothing could cheer you up?’’, ‘‘How oftendid you feel nervous?’’, ‘‘How often did you feel restless or fidgety?’’, and ‘‘How often did you feel that everything was aneffort?’’ Scores could range from 0 to 24. This scale is a good indicator of current anxiety and major mood disorders andhas better overall discriminatory power than the General Health Questionnaire-12 (GHQ-12) in detecting DSM-IV depressive

N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13 5

and anxiety disorders (Andrews & Slade, 2001; Furukawa, Kessler, Slade, & Andrews, 2003). The scale was found to stronglydiscriminate between community cases and non-cases of these disorders using DSM-IV/SCID criteria, and to have good pre-cision in the 90–99th percentile range of the population distribution (standard errors of standardized scores in the range0.20–0.25) (Kessler et al., 2002). Moreover, the K6 has been shown to be an accurate predictor of depression in the Canadianpopulation (Cairney, Veldhuizen, Wade, Kurdyak, & Streiner, 2007). Internal consistency, defined by Cronbach’s alpha, ran-ged from .76 to .85 in international studies (Arnaud et al., 2010; Baggaley et al., 2007; Haswell et al., 2010), and, in a studyusing the same NPHS data, was .80 for males and .81 for females (Gadalla, 2009).

2.2.3. Weekly alcohol consumptionWeekly alcohol consumption is the sum of all alcoholic beverages consumed during the past week, for individuals who

reported consuming at least one alcoholic beverage in the past 12 months (Statistics Canada, 2009). A drink was defined asone bottle of beer or a glass of draught, one glass of wine or a wine cooler, or one drink or cocktail with one and a half ouncesof liquor. In the present study, alcohol consumption was assessed in the wave prior to the MVI.

2.2.4. Demographic characteristicsAge was measured in years. Sex was coded such that male = 0 and female = 1. Marital status was coded, such that ‘‘mar-

ried or partnered’’ = 0 and ‘‘not married or partnered’’ = 1. Race was coded, such that ‘‘white’’ = 0 and ‘‘non-white’’ = 1. Edu-cation was coded, such that ‘‘high school education or less’’ was 0 and ‘‘at least some post-secondary education’’ was 1.Finally, household income was coded, such that a score of 0 corresponded to earning under $30, 000 per year, and a scoreof 1, to earning over $30,000 per year.

2.3. Statistical analyses

This study employed hierarchical linear modeling (HLM) as a general data analytic strategy, which is conceptualized ‘‘as ahierarchical system of regression equations’’ (Hox, 1998, p. 148). This method represents the most powerful way to examinechange in a continuous dimension over time within persons (Raudenbush & Bryk, 2002). This method for analyzinglongitudinal data essentially allows each participant in a sample to have a different growth trajectory (or growth curve),representing change over time. In the present analyses, assessment point [Time 1 (reporting MVI in past 12 months),Time 2(post-MVI assessment point 2), Time 3 (post-MVI assessment point 3), Time 4 (post-MVI assessment point 4), and Time 5(post-MVI assessment point 5)] was nested within individuals. This was done to account for dependency in an individual’sdistress over time, and to examine the trajectory of distress over time. An exploratory, model-building approach wasadopted. Data for continuous variables were grand-mean centered to improve interpretability and to reduce multicollinear-ity, per recommendations by Aiken and West (1991) and Raudenbush and Bryk (2002). Interactions were explored post hoc,using the pick-a-point method, as described by Preacher, Curran, and Bauer (2006). Prior to conducting analyses, data werecarefully examined. Data, including individual trajectories of distress over time, were visually inspected for the presence ofoutliers. No obvious outliers were visible, and the decision was made to retain somewhat more extreme data points, giventhat in a sample of this size, one cannot definitively conclude that an observation is not part of the population of interest, andto preserve statistical power. Diagnostics were performed, and data were examined to ensure they met the assumptions ofeach statistical test. Violations of assumptions were rare, and minor at most. Alpha was set at .05 for all statistical tests.

3. Results

The intra-class correlation coefficient (ICC) assesses the proportion of total observed variability in a measure that is attrib-utable to differences between groups (i.e., in this study, between individuals). The ICC also can be interpreted as a correlationcoefficient, pooling across groups (i.e., correlation between Time 1–5 scores across individuals). An ICC differentiates within-and between-group variability and also incorporates information about actual scale values, rather than simply covariance. Inthese ways, an ICC differs from a correlation coefficient, which is computed using overall scale variability and covariance inscores, without taking into account actual scores. The higher an individual’s ICC, the more similar that individual’s scores areacross time. Thus, a high individual ICC would suggest high similarity in an individual’s distress scores over time. In the pres-ent study, the ICC was 0.32, indicating that nestedness within person accounted for 32% of variance in distress scores overtime (Time 1–5).

HLM analyses were conducted using the proc mixed command within SAS software, using restricted maximum likelihood(REML) as the estimation method, rather than Full Information Maximum Likelihood (FIML), because REML provides moreaccurate estimates of random effects than does FIML. To determine the impact of demographic characteristics (i.e., age, sex,marital status, household income, education, and race) on distress over time, these variables were entered as predictors inthe first model. Pre-MVI (Time 0) distress also was entered as a predictor to control for its impact on post-MVI distress overtime. In addition, total weekly alcohol consumption was used to predict post-MVI distress. Descriptive data for these vari-ables and for distress over Times 1–5 are provided in Table 1.

An exploratory, model-building approach was adopted. First, an unconditional growth model for distress was run, inwhich both slope and intercept were allowed to vary randomly. This model revealed a significant fixed effect of time on

6 N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13

distress. That is, from Time 1–5, distress increased significantly, on average, 7.65 units of distress per time period (SE = 0.85,t(1237) = 9.02, p < .01). This suggests that, across participants, distress tends to increase over time. Random effects of bothintercept (s = 217.63, SE = 36.12, z = 6.02, p < .01) and slope/time (s = 141.04, SE = 19.17, z = 7.36, p < .01) were significant,suggesting that both average distress scores and trajectories of distress over time varied among participants. The advantageof growth curve modeling over more traditional examination of group data over time is demonstrated by the finding thatacross participants, growth curve modeling identified an increase in individual distress over time. However, examinationof group mean scores over time (as in Table 1), which does not take into account individual-level dependency in data, showsa different pattern altogether.

Next, an unconditional growth model allowing for heterogenous variance (that is, different variance structures in distressscores at different assessment points) was run through SAS. However, the model could not be completed, due to infinite like-lihood. As such, a homogeneous variance structure was assumed. Subsequently, an unconditional growth model was runassuming homogeneous variance structure but allowing both fixed and random effects of intercept (mean) and of timeand quadratic time (i.e., differential increase in distress scores across different time periods). However, the fixed quadraticeffect was nonsignificant (b = �0.33, SE = 0.51, t(1236) = �0.66, p > .05), even though the random quadratic effect was signif-icant (s = 35.12, SE = 6.49, z = 5.42, p < .01), and, as such, the quadratic effect of time was excluded in subsequent models.

3.1. The role of demographic characteristics

To determine the impact of demographic characteristics on distress over time, age, sex, race, marital status, income, andeducation were used as predictor variables in a growth model that incorporated both fixed and random effects of interceptand time. Random effects of both intercept/mean (s = 80.79, SE = 30.15, z = 2.68, p < .01) and slope (s = 218.91, SE = 42.76,z = 5.12, p < .01) remained, indicating that despite inclusion of demographic variables, unexplained inter-individual varianceremained in both average distress scores and in changes in distress scores over time. Fixed effects (presented in Table 2) indi-cate that both sex and marital status predicted levels of post-MVI distress. That is, men in this sample reported greater psy-chological distress than did women, and those participants who were nonpartnered reported greater distress than those whowere married or partnered pre-MVI. Because both sex and marital status significantly predicted distress, these were retainedas predictor variables in all subsequent analyses to control for their impact. Given the small correlation between these vari-ables (r = �.01), multicollinearity was not a concern. To conserve power, nonsignificant demographic variables were droppedfrom subsequent analyses.

3.2. Pre-MVI risk factors

3.2.1. The role of distressTo better understand the impact of pre-MVI distress on post-MVI distress, a growth model was run with the following

predictors: sex, pre-MVI marital status, pre-MVI distress, time, and interactions between all predictors. Correlations amongpredictor variables are provided in Table 3, with most relatively small (r = .10 between pre-MVI distress and sex and betweenpre-MVI distress and marital status).

Random effects of both intercept (s = 52.52, SE = 23.84, z = 2.20, p < .05) and slope/time (s = 204.53, SE = 38.90, z = 5.26,p < .01) were significant, indicating that even after accounting for the variance in post-MVI distress explained by thesepredictors, inter-individual differences remained both in average distress and changes in distress over time. Fixed effects(presented in Table 4) revealed significant interactions between (a) sex and time, and (b) pre-MVI marital status, time,and pre-MVI distress. As is demonstrated in Fig. 1, both men and women showed an increase in distress over time. However,that increase was steeper for men (simple slope = 10.87, p < .01) than for women (simple slope = 2.84, p = .29). In fact, theslope for women was nonsignificant. Next, the pre-MVI marital status � time � pre-MVI distress interaction, as seen inFig. 2, reveals that among those with partners, individuals experiencing greater pre-MVI distress (i.e., +1 SD) continued toexperience greater post-MVI distress than their lower-pre-MVI distress (i.e., �1 SD) peers over time. For those with partners,distress tended to increase relatively evenly over time, regardless of pre-MVI distress (for those with partners and greaterpre-MVI distress, simple slope = 12.73, p < .05; for those with partners and lower pre-MVI distress, simple slope = 9.01,p < .05). However, for those without partners, a different pattern emerged. Those with lower pre-MVI distress (compared

Table 1Descriptive data.

Variable M SD N

Time 0 (Pre-MVI) Distress 3.83 3.93 338Time 0 (Pre-MVI) Weekly Total Alcohol Consumption 4.51 8.57 298Time 1 Distress 4.30 4.52 503Time 2 Distress 3.38 3.88 376Time 3 Distress 3.56 3.78 270Time 4 Distress 3.46 3.58 175Time 5 Distress 3.41 3.76 88

1Distress scores could range from 0 to 24.

Table 2Fixed effects of demographic variables in predicting distress over time.

Predictor variable b SE df t p

Intercept �12.00 3.63 318 �3.30 <.01Time 8.15 1.40 519 5.84 <.01Age (Time 0) �0.06 0.09 318 �0.70 .48Sex (Time 0) �8.31 2.36 318 �3.53 <.01Race (Time 0) �3.33 4.21 318 �0.79 .43Marital Status (Time 0) 6.44 2.63 318 2.45 <.05Education (Time 0) �1.70 2.45 318 �0.70 .49Income (Time 0) �0.42 2.59 318 �0.16 .87

Table 3Correlations among predictor variables.

Distress Alcohol Sex Marital Status

Distress 1.00Alcohol �0.04 1.00Sex 0.10** �0.29** 1.00Marital Status 0.10** 0.02 �.01 1.00

* p < .05.** p < .01.

Table 4Fixed effects of pre-MVI distress and demographic variables predicting post-MVI distress.

Predictor variable b SE df t p

Intercept �13.00 2.70 354 �4.82 <.01Sex (Time 0) �4.14 3.52 354 �1.18 .24Marital status (Time 0) 9.90 3.57 354 2.77 <.01Sex �Marital Status (Time 0) �3.53 4.78 354 �0.74 .46Time 10.87 2.89 560 3.77 <.01Sex � Time �8.03 3.94 560 �2.04 <.05Marital Status � Time �1.59 3.89 560 �0.41 .68Sex �Marital Status � Time 7.59 5.46 560 1.39 0.17Distress (Time 0) 0.30 0.13 354 2.27 <.05Sex � Distress �0.25 0.16 354 �1.58 0.12Marital Status � Distress 0.27 0.15 354 1.84 0.07Sex �Marital Status � Distress 0.19 0.20 354 0.94 .35Time � Distress 0.06 0.13 560 0.44 .66Sex � Time � Distress �0.12 0.17 560 �0.68 .50Marital Status � Time � Distress �0.30 0.15 560 �2.01 <.05Sex �Marital Status � Time � Distress 0.17 0.24 560 0.70 .49

Time 1 Time 5

Post-MVI Distress

Assessment Time

Male

Female

Higher

Lower

Fig. 1. Sex by time interaction predicting post-MVI distress.

N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13 7

Time 1 Time 5

Post-MVI Distress

Assessment Time

Low Pre-MVI Distress, Partnered

High Pre-MVI Distress, Partnered

Low Pre-MVI Distress, Unpartnered

High Pre-MVI Distress, Unpartnered

Higher

Lower

Fig. 2. Marital status by time by pre-MVI distress interaction predicting post-MVI distress.

8 N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13

to their higher pre-MVI distress peers) showed lower distress at Time 1, but also a steeper increase in post-MVI distress(simple slope = 17.02, p < .01), such that ultimately, they experienced greater post-MVI distress than their higher pre-MVIdistress peers, who showed no significant increase in distress over time (simple slope = 1.54, p = .06).

3.2.2. The role of alcoholNext, the present study aimed to understand the impact of pre-MVI alcohol consumption on post-MVI distress, control-

ling for the effects of sex, pre-MVI marital status, and pre-MVI distress. As such, these variables were incorporated as pre-dictors in a growth model with fixed and random effects of intercept and time. The correlation between distress and alcoholconsumption at Time 0 was nonsignificant (r = �.04), and as such, multicollinearity did not pose a problem. Although thecorrelation between sex and pre-MVI alcohol consumption was significant (r = �.29), it was considered essential to controlfor the effects of sex in understanding the role of pre-MVI alcohol consumption in predicting post-MVI distress. As such, thispredictor variable was retained in this analysis.

Random effects of both intercept (s = 75.25, SE = 24.71, z = 3.05, p < .01) and slope/time (s = 218.03, SE = 40.81, z = 5.34,p < .01) remained, indicating that even after accounting for the variance in post-MVI distress explained by these predictors,inter-individual differences remained both in average distress and changes in distress over time. Fixed effects (presented inTable 5) revealed significant effects of sex, pre-MVI marital status, and two two-way interactions: (a) between pre-MVI alco-hol consumption and distress, and (b) between pre-MVI distress and time. As in the previous analysis, men reported greaterdistress than did women, and those individuals who were nonpartnered reported worse distress than those who were part-nered pre-MVI.

The interaction between pre-MVI alcohol consumption and pre-MVI distress (Fig. 3) indicates that, in general, those indi-viduals experiencing greater pre-MVI distress (i.e., +1 SD) also experienced greater post-MVI distress. For those with greaterpre-MVI weekly alcohol consumption (i.e., +1 SD), this relationship is strongest (i.e., the slope was steeper; simpleslope = 0.44, p < .01), followed by those with average pre-MVI alcohol consumption (simple slope = 0.27, p < .01), and finally,by those with lower (i.e.,�1 SD) pre-MVI alcohol consumption (simple slope = 0.11, p < .05). At low levels of pre-MVI distress(i.e., �1 SD), greater pre-MVI alcohol consumption may be associated with lower post-MVI distress. However, as pre-MVIdistress increased, this effect was reversed, such that those with the greater pre-MVI alcohol consumption (i.e., +1 SD) expe-rienced greater post-MVI distress.

In addition, the interaction between pre-MVI distress and time (controlling for alcohol effects) was explored. As demon-strated in Fig. 4, controlling for alcohol effects, distress tended to increase over time, but that increase was much steeper and

Table 5Fixed effects of alcohol consumption, distress, and demographic variables in predicting post-MVI distress.

Predictor variable b SE df t p

Intercept �11.91 2.09 356 �5.69 <.01Sex (Time 0) �6.04 2.22 356 �2.72 <.01Marital Status (Time 0) 6.32 2.22 356 2.84 <.01Time 7.22 1.45 564 4.99 <.01Distress (Time 0) 0.27 0.05 356 5.09 <.01Alcohol (Time 0) 0.00 0.00 356 1.56 .12Distress � Alcohol (Time 0) 0.0004 0.00 356 4.05 <.01Time � Distress (Time 0) �0.15 0.07 564 �2.23 <.05Time � Alcohol (Time 0) �0.00 0.00 564 �1.01 .31Time � Distress � Alcohol (Time 0) 0.00 0.00 564 0.08 .93

Post-MVI Distress

Low Pre-MVI Alcohol Consumption

Mean Pre-MVI Alcohol Consumption

High Pre-MVI Alcohol Consumption

Higher Pre-Injury Distress

Lower Pre-InjuryDistress

Higher

Lower

Fig. 3. Pre-MVI alcohol by pre-MVI distress interaction predicting post-MVI distress.

N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13 9

more pronounced for those with lower pre-MVI distress (i.e., �1 SD; simple slope = 11.93, p < .01) than for those with aver-age distress (simple slope = 7.22, p < .01) or higher pre-MVI distress (i.e., +1 SD; simple slope = 2.52, p = .39). For the higherpre-MVI distress group, this increase did not even attain statistical significance. Indeed, at post-MVI Time 1, those with lowerpre-MVI distress experienced lower post-MVI distress. However, by Time 5, this trend had reversed, such that those withgreater pre-MVI distress appeared to experience lower post-MVI distress than those with lower pre-MVI distress.

4. Discussion

The purpose of the current study was to explore trajectories of post-MVI distress and the effect of demographics and pre-MVI psychological functioning on post-MVI distress, using a prospective design. This represents an advance in the literature,because the NPHS allowed prospective exploration of the impact of MVI, rather than retrospective report of pre-MVI func-tioning, which may be biased. Second, the sample was large and participants were representative of the broader Canadianpopulation, rather than limited to treatment-seeking patients, among whom one might reasonably expect distress to beoverrepresented. In addition, this study followed participants up to nine years post-MVI, in contrast to previous research thatfollowed participants for much shorter time periods. Moreover, exceedingly few studies assessed interaction effects amongvariables to explore complex patterns and relationships.

Although distress scores in the current study were generally low, substantial variability existed within the sample. Theresults of analyses are suggestive of complex patterns of post-MVI distress over time. These may occur over many years,suggesting possible delayed onset of psychological sequelae, as others have found (Bonanno, 2004; Bonanno, 2005;deRoon-Cassini et al., 2010; Mayou et al., 1997). Indeed, in the present study, contrary to initial hypothesis, post-MVI distressincreased over time, although average distress scores and trajectories of distress over time varied. Individual characteristicscontributed sizably toward distress over time, with approximately one third of the variance in distress scores over timeattributable to nestedness withinindividual.

An examination of individual characteristics revealed two demographic characteristics, sex and marital status, were asso-ciated with post-MVI distress. Contrary to hypotheses, age, race, income and education were not associated with post-MVIdistress. Again, contrary to initial expectations, overall, women tended to report less long-term distress than men over time.As hypothesized, those who were married or partnered pre-MVI were less distressed than their nonpartnered counterparts.The finding that men reported greater post-MVI long-term distress over time stands in contrast to typical gender differencesin post-MVI distress reporting, either with greater distress in women (e.g., Blanchard et al., 2004; Bryant & Harvey, 2003;

Time 1 Time 5

Post-MVI Distress

Assessment Time

Low Pre-MVI Distress

Mean Pre-MVI Distress

High Pre-MVI Distress

Higher

Lower

Fig. 4. Pre-MVI distress by time interaction predicting post-MVI distress.

10 N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13

Ehlers et al., 1998; Frommberger et al., 1998; Fullerton et al., 2001; Ursano et al., 1999; Wu & Cheung, 2006), or with no gen-der difference (e.g., Chiu et al., 2011; Cieslak et al., 2011; Kupchik et al., 2007; Norris, 1992). It is unclear whether this findingmight represent delayed-onset post-MVI distress that was not captured by other studies, given that the vast majority of lon-gitudinal studies on post-trauma psychological sequelae have used shorter follow-up periods. Donahue (2007) found thatdespite an initial tendency for women to report greater distress than men at one-month post-MVI, these gender differencesdissipated over time. Alternatively, this finding could reflect differences between clinical and population-based samples; re-search finds that females are more likely than males to use medical services and thus may be overrepresented in clinicalsamples (Vingilis, Wade, & Seeley, 2006). Finally, the finding could be due to the fact that men are more likely than womento be injured as drivers than as passengers (Tavris, Kuhn, & Layde, 2001) and thus, men as drivers are more likely to be cul-pable in the crashes. Delahanty et al. (1997) have found culpability for MVI collisions to be associated with greater post-MVIdistress.

Apart from gender, the results suggest that individuals who may have fewer pre-MVI coping resources show greaterpost-MVI distress. It is not surprising that those who were married or partnered pre-MVI experienced less distress over time,given that marriage/partnership provides social support, buffering the individual against life stressors (Holt-Lunstad,Birmingham, & Jones, 2008). This is consistent with the findings of other researchers (Akhtar-Danesh & Landeen, 2007;Cairney, Corna, Veldhuizen, Kurdyak, & Streiner, 2008; Gabert-Quillen et al., 2012; Patten et al., 2006; Wang & El-Guebaly,2004). Research by Holt-Lunstad et al. (2008) showed that not only was marital status relevant, but also the quality of thatrelationship; those with higher-quality relationships showed better physical and mental health outcomes. Those with poorquality relationships fared worse in this regard than single people. The NPHS did not examine quality of the partnerrelationship on post-MVI distress – this would be an interesting avenue for future research.

4.1. Pre-MVI risk factors

When pre-MVI distress was used to predict post-MVI distress, complex findings emerged. It had been hypothesized thatgreater pre-MVI distress would predict greater post-MVI distress. Results revealed that for those who were partnered priorto the MVI, distress increased relatively evenly, regardless of pre-MVI distress. [It should be noted that, consistent with theliterature (e.g., Blanchard et al., 1994; Blanchard et al., 2004; Bryant & Harvey, 1995; Ehlers et al., 1998; Mayou et al., 2001;Ursano et al., 1999), and despite similar trajectories, those with higher pre-MVI distress experienced greater post-MVI dis-tress than their low pre-MVI distress counterparts.] In contrast, among those without partners, participants with lower pre-MVI distress showed a greater increase in distress over time than did their higher pre-MVI distress peers. This suggests thatrelationships serve as a support to buffer the effect of distress over time, particularly among those with below-average dis-tress, who may have less experience and thus fewer skills in managing distress, leading them to experience more distressover time. In addition, analyses revealed that, regardless of marital status, men, who were previously shown to experiencegreater distress overall, also experienced a greater increase in distress over time than did women.

When pre-MVI weekly alcohol consumption was used to predict post-MVI distress, men again reported greater distressover time than did women, and the nonpartnered reported worse distress than the partnered. Pre-MVI distress predictedpost-MVI distress, but that relationship was strongest for those with greater pre-MVI alcohol consumption. Indeed, at lowlevels of pre-MVI distress, it appeared that greater pre-MVI alcohol consumption was related to lower post-MVI distress,but at high levels of pre-MVI distress, greater pre-MVI alcohol consumption predicted increased post-MVI distress. This find-ing may be because alcohol consumption can be a maladaptive strategy for coping with high levels of distress, and ulti-mately, results in its own negative consequences which themselves serve to increase distress. Alcohol consumption, atlow levels of distress, may be a reasonable or neutral way to reduce distress, but at higher levels of distress, it may becomemore dysfunctional. This is consistent with the alcoholism literature, which indicates that in low to moderate amounts, alco-hol has a sedative or physiological ‘‘stress response dampening’’ effect (Levenson, Sher, Grossman, Newman, & Newlin, 1980,p. 536; Sher, Bartholow, & Peuser, 2007), whereas heavy alcohol consumption/intoxication contributes to depressive-likesymptoms (Schuckit, 1994) and reduces ability to recover from major depressive disorder (Mueller et al., 1994). With theeffect of pre-MVI alcohol use controlled for, post-MVI distress tended to increase over time, particularly among those withlower pre-MVI distress. This may be due to a distress inoculation effect (Seery, Holman, & Silver, 2010), such that those expe-riencing more distress pre-MVI had more experience with distress, and thereby were able to develop coping skills thathelped them to experience a lesser increase in subsequent distress over time (this effect may not hold for those with extre-mely high distress, whose coping resources may become overwhelmed). Alternatively, this may represent regression to themean. Previous findings that post-MVI distress generally increased over time more steeply for those more highly distressedpre-MVI may have been confounded by the effect of pre-MVI alcohol consumption on post-MVI distress. Thus, the role ofalcohol consumption in post-MVI psychological functioning may be complex and nuanced. Unfortunately, exceedinglyfew studies have actually examined pre-MVI alcohol use/abuse in relation to post-trauma (or, specifically, to post-MVI) psy-chological functioning. Vingilis et al. (1996) found that the total number of drinks per occasion pre-MVI predicted psycho-logical sequelae, which included measures of distress, at one year post-MVI. In their study of psychiatric patients withanxiety disorders, Zlotnick, Warshaw, Shea, and Keller (1997) found that those with trauma histories or PTSD were morelikely to have pre- and post-trauma alcohol abuse. Clearly, much more research is needed to better understand the relation-ship between pre-MVI patterns of and problems with alcohol use and post-MVI psychological functioning.

N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13 11

The current, preliminary study has limitations and thus, the results should be viewed with these in mind. The presentstudy examined self-reported distress over time, but did not examine the impact and/or trajectories of clinical syndromes(e.g., depression, anxiety, PTSD, etc.) diagnosed by a health professional. Future research might incorporate measures suchas the Structured Clinical Interview for DSM-IV (First, Gibbon, Spitzer, Williams, & Benjamin, 1997; First, Spitzer, Gibbon, &Williams, 1997). In addition, the NPHS measured distress only in two year intervals, and as such, the patterns of distress overthe first two years post-MVI are unmeasured; further, trajectories may vary within any two-year interval. Moreover, thesedata are based on self-reported ‘‘injuries serious enough to limit normal activities’’, and thus, actual injury severity is un-known. [It should be noted that the effect of injury severity on post-trauma psychological sequelae has been mixed, withtwo studies finding a relationship (Blanchard et al.,1996; Brewin et al., 2000), and other studies finding none (Braselet al., 2010; Chiu et al., 2011; Ehlers et al., 1998; Taylor & Koch, 1995; Zatzick et al., 2002). The present study also didnot examine risk factors for MVI (e.g., alcohol consumption in the context of driving while intoxicated), but rather factorspredicting post-MVI outcome. Finally, this study explored predictors of post-MVI distress among a representative sampleof Canadians who reported experiencing a MVI over a 10-year span, but did not compare distress of MVI victims with thosewho had not experienced a MVI. Thus, this study can only reflect trajectories of distress and predictors of distress within theMVI sample. It is important to note that findings may not generalize, for example, to those individuals who experienced amotor vehicle collision in which they were not injured, or to those who experienced minor injuries that did not limit theirnormal activities.

Despite its limitations, this study has shown complex patterns of post-MVI distress over time and complex relationshipswith pre-MVI distress, alcohol consumption, sex, and marital status over a nine year period. As such, clinicians working withindividuals with MVIs ought to attend to pre-MVI functioning (e.g., distress and alcohol consumption) and collaborate withclients in routine assessment of distress and the building of psychosocial resources to enhance coping. Given the trajectoriesof distress over time, screening could be warranted for several years post-MVI. In so doing, individuals might be identifiedwho could benefit from additional support. Nonetheless, given the paucity of long term studies on pre-MVI risk factors andpost-MVI distress, there is great need for future research examining the potentially complex relationships among risk factorsand post-MVI psychological distress over time, different distress trajectories, and particularly, delayed-onset distress.

Acknowledgements

This research was supported by a grant from AUTO21, a member of the Networks of Centres of Excellence (NCE) program,which is administered and funded by the Natural Sciences and Engineering Research Council (NSERC), the Canadian Insti-tutes of Health Research (CIHR), and the Social Sciences and Humanities Research Council (SSHRC), in partnership withIndustry Canada. The authors also wish to acknowledge SSHRC, Canadian Initiative on Social Statistics (CISS) Access to Re-search Data Centres. In addition, they would like to thank the following for their assistance: Drs. P. Wilk and J. Braun (fortheir sage advice during statistical consultation) and Ms. J. Seeley (for her assistance throughout the process). While the re-search and analysis are based on data from Statistics Canada, the opinions expressed do not necessarily represent the viewsof Statistics Canada.

References

Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. California: Sage Publications Inc..Akhtar-Danesh, N., & Landeen, J. (2007). Relationship between depression and sociodemographic factors. International Journal of Mental Health Systems, 1,

ArtID4.Andrews, G., & Slade, T. (2001). Interpreting scores on the Kessler Psychological Distress Scale (K10). Australia and New Zealand Journal of Public Health, 25(6),

494–497.Arnaud, B., Malet, L., Teissedre, F., Izaute, M., Moustafa, R., Geneste, et al (2010). Validity study of Kessler’s Psychological Distress Scales conducted among

patients admitted to a French emergency department for alcohol consumption-related disorders. Alcoholism: Clinical and Experimental Research, 34(7),1235–1245.

Baggaley, R. F., Ganaba, R., Filippi, V., Kere, M., Marshall, T., Sombié, I., et al (2007). Detecting depression after pregnancy: The validity of the K10 and K6 inBurkina Faso. Tropical Medicine and International Health, 12(10), 1225–1229.

Barsky, A. J. (2002). Forgetting, fabricating, and telescoping: The instability of the medical history. Archives of Internal Medicine, 162(9), 981–984.Beck, J. G., & Coffey, S. F. (2007). Assessment and treatment of posttraumatic stress disorder after a motor vehicle collision: Empirical findings and clinical

observations. Professional Psychology: Research and Practice, 38(6), 629–639.Blanchard, E. B., Hickling, E. J., Barton, K. A., Taylor, A. E., Loos, W. R., & Jones-Alexander, J. (1996). One-year prospective follow-up of motor vehicle accident

victims. Behaviour Research and Therapy, 34(10), 775–786.Blanchard, E. B., Hickling, E. J., Freidenberg, B. M., Malta, L. S., Kuhn, E., & Sykes, M. A. (2004). Two studies of psychiatric morbidity among motor vehicle

accident survivors 1 year after the crash. Behaviour Research and Therapy, 42(5), 569–583.Blanchard, E. B., Hickling, E. J., Taylor, A. E., Loos, W. R., & Gerardi, R. J. (1994). Psychological morbidity associated with motor vehicle accidents. Behaviour

Research and Therapy, 32(3), 283–290.Blaszczynski, A., Gordon, K., Silove, D., Sloane, D., Hillman, K., & Panasetis, P. (1998). Psychiatric morbidity following motor vehicle accidents: A review of

methodological issues. Comprehensive Psychiatry, 39(3), 111–121.Bonanno, G. A. (2004). Loss, trauma, and human resilience: Have we underestimated the human capacity to thrive after extremely aversive events?

American Psychologist, 59(1), 20–28.Bonanno, G. A. (2005). Resilience in the face of potential trauma. Current Directions in Psychological Science, 14(3), 135–138.Brasel, K. J., deRoon-Cassini & Bradley, C. T. (2010). Injury severity and quality of life: Whose perspective is important? Journal of Trauma, 68(2), 263–268.Breslau, N., Davis, G. C., Andreski, P., Peterson, E. L., & Schultz, L. R. (1997). Sex differences in posttraumatic stress disorder. Archives in General Psychiatry,

54(11), 1044–1048.

12 N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13

Brewin, C. R., Andrews, B., & Valentine, J. D. (2000). Meta-analysis of risk factors for posttraumatic stress disorder in trauma-exposed adults. Journal ofConsulting and Clinical Psychology, 68(5), 748–766.

Bromet, E., Sonnega, A., & Kessler, R. C. (1998). Risk factors for DSM-III-R posttraumatic stress disorder: Findings from the National Comorbidity Survey.American Journal of Epidemiology, 147(4), 353–361.

Bryant, R. A., & Harvey, A. G. (1995). Avoidant coping style and post-traumatic stress following motor vehicle accidents. Behaviour Research and Therapy,33(6), 631–635.

Bryant, R. A., & Harvey, A. G. (2003). Gender differences in the relationship between acute stress disorder and posttraumatic stress disorder following motorvehicle accidents. Australian and New Zealand Journal of Psychiatry, 37(2), 226–229.

Cairney, J., Corna, L. M., Veldhuizen, S., Kurdyak, P., & Streiner, D. L. (2008). The social epidemiology of affective and anxiety disorders in later life in Canada.Canadian Journal of Psychiatry, 53(2), 104–111.

Cairney, J., Veldhuizen, S., Wade, T. J., Kurdyak, P., & Streiner, D. L. (2007). Evaluation of two measures of psychological distress as screeners for depression inthe general population. Journal of Psychiatry, 52(2), 111–120.

Charlton, J. L., Koppel, S., O’Hare, M., Andrea, D., Smith, G., Khodr, B, et al. (2004). Influence of chronic illness on crash involvement of motor vehicle drivers.Victoria, Australia: Monash University Accident Research Centre (Rep No. 213).

Chiu, K. B., deRoon-Cassini & Brasel, K. J. (2011). Factors identifying risk for psychological distress in the civilian trauma population. Academic EmergencyMedicine, 18(11), 1156–1160.

Cieslak, R., Benight, C. C., Luszczynska, A., & Laudenslager, M. L. (2011). Longitudinal relationships between self-efficacy, post-traumatic distress and salivarycortisol among motor vehicle accident survivors. Stress and Health, 27(3), e261–e268.

Delahanty, D. L., Herberman, H., Craig, K. J., Hayward, M. C., Fullerton, C. S., Ursano, R. J., et al (1997). Acute and chronic distress and posttraumatic stressdisorder as a function of responsibility for serious motor vehicle accidents. Journal of Consulting and Clinical Psychology, 65(4), 560–567.

deRoon-Cassini, T. A., Mancini, A. D., Rusch, M. D., & Bonanno, G. A. (2010). Psychopathology and resilience following traumatic injury: A latent growthmixture model analysis. Rehabilitation Psychology, 55(1), 1–11.

Donahue, R. G. (2007). Predictors of psychological distress in young adult motor vehicle accident survivors: Gender differences and similarities. DissertationAbstracts International: Section B: The Sciences and, Engineering, 67(9-B), 5396.

Ehlers, A., Hofmann, S. G., Herda, C. A., & Roth, W. T. (1994). Clinical characteristics of driving phobia. Journal of Anxiety Disorders, 8(4), 323–339.Ehlers, A., Mayou, R. A., & Bryant, B. (1998). Psychological predictors of chronic posttraumatic stress disorder after motor vehicle accidents. Journal of

Abnormal Psychology, 107(3), 508–519.Ehring, T., Ehlers, A., & Glucksman, E. (2008). Do cognitive models help in predicting the severity of posttraumatic stress disorder, phobia, and depression

after motor vehicle accidents? A prospective longitudinal study. Journal of Consulting and Clinical Psychology, 76(2), 219–230.First, M. B., Gibbon, M., Spitzer, R. L., Williams, J. B., & Benjamin, L. S. (1997). Structured clinical interview for DSM-IV axis II personality disorders (SCID-II).

Washington, DC: American Psychiatric Publishing.First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. (1997). Structured clinical interview for DSM-IV axis I disorders (SCID-I), clinician version (administration

booklet). Washington, DC: American Psychiatric Publishing.Frans, Ö., Rimmö, P.-A., Åberg, L., & Fredrikson, M. (2005). Trauma exposure and post-traumatic stress disorder in the general population. Acta Psychiatrica

Scandinavica, 111(4), 291–299.Frommberger, U. H., Stieglitz, R. D., Nyberg, E., Schlickewei, W., Kuner, E., & Berger, M. (1998). Prediction of posttraumatic stress disorder by immediate

reactions to trauma: A prospective study in road traffic accident victims. European Archives of Psychiatry and Clinical Neuroscience, 248(6), 316–321.Fullerton, C. S., Ursano, R. J., Epstein, R. S., Crowley, B., Vance, K., Kao, T. C., et al (2001). Gender differences in Posttraumatic Stress Disorder after motor

vehicle accidents. American Journal of Psychiatry, 158(9), 1486–1491.Furukawa, T. A., Kessler, R. C., Slade, T., & Andrews, G. (2003). The performance of the K6 and K10 screening scales for psychological distress in the Australian

National Survey of Mental Health and Well-being. Psychological Medicine, 33(2), 357–362.Gabert-Quillen, C., Irish, L. A., Sledjeski, E., Fallon, W., Spoonseter, E., & Delahanty, D. (2012). The impact of social support on the relationship between

trauma history and posttraumatic stress disorder symptoms in motor vehicle accident victims. International Journal of Stress Management, 19(1), 69–79.Gadalla, T. (2009). Determinants, correlates and mediators of psychological distress: A longitudinal study. Social Science Medicine, 68(12), 2199–2205.Goldberg, L., & Gara, M. A. (1990). A typology of psychiatric reactions to motor vehicle accidents. Psychopathology, 23(1), 15–20.Harvey, A. G., & Bryant, R. A. (1999). Predictors of acute stress following motor vehicle accidents. Journal of Traumatic Stress, 12(3), 519–628.Haswell, M. R., Kavanagh, D., Tsey, K., Reilly, L., Cadet-James, Y., Laliberte, A., et al (2010). Psychometric validation of the growth and empowerment measure

(GEM) applied with Indigenous Australians. Australian and New Zealand Journal of Psychiatry, 44(9), 791–799.Holt-Lunstad, J., Birmingham, W., & Jones, B. Q. (2008). Is there something unique about marriage? The relative impact of marital status, relationship

quality, and network social support on ambulatory blood pressure and mental health. Annals of Behavioral Medicine, 35(2), 239–244.Hox, J. (1998). Multilevel modeling: When and why? In I. Balderjahn, R. Mathar, & M. Schader (Eds.), Classification, data analysis, and data highways

(pp. 147–154). New York, NY: Springer-Verlag.Kessler, R. C., Andrews, G., Colpe, L. J., Hiripi, E., Mroczek, D. K., Normand, S.-L. T., et al (2002). Short screening scales to monitor population prevalences and

trends in non-specific psychological distress. Psychological Medicine, 32(6), 959–976.King, D. W., Vogt, D. S., & King, L. A. (2004). Risk and resilience factors in the etiology of chronic post traumatic stress disorder. In B. T. Litz (Ed.), Early

intervention for trauma and traumatic loss (pp. 34–64). New York, NY: Guildford Press.Kupchik, M., Strous, R. D., Erez, R., Gonen, N., Weizman, A., & Spivak, B. (2007). Demographic and clinical characteristics of motor vehicle accident victims in

the community general health outpatient clinic: A comparison of PTSD and non-PTSD subjects. Depression and Anxiety, 24(4), 244–250.Levenson, R. W., Sher, K. J., Grossman, L. M., Newman, J., & Newlin, D. B. (1980). Alcohol and stress response dampening: Pharmacological effects,

expectancy, and tension reduction. Journal of Abnormal Psychology, 89(4), 528–538.Mayou, R., Bryant, B., & Duthie, R. (1993). Psychiatric consequences of road traffic accidents. British Medical Journal, 307(6905), 647–651.Mayou, R., Bryant, B., & Ehlers, A. (2001). Prediction of psychological outcomes one year after a motor vehicle accident. American Journal of Psychiatry, 158(8),

1231–1238.Mayou, R., Tyndel, S., & Bryant, B. (1997). Long-term outcome of motor vehicle accident injury. Psychosomatic Medicine, 59(6), 578–584.McCrady, B. S. (1993). Alcoholism. In D. H. Barlow (Ed.), Clinical handbook of psychological disorders (2nd ed., pp. 362–395). New York, NY: Guildford Press.Mueller, T. I., Lavori, P. W., Keller, M. B., Swartz, A., Warshaw, M., Hasin, D., et al (1994). Prognostic effect of the variable course of alcoholism on the 10-year

course of depression. American Journal of Psychiatry, 151(5), 701–706.Norris, F. H. (1992). Epidemiology of trauma: Frequency and impact of different potentially traumatic events on different demographic groups. Journal of

Consulting and Clinical Psychology, 60(3), 409–418.Patten, S. B., Wang, J. L., Williams, J. V. A., Currie, S., Beck, C. A., Maxwell, C. J., et al (2006). Descriptive epidemiology of major depression in Canada. Canadian

Journal of Psychiatry, 51(2), 84–90.Preacher, K. J., Curran, P. J., & Bauer, D. J. (2006). Computational tools for probing interaction effects in multiple linear regression, multilevel modeling, and

latent curve analysis. Journal of Educational and Behavioral Statistics, 31(4), 437–448.Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage.Roberts, S. E., Vingilis, E., Wilk, P., & Seeley, J. (2007). A comparison of self-reported motor vehicle collision injuries compared with official collision data:

Analysis of age and sex trends using the Canadian National Population Health Survey and Transport Canada data. Accident, Analysis and Prevention, 40(2),559–566.

Robins, L. N., Wing, J., Wittchen, H. U., Helzer, J. E., Babor, T. F., Burke, J., et al (1988). The Composite International Diagnostic Interview: An epidemiologicalinstrument suitable for use in conjunction with different diagnostic systems and in different cultures. Archives of General Psychiatry, 45(12), 1069–1077.

N. Wiesenthal, E. Vingilis / Transportation Research Part F 21 (2013) 1–13 13

Schuckit, M. A. (1994). The relationship between alcohol problems, substance abuse, and psychiatric syndromes. In T. A. Widiger, A. J. Frances, H. A. Pincus,M. B. First, R. Ross, & W. Davis (Eds.), DSM-IV sourcebook (pp. 45–66). Washington, DC: American Psychiatric Publishing.

Seery, M. D., Holman, E. A., & Silver, R. C. (2010). Whatever does not kill us: Cumulative lifetime adversity, vulnerability, and resilience. Journal of Personalityand Social Psychology, 99(6), 1025–1041.

Seethalakshmi, R., Dhavale, H. S., Gawande, S., & Dewan, M. (2006). Psychiatric morbidity following motor vehicle crashes: A pilot study from India.Psychiatric Practice, 12(6), 415–418.

Sher, K. J., Bartholow, B. D., & Peuser, K. (2007). Stress–Response-Dampening effects of alcohol: Attention as a mediator and moderator. Journal of AbnormalPsychology, 116(2), 362–377.

Statistics Canada (2009). National population health survey – Household component – Documentation for the derived variables and the constant longitudinalvariables (specifications) – Cycles 1–7 (1994/1995 to 2006/2007). <http://www.statcan.gc.ca/imdb-bmdi/document/3225_D10_T9_V3-eng.pdf>Retrieved 27.04.11.

Statistics Canada (2010). National population health survey – Household component – Longitudinal (NPHS). <http://www.statcan.gc.ca/cgi-bin/imdb/p2SV.pl?Function=getSurvey&SDDS=3225&lang=en&db=imdb&adm=8&dis=2> Retrieved 27.04.11.

Statistics Canada (2012). 1996 Census: Nation tables: Visible minorities. <http://www.statcan.gc.ca/c1996-r1996/feb17-17fev/minority-minorites1-eng.htm>Retrieved 04.04.13.

Tavris, D. R., Kuhn, E. M., & Layde, P. M. (2001). Age and gender patterns in motor vehicle crash injuries: Importance of type of crash and occupant role.Accident, Analysis and Prevention, 33(2), 167–172.

Taylor, J., Deane, F., & Podd, J. (2002). Driving-related fear: A review. Clinical Psychology Review, 22(5), 631–645.Taylor, S., & Koch, W. J. (1995). Anxiety disorders due to motor vehicle accidents: Nature and treatment. Clinical Psychology Review, 15(8), 721–738.Tolin, D. F., & Foa, E. B. (2006). Sex differences in trauma and posttraumatic stress disorder: A quantitative review of 25 years of research. Psychological

Bulletin, 132(6), 959–992.Ursano, R. J., Fullerton, C. S., Epstein, R. S., Crowley, B., Kao, T.-C., Vance, K., et al (1999). Acute and chronic posttraumatic stress disorder in motor vehicle

accident victims. American Journal of Psychiatry, 156(4), 589–595.Vingilis, E., Larkin, E., Stoduto, G., Parkinson-Heyes, A., & McLellan, B. (1996). Psychosocial sequelae of motor vehicle collisions: A follow-up study. Accident,

Analysis and Prevention, 28(5), 637–645.Vingilis, E., Wade, T. J., & Seeley, J. (2006). Predictors of adolescent health care utilization. Journal of Adolescence, 30(5), 773–800.Wang, J. L., & El-Guebaly, N. (2004). Sociodemographic factors associated with comorbid major depressive episodes and alcohol dependence in the general

population. Canadian Journal of Psychiatry, 49(1), 37–44.World Health Organization (WHO) (2009). Decade of action for road safety 2011–2020. <http://www.who.int/violence_injury_prevention/publications/

road_traffic/saving_millions_lives_en.pdf> Retrieved 06.02.12.Wu, K. K., & Cheung, M. W. L. (2006). Posttraumatic stress after a motor vehicle accident: A six-month follow-up study utilizing latent growth modeling.

Journal of Traumatic Stress, 19(6), 923–936.Zatzick, D. F., Kang, S.-M., Müller, H.-G., Rivara, F. P., Katon, W., Jurkovich, G. J., et al (2002). Predicting posttraumatic distress in hospitalized trauma

survivors with acute injuries. American Journal of Psychiatry, 159(6), 941–946.Zlotnick, C., Warshaw, M., Shea, M. T., & Keller, M. B. (1997). Trauma and chronic depression among patients with anxiety disorders. Journal of Consulting and

Clinical Psychology, 65(2), 333–336.