Accident Analysis and Prevention 202 (2024) 107574

Contents lists available at ScienceDirect

Accident Analysis and Prevention

journal homepage: [Link]/locate/aap

Trajectories of health-related quality of life following road trauma: Latent

growth mixture modeling across a 12-month cohort study
Somayeh Momenyan a, Herbert Chan a, Shannon Erdelyi a, Lulu X. Pei a, Leona K. Shum a,
Lina Jae a, John Taylor a, John A. Staples a, c, Stirling Bryan b, c, Jeffrey R. Brubacher a, *
a
Department of Emergency Medicine, University of British Columbia, Vancouver, British Columbia, Canada
b
School of Population & Public Health, University of British Columbia, Vancouver, British Columbia, Canada
c
Centre for Clinical Epidemiology & Evaluation, Vancouver, British Columbia, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: Introduction: Health-related quality of life (HRQoL) should be considered when evaluating the burden of road
Health-related quality of life trauma (RT) injuries. This study aimed to identify distinct HRQoL trajectories following minor to severe RT
Road trauma injury injury and determine characteristics of trajectory membership.
Cohort study
Methods: This prospective inception cohort study recruited 1480 RT survivors from three emergency departments
in British Columbia, Canada (July 2018 - March 2020). HRQoL outcome was measured with the Short Form 12
survey (SF-12) and the 5-level version of the EuroQol instrument (EQ-5D-5L) at baseline (pre-injury) and at 2, 4,
6, and 12 months post-injury. Potential predictors of outcome trajectory included sociodemographic, psycho­
logical, medical, crash, and injury factors collected at baseline. We used a latent growth mixture model to
identify distinct recovery trajectories and multinomial logistic regression to determine predictors of trajectory
membership.
Results: Three distinct HRQoL trajectories were identified for SF-12 subscales and EQ-5D-5L measures: Low/
Moderate-Stable, High-Large decline, and High-Slight decline. Participants in the Low/Moderate-Stable trajec­
tory had persistent low to moderate HRQoL before and after the injury. Those in the High-Large decline tra­
jectory had good pre-injury HRQoL followed by persistently decreased HRQoL afterwards. The High-Slight
decline trajectory was characterized by good pre-injury HRQoL and only a slight decline afterwards. Participants
in the Low/Moderate-Stable and High-Large decline trajectories were considered at risk of permanently poor
HRQoL following RT injury given their low HRQoL over a long period of time. Characteristics that placed
participants in the Low/Moderate-Stable trajectory were older age, female gender, poor pre-injury health
(medical comorbidity, prescribed medication use, complaints in the injured body area(s)), pre-injury somatic
symptoms, pain catastrophizing or psychological distress, injury severity (ISS) and injury pain. Patients with
head injury were less likely to be in the Low/Moderate-Stable trajectory. Risk factors for membership in the
High-Large decline trajectory included older age (for physical HRQoL), younger age (for mental HRQoL), female
gender, living alone, pre-injury psychological distress, ISS, injury pain, no expectations for a fast recovery, as
well as head injuries, spine/back injuries or lower extremity injuries.
Conclusions: This study highlighted the heterogeneity of HRQoL trajectories following RT injury and the
importance of considering differences between characteristics of survivors. In addition to injury type and
severity, outcome is related to demographic factors, pre-injury health and pre-injury psychological factors.

1. Introduction shifted from mortality to disability (Polinder et al., 2010; Rissanen et al.,
2020). Non-fatal but disabling injuries have a substantial impact on
Each year, road trauma (RT) causes 20 to 50 million injuries and 1.3 health, social, and economic outcomes (Gopinath et al., 2015; Connelly
million fatalities globally (World Health Organization, 2018). With and Supangan, 2006). Although most research has focused on severe
improvements in emergency medical care, the burden of RT injuries has injuries which can result in permanent disability, it is recognized that

* Corresponding author.
E-mail address: [Link]@[Link] (J.R. Brubacher).

[Link]
Received 27 February 2023; Received in revised form 26 March 2024; Accepted 8 April 2024
0001-4575/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license ([Link]
nc-nd/4.0/).
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

even minor RT injuries can have long-term consequences, such as Canada. The second aim was to examine a range of sociodemographic,
physical disability, psychological disorders, prolonged absenteeism psychological, medical, crash, and injury factors as potential predictors
from work, and reduced health-related quality of life (HRQoL) (Hours of trajectory membership. To accomplish these aims, we used an ana­
et al., 2013; Khati et al., 2013; Chossegros et al., 2011; Murgatroyd et al., lytic method known as a latent growth mixture model (LGMM). LGMM is a
2016). clustering method used increasingly to identify latent (unobserved)
Self-reported HRQoL is an important metric of health status that subpopulations of similar individuals based on their outcome trajectory
incorporates a comprehensive picture of an individual’s living situation across time (Muthén and Muthén, 2000; Jung and Wickrama, 2008).
(e.g., ability to function and perception of well-being in the physical and
mental domains of life). A systematic review confirmed that HRQoL 2. Methods
scores following RT declined significantly from pre-injury values (Ris­
sanen et al., 2017). Understanding predictive factors for poor HRQoL 2.1. Participants and procedures
following RT injury is important as the highest costs are incurred by
people with poor recovery (Connelly and Supangan, 2006; Littleton This prospective cohort study recruited RT survivors from three BC
et al., 2011). HRQoL trajectories following RT injury vary depending on EDs (Vancouver General Hospital, Vancouver; Royal Columbian Hos­
a range of pre-injury, injury-related, and post-injury factors. Psycho­ pital, New Westminster; and Kelowna General Hospital, Kelowna) be­
logical factors, including anxiety and depression (Littleton et al., 2011; tween July 2018 and March 2020 and followed them for 12 months post-
Gopinath et al., 2020; Wang et al., 2005; Hung et al., 2022) and post- injury. Methods have been described in detail previously (Shum et al.,
traumatic stress disorder post-injury (Wang et al., 2005; Kenardy 2021). RT survivors were eligible if they were at least 16 years old and
et al., 2017), affect the way people cope with injury-related pain and entered the ED within 24 h following a collision involving at least one
disability and therefore influence HRQoL. Pain experience at baseline, motorized vehicle. Non-BC residents and those who died within 30 days
such as injury pain (Gopinath et al., 2020) and pain catastrophizing were excluded. The present study analyzed data collected at baseline
(Gopinath et al., 2020; Hung et al., 2022), is also associated with poor and at 2, 4, 6, and 12 months post RT injury for five different road user
HRQoL at follow-up. The effects of sociodemographic factors and injury types (cyclists, pedestrians, motorcyclists, and motor vehicle drivers and
severity on HRQoL outcome are varied across studies. Some studies passengers). Information was collected using interviews and medical
found that older individuals (Gopinath et al., 2020; Doan et al., 2020) chart reviews. Baseline interviews were conducted in-person by a
and females (Doan et al., 2020; Rissanen et al., 2020; Fitzharris et al., research assistant during ED visits or hospital admissions, or by tele­
2007) had significantly poorer HRQoL. However, other studies found no phone in some cases. Interviews were completed within one week of
significant effect for age or sex (Jagnoor et al., 2015; Sharwood et al., injury except for in patients admitted to hospital with severe injuries
2021). Unsurprisingly, most studies found that individuals with higher who were not capable of being interviewed within this timeframe; over
injury severity had poorer HRQoL outcomes (Jagnoor et al., 2015; 98% of the interviews were conducted within two weeks after the injury
Barnes and Thomas, 2006), although other studies found that partici­ and all interviews were completed within 35 days. Follow-up interviews
pants with minor injuries had worse mental health outcomes (Kenardy were conducted by telephone, online survey, self-filled paper question­
et al., 2017; Kenardy et al., 2015) and one study reported no significant naire or in-person (using a translator if required) depending on partic­
association between injury severity and HRQoL outcome (Doan et al., ipant preference. For those unable to complete interviews independently
2020). Further research is required to identify risk factors for poor (e.g., cognitive disability, language barrier), a proxy either assisted the
HRQoL following RT injury. participant or completed the questionnaire on the participant’s behalf.
Much previous research on HRQoL following RT investigated highly Non-English speakers were interviewed through a translator (e.g.,
selected groups and may not apply to the general RT injury population. family) or multilingual research assistant in Cantonese, French, Korean,
Some studies included only RT survivors with musculoskeletal injury Mandarin, Punjabi, and Vietnamese (reflecting the common languages
(Littleton et al., 2011; Littleton et al., 2014), one study enrolled only spoken in Greater Vancouver).
survivors with personal-injury insurance (Gopinath et al., 2020). Other
studies excluded minor or severe injury crashes (Littleton et al., 2011; 2.2. Outcome variables
Gopinath et al., 2020; Kenardy et al., 2017; Littleton et al., 2014; Smits
et al., 2019; Fitzharris et al., 2010; Paiva et al., 2016; Lindahl et al., The continuous outcome was HRQoL score, measured with the Short
2021), some road users (Littleton et al., 2011; Doan et al., 2020; Littleton Form 12 survey (SF-12) (Ware, 1993) and the 5-level version of EuroQol
et al., 2014), or people who did not speak the dominant language instrument (EQ-5D-5L) (Cheung et al., 2009) at each time point. The
(Kenardy et al., 2017; Fitzharris et al., 2007; Jagnoor et al., 2015; baseline SF-12 asks about HRQoL 4 weeks prior to the injury; the
Gopinath et al., 2015; Littleton et al., 2014; Smits et al., 2019; Fitzharris baseline EQ-5D-5L asks about HRQoL the day before the injury.
et al., 2010). The validity of some previous studies is limited because The SF-12 is a 12-item questionnaire which measures health in eight
they enrolled RT survivors weeks after the crash when they filed in­ dimensions: general health, physical role and functioning, social func­
surance claims (Kenardy et al., 2017; Jagnoor et al., 2015; Kenardy tioning, bodily pain, mental health, vitality, and emotional role. The
et al., 2015; Smits et al., 2019; Gopinath et al., 2015; Rebbeck et al., Physical Component Summary (PCS) and Mental Component Summary
2006). Such designs could result in selection bias and recall bias as in­ (MCS) were calculated from the eight dimensions of the SF-12 and
dividuals with minor injuries are more likely to be excluded and standardized to the 1998 U.S. population (Maruish, 2012). Standardized
enrolment delays may worsen recall of pre-injury health status. Also, to PCS and MCS scores have a mean of 50 and a standard deviation of 10,
date, the majority of RT outcome studies have been conducted in with higher scores suggesting better physical and mental well-being,
Australia or Europe with relatively little North American research. This respectively. The EQ-5D-5L consists of two parts: a short descriptive
is an important limitation as some risk factors for poor HRQoL, such as system questionnaire (EQ-5D) and a visual analog scale (EQ-VAS). The
recovery expectations, are likely related to cultural factors that may vary EQ-5D measures health status along five dimensions (mobility, self-care,
from country to country. usual activities, pain/discomfort, and anxiety/depression), each with
Identifying characteristics of distinct groups at risk for poorer HRQoL five response levels (no problems, slight problems, moderate problems,
could facilitate the delivery of targeted interventions along the recovery severe problems, and unable to perform/extreme problems). The EQ-5D-
process following RT injury. Therefore, our first aim was to identify 5L summary score was calculated by combining the responses to the 5
distinct HRQoL trajectories over a 12-month period in RT survivors dimensions and standardizing based on the Canadian value set (Xie
(including all road users and all injury severity levels) who visited a et al., 2016). Standardized EQ-5D-5L summary scores range from − 0.14
participating emergency department (ED) in British Columbia (BC), to 0.94, with higher scores suggesting better health. The EQ-VAS is an

2
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

index of overall health state that ranges from 0 to 100, where 0 denotes test, entropy, posterior probabilities, classes size, and clinical inter­
the worst possible health state and 100 the best possible health state. pretability (Nylund et al., 2007). Lower values of AIC, BIC, and ABIC
indicate better model fitness. The ALMR LR test compares the LGMM of
2.3. Predictor variables the k-class vs the (k + 1)-class, where a significant P-value indicates that
the (k + 1)-class model is preferred. Entropy is a criterion to assess the
Potential predictors of HRQoL outcome following RT injury con­ precision of classification ranging from 0 to 1. The entropy values of
sisted of the following information collected at baseline: (1) socio­ 0.80 and over indicate good class separation (Clark, 2010). Individuals
demographic factors; (2) psychological and medical factors; (3) crash are assigned to the class based on their most likely posterior probabili­
and injury factors. Sociodemographic factors included age, gender, ties. An average posterior probability is recommended to be more than
employment status, living situation, education level, ethnicity, years 0.7 (Nagin and Nagin, 2005). Finally, once individuals are assigned to
lived in Canada, and alcohol and/or cannabis use (in the six hours prior classes, it is important to check the number of individuals in each class as
to the injury). Psychological and medical factors included somatic well as the interpretability of classes. We rejected models that were
symptoms, pain catastrophizing, psychological distress, recovery ex­ clinically uninterpretable or that included classes with <1% of total
pectations, number of pre-injury comorbidities, pre-injury complaints in sample size.
the injured body area(s), and use of any prescription medication use. All LGMMs were obtained using 1000 random sets of starting values,
Crash and injury factors included time of ED visit, road user type, Injury 250 final optimizations, and 20 iterations. To confirm that model esti­
Severity Score (ISS) (Osler et al., 1996; Barnard et al., 2013); injury pain, mations converge on the global solution instead of local maxima, we
and injury location. checked that the best log-likelihood value was replicated (Muthén and
Somatic symptoms were evaluated with the Patient Health Muthen, 2017). We also ran models with two different seeds from the
Questionnaire-15 (PHQ-15) which ranges from 0 to 30; higher scores best log-likelihood value to ensure model estimations were identical for
indicate greater symptom severity (Kocalevent et al., 2013). Pain cata­ different seeds. Participants with at least 2 measurements over time
strophizing (2 weeks prior to the injury) was measured with the Pain were included in the LGMM for each outcome as trajectory analyses
Catastrophizing Scale, a validated 13-item 5-point Likert scale with were more stable and precise. The flow diagram of enrolled participants
higher scores indicate higher degrees of catastrophic thinking styles as well as exclusions for LGMM is presented in Fig. 1. We considered
(range 0–52) (Sullivan et al., 1995). The Patient Health Questionnaire-4 predictors for the multivariable analysis if they were univariately sig­
(PHQ-4) was used to assess psychological distress; higher scores suggest nificant at 0.2 level. The final model was built using a backward elim­
more severe depression and/or anxiety (range 0–12) (Löwe et al., 2010; ination algorithm by keeping predictors that were significant at 0.05
Kroenke et al., 2009). PHQ-15 and PHQ-4 ask about somatic symptoms level (based on the likelihood ratio test). As the nature of this analysis is
and psychological distress 4 and 2 weeks before the injury, respectively. exploratory, P-values were not adjusted for multiple comparisons.
Pre-injury comorbidities included eye disease, arthritis, diabetes, res­
piratory disease, heart disease, hypertension, stroke, epilepsy, kidney 2.5. Missing data
disease, psychiatric disease, and other diseases that participants re­
ported to the interviewers. Post-injury pain reported during the baseline LGMM analysis handles missing data on outcome variables using a
interview was measured on a visual analog scale (VAS) ranging from robust full-information maximum-likelihood (FIML) estimation
0 (no pain) to 10 (worst pain imaginable) (Todd et al., 1996). approach under the assumption of a missing at random mechanism. The
assumption of missing at random was shown to be reasonable in missing
2.4. Statistical analysis data on outcomes (Supplementary Table S8). Missing values for baseline
covariates were imputed using the Markov chain Monte Carlo (MCMC)
Data cleaning and statistical analyses were performed using R method. The assumption of missing at random was shown to be
version 4.0.5 and Mplus version 7.4. Descriptive statistics (mean and reasonable in missing data on covariates (Supplementary Table S9). A
standard deviation (SD) or number and percent) were employed to sensitivity analysis was performed by comparing analysis of data with
describe the participants at baseline and their PCS, MCS, EQ-5D-5L (Tables 3 and 4) and without imputations (Supplementary Tables S10
summary score, and EQ-VAS outcomes across time. We used three and S11).
steps to analyze data: first we plotted trajectories of PCS, MCS, EQ-5D-5L
summary score, and EQ-VAS across participants to describe patterns 3. Results
during a 12-month follow-up (Supplementary Figs. S1–S4). Second, an
LGMM was applied to identify distinct trajectory classes, separately for A total of 1480 participants were enrolled. The mean ± SD age was
each outcome (see below). Third, predictors of membership in each class 43.1 ± 18.2 years and 54.1% were males. Other baseline characteristics
were identified using multinomial logistic regression. The results of the of participants are summarized in Table 1. Descriptive statistics of PCS,
multinomial logistic regression were presented as adjusted odds ratios MCS, EQ-5D-5L summary score, and EQ-VAS outcomes across time are
(ORs) with 95% confidence intervals. presented in Table 2. The differences in baseline characteristics of par­
To identify trajectory classes using LGMM, a latent categorical var­ ticipants included in the LGMM and those not-included are presented in
iable (class) is regressed on latent growth parameters (such as intercept Supplementary Table S7.
and slope) that incorporate information from repeated measures of an
outcome over time. The LOESS (locally estimated scatterplot smoothing) 3.1. Patterns of HRQoL across time
curves of all outcomes decreased relative to baseline, rapidly increased
in the first 4-month period post-injury, and then slowed down thereafter Supplementary Tables S1–S4 show the fit criteria of variant models
(Supplementary Figs. S1–S4). Polynomials of any order cannot well with free time scores, varying from two to five classes for the PCS, MCS,
approximate this nonlinear growth (Fitzmaurice et al., 2012). In this EQ-5D-5L summary score, and EQ-VAS outcomes, respectively. Ac­
study, a model with free time scores was specified to account for the cording to fit criteria, the optimal model was identified for each
nonlinear trend in outcome values (see Supplementary Section 4) (Wang outcome, and its classes were titled based on starting values at baseline
and Wang, 2019). To identify the optimal class number, a sequence of (Low, Moderate, or High) and the overall shape of the 12-month tra­
LGMMs, ranging from 2 to 5 classes were compared according to a series jectory (Stable, Large decline, or Slight decline).
of criteria, including the Akaike information criterion (AIC), Bayesian The three-class model was most appropriate for PCS trajectories
information criterion (BIC), sample size-adjusted Bayesian information (Fig. 2). The Low-Stable class indicates a group of participants (n = 128,
criterion (ABIC), adjusted Lo-Mendell-Rubin likelihood ratio (ALMR LR) 10.9%) who had a constantly low PCS across the 12-month follow-up.

3
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

Fig. 1. Recruitment flow diagram of the current study.

The High-Large decline class represents a group of participants (n = 577, participants with no expectations for a fast recovery, higher ISS and
48.9%) with high baseline PCS scores that had decreased dramatically at injury pain scores, and spine/back or lower extremity injuries had
two-month follow-up and then increased slightly but remained low. The increased probability of being in the High-Large decline class. For
High-Slight decline class (n = 481, 40.2%) indicates a PCS trajectory instance, with 1 score increase in injury pain, the odds of being in the
that decreased slightly from a high level during the first two months of High-Large decline class rather than being in the High-Slight decline PCS
follow-up and then increased rapidly. Similarly, three trajectory pat­ class is estimated to increase by 27% (OR = 1.27 per 1 score; 95% CI =
terns for MCS scores were identified: Low-Stable (n = 97, 8.5%), High- 1.18–1.37), while this chance for patients with spine/back injury rela­
Large decline (n = 417, 35.2%) and High-Slight decline (n = 672, tive to others is estimated to increase by 80% (OR = 1.80; 95% CI =
56.3%). (Fig. 3). For EQ-5D-5L summary scores three classes were 1.30–2.49). In comparison with the High-Slight decline EQ-VAS class,
identified: Moderate-Stable (n = 39, 3.2%), High-Large decline (n = 222, the High-Large decline class was characterized by participants who had
18.4%) and High-Slight decline (n = 944, 78.3%) (Fig. 4). The 3-class no expectations for a fast recovery and higher injury pain scores. In
model was also optimal for EQ-VAS trajectories (Fig. 5). The comparison with the High-Large decline EQ-VAS class, the Moderate-
Moderate-Stable class for EQ-VAS trajectories included 60 participants Stable class was characterized by participants with higher somatic
(5.0%), the High-Large decline class had 222 participants (18.5%) and symptoms and pain catastrophizing scores, more comorbidities, and
the High-Slight decline class had 919 participants (76.5%). who used prescribed medication pre-injury. Comparisons between other
EQ-5D-5L trajectory classes are shown in Table 4.
3.2. Predictors of HRQoL trajectories
4. Discussion
Baseline characteristics of participants by the SF-12 subscales and
This study identified 3 heterogeneous classes of Low/Moderate-
EQ-5D-5L measures trajectory classes are summarized in Supplementary
Stable, High-Large decline, and High-Slight decline for PCS, MCS, EQ-
Tables S5 and S6, respectively. Predictors which differentiate between
5D-5L summary score, and EQ-VAS outcomes in RT survivors. The
the SF-12 subscales and EQ-5D-5L measures trajectory classes were
Low/Moderate-Stable (3.2–10.9%), High-Large decline (18.4–48.9%),
determined using multinomial logistic regression (Tables 3 and 4). For
and High-Slight decline (40.2–78.3%) classes represent groups of par­
each outcome, all pairwise comparisons between classes were per­
ticipants with HRQoL that was i) low-moderate pre-injury and relatively
formed as these comparisons were clinically relevant.
unaffected by the injury itself, ii) decreased substantially after the
injury, and iii) decreased slightly after the injury, respectively. The Low/
3.2.1. SF-12 trajectories
Moderate-Stable class included patients with minor injuries whose
Compared to the High-Slight decline PCS class, being older, being
HRQoL did not change following the collision. In this group, poor to
female, having higher ISS and injury pain scores, and having spine/back
moderate post-injury HRQoL may simply be a continuation of their pre-
and lower extremity injuries were associated with an increased proba­
injury HRQoL. In both the High-Slight decline and High-Large decline
bility of being in the High-Large decline class. For instance, with 10
classes, the loss of HRQoL scores was largest during the first 2 months of
years increase in age, the odds of being in the High-Large decline class
follow-up and HRQoL scores did not return to pre-injury levels at 12-
rather than being in the High-Slight decline PCS class is estimated to
month follow-up. This pattern is consistent with previous research
increase by 18% (OR = 1.18 per 10-year; 95% CI = 1.10–1.27) while this
showing that the decline in HRQoL is greatest shortly after the injury
chance for females relative to males is estimated to increase by 62% (OR
and many patients do not attain their pre-injury HRQoL at last follow-up
= 1.62; 95% CI = 1.23–2.13). Compared with the participants in the
(Littleton et al., 2011; Gopinath et al., 2020; Doan et al., 2020; Fitzharris
High-Slight decline MCS class, participants in the High-Large decline
et al., 2007; Littleton et al., 2014; Fitzmaurice et al., 2012). We also
class were more likely to be younger, female, live alone, have higher
found that the reduction in PCS scores post-injury was more than that of
psychological distress scores, have no expectations for a fast recovery,
MCS scores in both the High-Slight decline and High-Large decline
have higher injury pain scores, and have a spine/back injury. Compar­
classes. This finding indicates that, as might be expected, the injury had
isons between the other PCS-trajectory and MCS-trajectory classes are
a greater impact on physical health than mental health, which is
shown in Table 3.
consistent with previous research from Australia and the UK (Gopinath
et al., 2020; Fitzharris et al., 2007; Sharwood et al., 2021; Barnes and
3.2.2. EQ-5D-5L trajectories
Thomas, 2006; Fitzmaurice et al., 2012).
Compared to the High-Slight decline EQ-5D-5L summary score class,

4
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

Table 1 Table 2
Baseline characteristics of participants (n = 1480). Descriptive statistics of outcomes under study (n = 1480).
Characteristic Number Mean (SD) or Number Outcomes Number Mean (SD)
(percent) median [Min-Max] (percent)
Physical Component Summary (PCS)
missing
Baseline 1459 52.7 (7.2)
Sociodemographic factors Month 2 1012 36.8 (11.3)
Age in years 43.1 (18.2) 0 (0.0) Month 4 898 40.1 (11.4)
Gender (Male) 800 (54.1) 0 (0.0) Month 6 883 42.4 (11.2)
Employment status 48 (3.2) Month 12 826 43.4 (11.5)
Employed 981 (66.3)
School 154 (10.4)
Mental Component Summary (MCS)
Retired 200 (13.5)
Baseline 1459 53.9 (8.3)
Others 97 (6.6)
Month 2 1012 43.1 (11.9)
Living situation 23 (1.6)
Month 4 898 44.8 (11.5)
Alone 351 (23.7)
Month 6 883 45.3 (11.3)
Others 1106 (74.7)
Month 12 826 45.9 (11.5)
Education level 17 (1.1)
Less than high school 96 (6.5)
High school and 569 (38.4) EQ-5D-5L summary score
vocational Baseline 1473 0.92 (0.07)
University 798 (53.9) Month 2 1044 0.69 (0.23)
Ethnicity 20 (1.4) Month 4 929 0.73 (0.21)
Caucasian 739 (49.9) Month 6 908 0.76 (0.20)
Asian 364 (24.6) Month 12 855 0.77 (0.19)
Others 357 (24.1)
Years lived in Canada 19 (1.3)
>10 years 1252 (84.6) EQ-VAS
<10 years 209 (14.1) Baseline 1470 86.8 (13.6)
Pre-existing alcohol use 740 (50.0) 323 (21.8) Month 2 1037 64.7 (21.2)
(Yes) Month 4 915 67.6 (20.6)
Pre-existing cannabis use 297 (20.1) 324 (21.9) Month 6 892 70.3 (19.6)
(Yes) Month 12 851 71.1 (19.3)

SD: standard deviation; EQ-5D-5L: European Quality of Life-5 Dimensions; EQ-

Psychological and medical factors VAS: EQ-5D-5L visual analog scale.
Somatic symptoms (PHQ- 2.1 [0.0–25.0] 47 (3.2)
15)
Pain catastrophizing 4.0 [0.0–52.0] 45 (3.0)
Kenardy et al. described HRQoL trajectories between 6 and 24
Psychological distress 0.0 [0.0–12.0] 5 (0.3) months in Australian auto-insurance claimants with non-catastrophic
(PHQ-4) RT injury (Kenardy et al., 2017). They described three recovery trajec­
Recovery expectations 324 (21.9) tories for PCS (“severe and chronic”, “low but improving”, and “gradual
Less than 1 month 395 (26.7)
recovery”) and four trajectories for MCS (“unaffected”, “severe but
More than 1 month 263 (17.8)
No idea 498 (33.6) improving”, “severe and declining”, and “severe and chronic”). Our re­
Pre-injury comorbidities 1.0 [0.0–9.0] 0 (0.0) sults cannot be compared directly with Kenardy’s findings (Kenardy
number et al., 2017), because their study did not collect pre-injury HRQoL and
Pre-injury body 319 (21.6) 14 (0.9) their follow-up was at 6, 12 and 24 months post-injury. That said, the
complaints (Yes)
Pre-injury prescribed 615 (41.6) 0 (0.0)
PCS trajectories suggested by Kenardy had similar shapes to what we
medication use (Yes) found in the present study in a common follow-up period (i.e. between 6
and 12 months). Our “Low-Stable”, “High-Large decline”, and “High-
Crash and injury factors
Slight decline” trajectories correspond to their “severe and chronic”,
Time of ED visit 0 (0.0) “low but improving”, and “gradual recovery” trajectories, respectively.
Daytime 966 (65.3) Kenardy’s four MCS trajectories were partially found in the present
Nighttime 514 (34.7) study. Our “High-Slight decline” corresponds to their “unaffected” tra­
Road user type 0 (0.0)
jectory and our “Low-Stable” and “High-Large decline” trajectories
Driver 683 (46.1)
Passenger 225 (15.2) correspond to their “low but improving” trajectory in a common follow-
Motorcyclist 118 (8.0) up period. The differences in the number of MCS trajectories could be
Pedestrian 280 (18.9) related to sample differences and/or differences in follow-up period.
Cyclist 174 (11.8)
The second aim of the current study was to identify the character­
ISS 3.0 [0.0–66.0] 1 (0.1)
1–3 828 (56.0)
istics of those participants at risk of poor HRQoL following RT injury.
4–8 241 (16.3) These are participants in the “Low/Moderate-Stable” and “High-Large
9–11 151 (10.2) decline” trajectories which included 22–60% of total participants.
≥12 259 (17.5) Characteristics that placed participants in the “Low/Moderate-Stable”
Injury pain (VAS) 5.3 (2.4) 330 (22.3)
trajectory were older age, female gender, medical comorbidities, pre-
Injury location
Head (Yes) 567 (38.3) 0 (0.0) injury prescribed medication use, pre-injury complaints in the injured
Neck (Yes) 558 (37.7) 0 (0.0) body area(s), more somatic symptoms, pain catastrophizing, psycho­
Torso (Yes) 567 (38.3) 0 (0.0) logical distress, higher ISS, injury pain, and injury other than head
Spine/back (Yes) 501 (33.9) 0 (0.0)
injury. Characteristics that placed participants in the “High-Large
Upper extremity (Yes) 744 (50.3) 0 (0.0)
Lower extremity (Yes) 705 (47.6) 0 (0.0)
decline” trajectory included older age (for physical HRQoL), younger
age (for mental HRQoL), female gender, more psychological distress,
SD: standard deviation; PHQ-15: Patient Health Questionnaire-15; PHQ-4: Pa­ living alone, no expectations for a fast recovery, greater ISS, greater
tient Health Questionnaire-4; ED: emergency department; ISS: injury severity
injury pain, and injuries to head, spine/back or lower extremity.
score; VAS: visual analog scale.
The results of present study were consistent with other studies that

5
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

Fig. 2. 3-class model for Physical Component Summary (PCS) (n = 1186).

Fig. 3. 3-class model for Mental Component Summary (MCS) (n = 1186).

Fig. 4. 3-class model for EQ-5D-5L summary score (n = 1205).

also reported lower HRQoL scores for females and lower scores of individuals have greater life satisfaction and resilience to common
physical HRQoL for older patients (Littleton et al., 2011; Gopinath et al., physical and social stresses (Thomas et al., 2016). Although it makes
2020; Doan et al., 2020). It is not surprising that older patients had sense that pre-existing symptoms in the injured body area could lower
poorer physical HRQoL in our study, as even minor injuries in older HRQoL, we found no previous study that assessed this association. Pain
individuals could result in a significantly reduced functioning (Gopinath catastrophizing is hypothesized to be associated with poor recovery
et al., 2015). Despite this, older patients in our study had better mental following injury as it undermines behavioral and medical treatments
health outcomes than younger patients. This may be because older (Wertli et al., 2014; Vissers et al., 2012). Consistent with previous

6
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

Fig. 5. 3-class model for EQ-VAS (n = 1201).

research (Gopinath et al., 2020; Samoborec et al., 2018), our results because there were differences in baseline characteristics of participants
support this hypothesis. Previous research has also found that increased included in the analysis of LGMM and those non-included. Also, older
levels of anxiety and depression were associated with poor HRQoL individuals and some individuals with “too much pain” may have been
scores at follow-up, which concurs with our results (Sharwood et al., less likely to participate. Moreover, we may have missed survivors who
2021; Littleton et al., 2011; Gopinath et al., 2020; Wang et al., 2005). never sought hospital treatment and survivors who were discharged
Our study emphasized the importance of pre-injury medical factors rapidly from the emergency department. Fourth, this study could be
including comorbidities, pre-injury complaints in the injured body improved by including a group of non-injured participants to compare
areas, and prescribed medication use in experiencing poor HRQoL. HRQoL trajectories. Fifth, our first post-injury assessment of HRQoL was
These results are in agreement with the well-established detrimental at two months post-injury. It is possible that some participants, espe­
effects of poor health and chronic pain on HRQoL (Tüzün, 2007). Similar cially those with minor injuries, may have had a marked decrease in
to previous research, our findings also confirmed that ISS was inversely HRQoL immediately after the injury but went on to have significant
associated with HRQoL (Khati et al., 2013; Rissanen et al., 2017; Fitz­ improvement by two months. Our study would not be able to detect this
harris et al., 2007). Finally, patients who had suffered head, spine/back, trajectory. Finally, a study with a larger sample size for each road user
and lower extremity injuries were more susceptible to poor HRQoL in type is needed to examine whether the number and type of latent HRQoL
our study, which is in line with other cohort studies (Khati et al., 2013; trajectories following RT injury differ according to road user type.
Gopinath et al., 2020; Fitzharris et al., 2010).
These findings are promising, they indicate that RT survivors at risk 5. Conclusions
of a poor HRQoL outcome might be identified at the time of hospital
admission based on these factors, some of which may be modifiable This prospective study highlighted the heterogeneity of HRQoL tra­
through targeted treatment. However, it should be mentioned that the jectories following RT injury and the importance of considering differ­
predictive performance of these factors on HRQoL outcome was not ences between characteristics of survivors. Based on the distinct HRQoL
assessed in this study. Our findings reinforce the fact that RT survivors trajectories identified in this study, a particular participant group was
with poor pre-injury health are at risk of a poor outcome and warrant found to have long-term loss of HRQoL following RT injury. Membership
special rehabilitation measures. However, it is also important to recog­ in this trajectory was more likely among participants with older age (for
nize that younger patients and those who live alone may benefit from physical HRQoL), younger age (for mental HRQoL), female gender,
psychological therapy to prevent poor mental health outcomes living alone, pre-injury psychological distress, ISS, injury pain, no ex­
following RT. Patients with pre-injury psychological distress, pain cat­ pectations for a fast recovery, as well as head injuries, spine/back in­
astrophizing, or multiple somatic complaints may also benefit from post- juries or lower extremity injuries. Moreover, another trajectory of
injury psychological therapy. The association between traumatic brain change in HRQoL scores was identified, which represents participants
injury and poor outcome is well-recognized but certain other injury who had poor HRQoL pre- and post-injury. Membership of this trajec­
types, including lower extremity injuries and spine/back injuries are tory was more likely among participants with older age, female gender,
also associated with poor outcomes and may warrant special rehabili­ poor pre-injury health (medical comorbidity, prescribed medication use,
tation measures. complaints in the injured body area(s)), pre-injury somatic symptoms,
The greatest strengths of this study are its prospective design over pain catastrophizing or psychological distress. Participants with higher
five waves, spanning 12 months, and its use of validated scales to collect ISS and greater injury pain were more likely to be in this trajectory, and
HRQoL measures and potential predictors. In addition, to maximize the participants with head injury were less likely to be in this trajectory.
generalizability of results, we did not place restrictions on road user These factors may help clinicians identify patients at risk of long-term
type, injury severity level, or language. The current study also has some impaired HRQoL for personalizing care plans. Moreover, it is possible
limitations that should be mentioned. First, although self-reported to modify many of these factors through early interventions to improve
questionnaires are a common and convenient method in research, they patients’ HRQoL following RT injury.
do come with some drawbacks such as misinterpretation of questions.
Also, there is the potential recall bias regarding self-report of pre-injury Financial Disclosure
HRQoL, PHQ-15, and PHQ-4 despite efforts to minimize this by con­
ducting baseline interviews within 7 days following the injury in most This study was funded by a research grant from the Canadian In­
cases. Second, conducting some interviews through a translator may stitutes of Health Research.
affect the validity of data because of miscommunication between the
translator and participant. Third, selection bias may have occurred

7
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

**
Table 3 P < 0.01.
***
Results of multivariable multinomial logistic regression model for assessing the P < 0.001.
a
concurrent effect of predictors on trajectory membership for SF-12 subscales Likelihood ratio test (final model versus model without the predictor under
(imputed data). consideration).
Predictor Low-Stable High-Large Low-Stable
vs. High- decline vs. vs. High-
Slight High-Slight Large Table 4
decline decline decline Results of multivariable multinomial logistic regression model for assessing the
OR (95% CI) OR (95% CI) OR (95% CI) P- concurrent effect of predictors on trajectory membership for EQ-5D-5L measures
valuea (imputed data).
Physical Component Summary (PCS) (n ¼ 1186) Predictor Moderate- High-Large Moderate-
Age (each 10-year 1.50 (1.29, 1.18 (1.10, 1.28 (1.10, <0.001 Stable vs. decline vs. Stable vs.
increase) 1.75)*** 1.27)*** 1.49)** High-Slight High-Slight High-Large
Gender (Ref: Male) decline decline decline
Female 1.33 (0.80, 1.62 (1.23, 0.82 (0.50, 0.002 OR (95% CI) OR (95% OR (95% CI) P-
2.19) 2.13)** 1.32) CI) valuea
Somatic symptoms 1.36 (1.26, 0.98 (0.93, 1.38 (1.28, <0.001
(PHQ-15) (each 1.46)*** 1.03) 1.48)*** EQ-5D-5L summary score (n ¼ 1205)
score increase) Somatic symptoms 1.35 (1.24, 1.04 (0.99, 1.29 (1.17, <0.001
Pain catastrophizing 1.03 (1.01, 1.01 (0.99, 1.02 (0.99, 0.046 (PHQ-15) (each 1.47)*** 1.10) 1.41)***
(each score 1.06)* 1.03) 1.04) score increase)
increase) Psychological 1.25 (1.11, 1.03 (0.95, 1.20 (1.05, 0.001
Pre-injury 1.36 (1.12, 1.14 (0.99, 1.18 (1.01, 0.005 distress (PHQ-4) 1.40)*** 1.12) 1.37)**
comorbidities 1.64)** 1.32) 1.39)* (each score
increase)
number
Pre-injury body complaints (Ref: No) Recovery expectations (Ref: Less than 1 month)
Yes 3.21 (1.89, 1.10 (0.78, 2.91 (1.77, <0.001 More than 1 month 0.95 (0.31, 2.74 (1.66, 0.34 (0.10, <0.001
5.44)** 1.55) 4.77)*** 2.92) 4.51)*** 1.13)
ISS (each score 1.03 (1.01, 1.05 (1.03, 0.98 (0.95, <0.001 No idea 1.03 (0.42, 2.51 (1.60, 0.41 (0.15,
increase) 1.06) * 1.07)*** 1.01) 2.52) 3.92)*** 1.08)
Injury pain (VAS) 1.26 (1.13, 1.25 (1.18, 1.01 (0.91, <0.001 ISS (each score 1.01 (0.95, 1.03 (1.01, 0.97 (0.92, <0.001
increase) 1.06) 1.05)*** 1.02)
(each score 1.40)*** 1.33)*** 1.11)
increase) Injury pain (VAS) 0.97 (0.82, 1.27 (1.18, 0.76 (0.64, <0.001
Head (Ref: No) (each score 1.16) 1.37)*** 0.91)**
Yes 0.39 (0.23, 0.92 (0.70, 0.43 (0.25, 0.002 increase)
0.69)** 1.21) 0.73)** Spine/back (Ref: No)
Spine/back (Ref: No) Yes 1.15 (0.51, 1.80 (1.30, 0.63 (0.27, 0.002
Yes 0.96 (0.56, 1.48 (1.10, 0.65 (0.39, 0.015 2.54) 2.49)*** 1.45)
1.65) 1.98) ** 1.08) Lower extremity (Ref: No)
Yes 0.86 (0.39, 1.50 (1.08, 0.57 (0.25, 0.042
Lower extremity (Ref:
No) 1.91) 2.07)* 1.32)
Yes 1.26 (0.76, 1.63 (1.24, 0.77 (0.48, 0.002
2.09) 2.15)*** 1.25) EQ-VAS (n ¼ 1201)
Somatic symptoms 1.21 (1.13, 1.05 (0.99, 1.15 (1.07, <0.001
Mental Component Summary (MCS) (n ¼ 1186) (PHQ-15) (each 1.30)*** 1.10) 1.23)***
Age (each 10-year 0.86 (0.74, 0.86 (0.81, 0.99 (0.85, 0.001 score increase)
increase) 1.01) 0.93)*** 1.15) Pain catastrophizing 1.05 (1.02, 1.01 (0.99, 1.03 (1.01, <0.001
Gender (Ref: Male) (each score 1.08)*** 1.03) 1.06)*
increase)
Female 2.37 (1.34, 1.40 (1.07, 1.68 (0.96, 0.002
4.20)** 1.81)* 3.01) Recovery expectations (Ref: Less than 1 month)
Living situation (Ref: Others) More than 1 month 1.54 (0.61, 2.91 (1.78, 0.53 (0.19, <0.001
Alone 1.68 (0.89, 1.56 (1.14, 1.08 (0.57, 0.013 3.90) 4.75)*** 1.44)
3.18) 2.12)** 2.02) No idea 1.98 (0.93, 3.51 (2.29, 0.66 (0.28,
Somatic symptoms 1.19 (1.10, 1.04 (0.99, 1.14 (1.06, <0.001 4.19) 5.39)*** 1.55)
(PHQ-15) (each 1.28)*** 1.09) 1.22)*** Pre-injury 1.27 (1.05, 0.98 (0.85, 1.29 (1.05, 0.033
score increase) comorbidities 1.54)* 1.13) 1.59)*
number
Pain catastrophizing 1.03 (1.01, 1.01 (0.99, 1.02 (0.99, 0.027
(each score 1.06)** 1.03) 1.05) Pre-injury prescribed medication use (Ref: No)
increase) Yes 2.71 (1.29, 1.05 (0.74, 2.58 (1.19, 0.022
Psychological distress 1.68 (1.50, 1.12 (1.02, 1.49 (1.34, <0.001 5.65)** 1.48) 5.56)*
(PHQ-4) (each score 1.89)*** 1.23)* 1.66)*** Injury pain (VAS) 1.15 (1.01, 1.14 (1.06, 1.01 (0.87, <0.001
increase) (each score 1.31)* 1.22)*** 1.16)
Recovery expectations (Ref: Less than 1 month) increase)
More than 1 month 0.88 (0.39, 1.50 (1.05, 0.58 (0.26, 0.006 OR: odds ratio; CI: Confidence interval; Ref: reference group; EQ-5D-5L: Euro­
1.96) 2.14)* 1.30)
pean Quality of Life-5 Dimensions; EQ-VAS: EQ-5D-5L visual analog scale; PHQ-
No idea 1.65 (0.86, 1.74 (1.27, 0.95 (0.49,
15: Patient Health Questionnaire-15; PHQ-4: Patient Health Questionnaire-4;
3.15) 2.38)*** 1.81)
Injury pain (VAS) 0.99 (0.88, 1.14 (1.07, 0.87 (0.77, <0.001 ISS: injury severity score; VAS: visual analog scale.
*
(each score 1.12) 1.20)*** 0.98)* P < 0.05.
**
increase) P < 0.01.
***
Spine/back (Ref: No) P < 0.001.
a
Yes 0.87 (0.47, 1.38 (1.05, 0.63 (0.34, 0.036 Likelihood ratio test (final model versus model without the predictor under
1.58) 1.81)* 1.14) consideration).
OR: odds ratio; CI: confidence interval; Ref: reference group; PHQ-15: Patient
Health Questionnaire-15; ISS: injury severity score; VAS: visual analog scale;
PHQ-4: Patient Health Questionnaire-4.
*
P < 0.05.

8
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

CRediT authorship contribution statement Jung, T., Wickrama, K.A., 2008. An introduction to latent class growth analysis and
growth mixture modeling. Soc. Pers. Psychol. Compass 2 (1), 302–317.
Kenardy, J., Heron-Delaney, M., Warren, J., Brown, E., 2015. The effect of mental health
Somayeh Momenyan: Writing – original draft, Formal analysis. on long-term health-related quality of life following a road traffic crash: results from
Herbert Chan: Writing – review & editing, Conceptualization. Shannon the UQ SuPPORT study. Injury 46 (5), 883–890.
Erdelyi: Formal analysis, Conceptualization. Lulu X. Pei: Writing – Kenardy, J., Heron-Delaney, M., Hendrikz, J., Warren, J., Edmed, S.L., Brown, E., 2017.
Recovery trajectories for long-term health-related quality of life following a road
review & editing, Methodology. Leona K. Shum: Project administra­ traffic crash injury: results from the UQ SuPPORT study. J. Affect. Disord. 214, 8–14.
tion, Investigation. Lina Jae: Project administration, Investigation. Khati, I., Hours, M., Charnay, P., Chossegros, L., Tardy, H., Nhac-Vu, H.-T., Boisson, D.,
John Taylor: Investigation, Conceptualization. John A. Staples: Laumon, B., 2013. Quality of life one year after a road accident: results from the
adult ESPARR cohort. J. Trauma Acute Care Surg. 74 (1), 301–311.
Investigation, Conceptualization. Stirling Bryan: Writing – review & Kocalevent, R.-D., Hinz, A., Brähler, E., 2013. Standardization of a screening instrument
editing, Conceptualization. Jeffrey R. Brubacher: Writing – review & (PHQ-15) for somatization syndromes in the general population. BMC Psychiatry 13
editing, Resources, Methodology, Investigation, Funding acquisition, (1).
Kroenke, K., Spitzer, R.L., Williams, J.B.W., Lowe, B., 2009. An ultra-brief screening scale
Conceptualization. for anxiety and depression: the PHQ-4. Psychosomatics 50 (6), 613–621.
Lindahl, M., Juneja, H., Teljigovic, S., Rafn, J., Nielsen, N.O., 2021. Change in sense of
coherence and health-related quality of life after injury–a prospective cohort study.
Declaration of competing interest Disabil. Rehabil. 43 (11), 1532–1541.
Littleton, S.M., Cameron, I.D., Poustie, S.J., Hughes, D.C., Robinson, B.J., Neeman, T.,
Smith, P.N., 2011. The association of compensation on longer term health status for
The authors declare that they have no known competing financial people with musculoskeletal injuries following road traffic crashes: emergency
interests or personal relationships that could have appeared to influence department inception cohort study. Injury 42 (9), 927–933.
the work reported in this paper. Littleton, S.M., Hughes, D.C., Gopinath, B., Robinson, B.J., Poustie, S.J., Smith, P.N.,
Cameron, I.D., 2014. The health status of people claiming compensation for
musculoskeletal injuries following road traffic crashes is not altered by an early
Data availability intervention programme: a comparative study. Injury 45 (9), 1493–1499.
Löwe, B., Wahl, I., Rose, M., Spitzer, C., Glaesmer, H., Wingenfeld, K., Schneider, A.,
Brähler, E., 2010. A 4-item measure of depression and anxiety: validation and
The data that has been used is confidential. standardization of the Patient Health Questionnaire-4 (PHQ-4) in the general
population. J. Affect. Disord. 122 (1-2), 86–95.
Maruish, M.E., 2012. User’s Manual for the SF-12v2 Health Survey, third ed.
Appendix A. Supplementary data
QualityMetric Incorporated, Lincoln, RI.
Murgatroyd, D.F., Harris, I.A., Tran, Y., Cameron, I.D., Murgatroyd, D., 2016. Predictors
Supplementary data to this article can be found online at [Link] of return to work following motor vehicle related orthopaedic trauma. BMC
org/10.1016/[Link].2024.107574. Musculoskelet. Disord. 17, 171.
Muthén, B., Muthén, L.K., 2000. Integrating person-centered and variable-centered
analyses: growth mixture modeling with latent trajectory classes. Alcohol. Clin. Exp.
References Res. 24 (6), 882–891.
Muthén, L.K., Muthen, B., 2017. Mplus User’s Guide: Statistical Analysis with Latent
Variables, User’s Guide. Muthén & Muthén.
Barnard, R.T., Loftis, K.L., Martin, R.S., Stitzel, J.D., 2013. Development of a robust
Nagin, D.S., Nagin, D., 2005. Group-Based Modeling of Development. Harvard University
mapping between AIS 2+ and ICD-9 injury codes. Accid. Anal. Prev. 52, 133–143.
Press.
Barnes, J., Thomas, P., (Eds.), 2006. Quality of life outcomes in a hospitalized sample of
Nylund, K.L., Asparouhov, T., Muthén, B.O., 2007. Deciding on the number of classes in
road users involved in crashes. In: Annual Proceedings/Association for the
latent class analysis and growth mixture modeling: a Monte Carlo simulation study.
Advancement of Automotive Medicine, Association for the Advancement of
Struct. Equ. Model. Multidiscip. J. 14 (4), 535–569.
Automotive Medicine.
Osler, T., Rutledge, R., Deis, J., Bedrick, E., 1996. ICISS: an international classification of
Cheung, K., Oemar, M., Oppe, M., Rabin, R., 2009. EQ-5D User Guide: Basic Information
disease-9 based injury severity score. J. Trauma 41 (3), 380–386.
on How to Use EQ-5D. EuroQol Group, Rotterdam.
Paiva, L., Pompeo, D.A., Ciol, M.A., Arduini, G.O., Dantas, R.A.S., Senne, E.C.V., et al.,
Chossegros, L., Hours, M., Charnay, P., Bernard, M., Fort, E., Boisson, D., Sancho, P.-O.,
2016. Health status and the return to work after traffic accidents. Rev. Bras. Enferm.
Yao, S.N., Laumon, B., 2011. Predictive factors of chronic post-traumatic stress
69, 443–450.
disorder 6 months after a road traffic accident. Accid. Anal. Prev. 43 (1), 471–477.
Polinder, S., Haagsma, J., Belt, E., Lyons, R., Erasmus, V., Lund, J., et al., 2010.
Clark, S., 2010. Mixture Modeling with Behavioral Data Doctoral Dissertation. University
A systematic review of studies measuring health-related quality of life of general
of California.
injury populations. BMC Public Health 10, 1–13.
Connelly, L.B., Supangan, R., 2006. The economic costs of road traffic crashes: Australia,
Rebbeck, T., Sindhusake, D., Cameron, I., Rubin, G., Feyer, A., Walsh, J., et al., 2006.
states and territories. Accid. Anal. Prev. 38 (6), 1087–1093.
A prospective cohort study of health outcomes following whiplash associated
Doan, H.T.N., Hobday, M.B., Leavy, J.E., Jancey, J., 2020. Health-related quality of life
disorders in an Australian population. Inj. Prev. 12 (2), 93–98.
in motorcycle crash victims one year after injury: a longitudinal study in Ho Chi
Rissanen, R., Berg, H.-Y., Hasselberg, M., 2017. Quality of life following road traffic
Minh City, Vietnam. Asia Pac. J. Public Health. 32 (2–3), 118–125.
injury: a systematic literature review. Accid. Anal. Prev. 108, 308–320.
Fitzharris, M., Fildes, B., Charlton, J., Kossmann, T., 2007. General health status and
Rissanen, R., Ifver, J., Hasselberg, M., Berg, H.Y., 2020. Quality of life following road
functional disability following injury in traffic crashes. Traffic Inj. Prev. 8 (3),
traffic injury: the impact of age and gender. Qual. Life Res. 29 (6), 1587–1596.
309–320.
Samoborec, S., Ruseckaite, R., Ayton, D., Evans, S., Puebla, I., 2018. Biopsychosocial
Fitzharris, M., Bowman, D., Ludlow, K., 2010. Factors associated with return-to-work and
factors associated with non-recovery after a minor transport-related injury: a
health outcomes among survivors of road crashes in Victoria. Aust. N. Z. J. Public
systematic review. PLoS One 13 (6), e0198352.
Health 34 (2), 153–159.
Sharwood, L.N., Kifley, A., Craig, A., Gopinath, B., Jagnoor, J., Cameron, I.D., 2021.
Fitzmaurice, G.M., Laird, N.M., Ware, J.H., 2012. Applied Longitudinal Analysis. John
Comparison of physical and psychological health outcomes for motorcyclists and
Wiley & Sons.
other road users after land transport crashes: an inception cohort study. BMC Public
Gopinath, B., Jagnoor, J., Harris, I.A., Nicholas, M., Casey, P., Blyth, F., Maher, C.G.,
Health 21 (1), 1–16.
Cameron, I.D., 2015. Prognostic indicators of social outcomes in persons who
Shum, L.K., Chan, H., Erdelyi, S., Pei, L.X., Brubacher, J.R., 2021. Predictors of poor
sustained an injury in a road traffic crash. Injury 46 (5), 909–917.
health and functional recovery following road trauma: protocol of a British
Gopinath, B., Harris, I.A., Nicholas, M., Casey, P., Blyth, F., Maher, C.G., Cameron, I.D.,
Columbian inception cohort study. BMJ Open 11 (4), e049623.
Sterling, M., 2015. A comparison of health outcomes in older versus younger adults
Smits, E., Brakenridge, C., Gane, E., Warren, J., Heron-Delaney, M., Kenardy, J.,
following a road traffic crash injury: a cohort study. PLoS One 10 (4), e0122732.
Johnston, V., 2019. Identifying risk of poor physical and mental health recovery
Gopinath, B., Jagnoor, J., Kifley, A., Dinh, M., Craig, A., Cameron, I.D., 2020. Predictors
following a road traffic crash: an industry-specific screening tool. Accid. Anal. Prev.
of health-related quality of life after non-catastrophic injury sustained in a road
132, 105280.
traffic crash. Ann. Phys. Rehabil. Med. 63 (4), 280–287.
Sullivan, M.J., Bishop, S.R., Pivik, J., 1995. The pain catastrophizing scale: development
Hours, M., Chossegros, L., Charnay, P., Tardy, H., Nhac-Vu, H.-T., Boisson, D., Luauté, J.,
and validation. Psychol. Assess. 7 (4), 524–532.
Laumon, B., 2013. Outcomes one year after a road accident: results from the ESPARR
Thomas, M.L., Kaufmann, C.N., Palmer, B.W., Depp, C.A., Martin, A.S., Glorioso, D.K.,
cohort. Accid. Anal. Prev. 50, 92–102.
Thompson, W.K., Jeste, D.V., 2016. Paradoxical trend for improvement in mental
Hung, K.K.C., Kifley, A., Brown, K., Jagnoor, J., Craig, A., Gabbe, B., Derrett, S.,
health with aging: a community-based study of 1,546 adults aged 21–100 years.
Collie, A., Dinh, M., Gopinath, B., Cameron, I.D., FISH Investigators, O.B.o.T., 2022.
J. Clin. Psychiatry 77 (08), e1019–e1025.
Psychological distress, pain and insurance claims negatively affect long-term health-
Todd, K.H., Funk, K.G., Funk, J.P., Bonacci, R., 1996. Clinical significance of reported
related quality of life after road traffic injuries. J. Rehabil. Med. 54, jrm00310.
changes in pain severity. Ann. Emerg. Med. 27 (4), 485–489.
Jagnoor, J., De Wolf, A., Nicholas, M., Maher, C.G., Casey, P., Blyth, F., et al., 2015.
Tüzün, E.H., 2007. Quality of life in chronic musculoskeletal pain. Best Pract. Res. Clin.
Restriction in functioning and quality of life is common in people 2 months after
Rheumatol. 21 (3), 567–579.
compensable motor vehicle crashes: prospective cohort study. Injury Epidemiol. 2
(1), 1–11.

9
S. Momenyan et al. Accident Analysis and Prevention 202 (2024) 107574

Vissers, M.M., Bussmann, J.B., Verhaar, J.A.N., Busschbach, J.J.V., Bierma-Zeinstra, S.M. Wertli, M.M., Burgstaller, J.M., Weiser, S., Steurer, J., Kofmehl, R., Held, U., 2014.
A., Reijman, M., 2012. Psychological factors affecting the outcome of total hip and Influence of catastrophizing on treatment outcome in patients with nonspecific low
knee arthroplasty: a systematic review. Semin. Arthritis Rheum. 41 (4), 576–588. back pain: a systematic review. Spine 39 (3), 263–273.
Wang, C.H., Tsay, S.L., Elaine, B.A., 2005. Post-traumatic stress disorder, depression, World Health Organization. Global Road Safety Report 2018 [June 17, 2018]. Available
anxiety and quality of life in patients with traffic-related injuries. J. Adv. Nurs. 52 from: [Link]
(1), 22–30. Xie, F., Pullenayegum, E., Gaebel, K., Bansback, N., Bryan, S., Ohinmaa, A., Poissant, L.,
Wang, J., Wang, X., 2019. Structural Equation Modeling: Applications using Mplus. John Johnson, J.A., 2016. A time trade-off-derived value set of the EQ-5D-5L for Canada.
Wiley & Sons. Med. Care 54 (1), 98–105.
Ware, J.E., 1993. SF-36 Health Survey: Manual and Interpretation Guide. Health
Institute.

10