The ESPARR cohort
The ESPARR project is a prospective cohort study of road accident casualties, which seeks to identify long-term consequences and to better define what constitutes serious injury. It is based on the Rhône Registry of Road Traffic Casualties , which, since 1995, has recorded all casualties receiving medical care in public or private health facilities in the Rhône administrative area of France. The inclusion criteria are: (1) having had a road traffic accident in the Rhône administrative area involving at least one mechanical means of transport; (2) living in the Rhône administrative area; (3) having survived the accident at least up to hospital admission; and (4) having received care in one of the area’s hospitals. The injured are followed up for 5 years after the accident. All subjects have provided written consent.
Each individual lesion is coded using the Abbreviated Injury Scale (AIS) : from 1 (minor injury) to 6 (fatal injury). The Maximum-AIS (M.AIS) is the injury’s highest AIS score and defines overall initial injury severity.
The inclusion period for road accidents was initially from October 1st, 2004 to December 31st, 2005, later extended to July 31st, 2006 for the most seriously injured (M.AIS ≥ 3). The protocol has been described in detail in a previous publication .
The cohort comprises 1,168 adults, aged ≥16 years at the time of the accident. At inclusion, the injured were asked to answer a questionnaire specifically drawn up for the ESPARR study, administered in a face-to-face interview.
The questionnaire gathered data on the accident and previous familial, occupational and health status. In addition, initial lesion assessment and other medical data (radiology, intensive care, etc.) were collected from medical records. Lesions were coded by the trained physician of the Registry from which the cohort was derived.
The present study selected the 546 adults in the ESPARR cohort who had sustained only mild injury, defined as maximum AIS level 1 (MAIS-1), excluding cases with associated AIS ≥ 2 lesions in other body regions; 253 of these subjects had sustained whiplash injury and the other 293 had any other type of MAIS-1 lesion, such as ankle or shoulder sprain, superficial wounds or contusions, tendon tear, etc. In all, 352 subjects (64.5 %) responded to the 5-year follow-up questionnaire (between October 2009 and December 2010): 167 of the 253 whiplash cases (66.0 %), comprising 63 grade-1 and 104 grade-2 injuries, and 185 of the 293 other mild injury casualties (63.1 %). 16 % of non-whiplash and 20 % of whiplash casualties failed to respond to the 1-year questionnaire.
Clinical definition of whiplash
In the present study, all subjects with lesions classified as cervical contusion (AIS code 310402) or neck sprain (AIS code 640278) were considered as whiplash casualties. Diagnosis was made by physicians, at the outset of hospital care, based on interview, clinical findings and X-ray. In the AIS classification, Code 310402 is attributed to neck pain following a road accident with painful neck on palpation without other objective signs, and corresponds to grade 1 in the Quebec classification; Code 640278 is attributed to neck pain associated with cervical stiffness and radiologic loss of cervical lordosis, and corresponds to grade 2 in the Quebec classification . Cervical spine lesions graded AIS-1 (code 640278) but with associated neurologic abnormality (Quebec grade 3) (n = 2) were excluded.
Variables and measurements
Outcome measurements at 5 years
QoL was evaluated on the World Health Organization Quality of Life (Brief) (WHOQoL-Bref) questionnaire . This tool contains 26 questions. The first two assess perception of overall QoL and perception of overall health, respectively. The other questions are categorized in four domains: physical, psychological, social and environmental . Responses to each question are graded on a five-point Likert scale for intensity, capacity, frequency or evaluation, as the case may be (from 1 = poor QoL, to 5 = good QoL). Each of the four scores (one per domain) ranges from 4 to 20, modified in the present study to 0–100 for comparison with WHOQOL-100 scores used in other studies. A high score indicates good QoL.
For analysis, the two generic questions (overall QoL and general health status) of the WHOQoL-Bref were coded as dichotomous variables: good or very good versus neither good nor bad, bad, very bad or no reply for overall QoL; and satisfied or very satisfied versus neither satisfied nor dissatisfied, dissatisfied, very dissatisfied or no reply for perception of overall health.
Variable of interest
The variable of interest for analysis was whiplash status, in three categories: non-whiplash, grade-1 whiplash, and grade-2 whiplash.
At inclusion, the questionnaire collected sociodemographic and accident-related data, plus some medical information and psychological history. Specifically, sociodemographic data comprised age, gender, family situation, educational level and socio-occupational category. Accident-related information comprised type of road-user with position inside the vehicle and impact direction, antagonist, reason for travel at time of accident, subjective responsibility for accident (admitting being at fault in the accident or not), friend or family member also injured in the same accident, and intention to lodge a complaintFootnote 1. Financial problems (job loss, financial difficulties, failure) in the year before the accident and psychological history (sleep disorder, consumption of antidepressants/anxiolytics, regular appointments with a psychologist) in the year before the accident were also considered in this analysis.
Post-traumatic stress disorder (PTSD) at 6 months and/or 1 year after the accident was also taken into account and was assessed using the Post-traumatic CheckList Scale (PCLS) . The PCLS includes 17 items relating to the 3 dimensions of the disorder (re-experiencing, avoidance and hyper-arousal), and has been shown to have good specificity for the diagnosis of PTSD; validation studies of the PCLS [29–32] have shown good psychometric properties. The validation of the French version  showed that PCLS score ≥ 44 indicates presence of PTSD and the existence of disturbances which will necessarily affect lifestyle.
The presence of pain and sequelae of accident-related injuries was collected in the 5-year questionnaire. A free text about pain location allowed data to be classified according to body region (head, neck, face, spine, thorax, abdomen, upper limbs, and lower limbs); a dichotomous variable was created for each region (pain vs. no pain).
The representativeness of the study population (whiplash and non-whiplash subjects) was assessed by comparing sociodemographic and accident data between respondents and non-respondents at 5 years’ follow-up. Descriptive statistics were used to describe the distribution of variables. Chi2 test (significance level, 5 %) or Fisher’s exact test (small samples) were used for categoric variables; Student’s test (normal data distribution) or Kruskal-Wallis test (non-normal data distribution) were used for continuous variables.
QoL data (the two generic questions and mean scores in the 4 domains) collected at 1 and 5 years after the accident were compared for subjects responding at the both follow-up steps. Thus, it was possible to detect a significant improvement or deterioration in QoL for each group, using McNemar’s test (significance level, 5 %) or Student’s test for matched data.
The next step was to identify factors for impaired QoL, with whiplash status as the variable of interest. Two modified Poisson regression models were built for each of the first two variables of the QoL scale: i.e., poor overall QoL and unsatisfactory overall health. Variance analysis was used to study each of the four QoL domains. Analysis strategy was identical in all models. Age and gender, considered as adjustment variables, were included in the multivariate analysis regardless of their significance level. The variable of interest (non-whiplash, grade-1 whiplash, grade-2 whiplash) and explanatory factors at time of accident or at 1-year follow-up significantly associated (≥10 %) with outcome on univariate analysis were included in the multivariate analysis, after checking for collinearity between explanatory variables. Stepwise selection with backward elimination was applied, with p > 0.05 for exclusion (model 1). For each outcome, the final model was built from model 1, to which the variable “pain at 5 years” was added (model 2).
QoL was also analyzed separately for casualties with and without PTSD, and interactions between whiplash and pain were investigated.
Statistical Analysis System software, version 9.3 for Windows (SAS Institute Inc., Cary, NC, USA) (proc genmod and proc glm) was used for all analyses.
Ethics and consent
The study protocol was submitted to and approved by the French Ministry of Research (CCTIRS: Advisory Committee on Information Processing in Material Research in the Field of Health) (CCTIRS Number 04.159). Data collection and analysis were approved by the national data protection authority (CNIL: CNIL Number 04–1417). Lastly, only patients (or their family) who gave written consent for follow-up were included in the cohort. At any time during the follow-up period, subjects were free to cease participation, and, in that case, to be totally withdrawn from the study files and analyses.
Availability of data
Data are available from the authors upon request (CNIL’s requirement).