The methods for the FISH have been previously described [17]. In summary, NSW residents aged 17 or older, injured in a MVC in NSW, Australia, between August 2013 and December 2016 were recruited from selected public hospital emergency departments (5% from other sources), within 1 month of the crash. Patients with pre-existing cognitive impairment e.g. dementia, injuries resulting from intentional self-harm and death of an immediate family member in the crash were excluded. Isolated, superficial soft tissue (very minor) injuries or extremely severe injuries defined by eligibility for the NSW Lifetime Care and Support Scheme including very severe traumatic brain injuries, spinal cord injuries, extensive burns or multiple amputations were also excluded from the study.
Eligible patients were contacted by telephone and a structured interview was conducted at baseline (within 28 days of the injury), 6 and 12 months following informed consent. The study was conducted according to the Declaration of Helsinki and approved by the Central Sydney (Concord Hospital) Local Health District Human Research Ethics Committee. Outcomes including return to work, HRQOL, health status, disability and functioning, psychological factors, pain and compensation were collected using validated tools at 6 and 12 months.
Measures of injury severity
The baseline questionnaire asked participants if they had presented to the hospital due to MVC-related injury, and if so, the length of hospitalisation. Participants were divided into three groups according to self-reported hospital LOS (LOS ≤ 1 day – including those not presenting to hospital, LOS 2–6 days and LOS ≥ 7 days). These groups were based on the commonly defined cut-off for serious injuries according to the International Traffic Safety Data and Analysis Group [4].
We also used ISS to classify injury severity in the supplementary analyses. ISS is based on the anatomical injury severity classification of AIS [18]. The AIS classifies individual injuries by body region into 6 severity categories, with AIS 1 being minor and AIS 6 being maximal and untreatable [19]. ISS is calculated as the sum of the squares of the highest AIS code in each of the 3 most severely injured body regions of head or neck, face, thorax, abdomen, extremities, and external [18]. In the current study, we define ISS 1–3 as minor (with maximum AIS of 1), and ISS 12+ as severe (with at least 2 moderate or serious injuries).
ISS were derived using standard methods [18, 19] by a trained coder (KB). Data sources for the ISS coding were the AIS with all injuries specified for 51.3% of records, text data from the research data set for 34.7% of records, existing ISS data for 12.1% of records, a combination of ISS and AIS data for 0.4% of records where there were discrepancies between the two measures; an ISS of 1 was assigned to 0.4% of records with no injury information. ISS was directly calculated from the AIS in the first category above, in the second category the coder read all available text information and completed the standard calculation to derive the ISS. In the third category, the pre-existing ISS and AIS data were inconsistent, so all available data was used to assign the lowest feasible ISS.
All coding was completed in accordance with the AIS coding guideline of coding conservatively, that is, when multiple AIS codes could apply to the available data, assign the least severe AIS code in that injury category” [19].
Baseline variables, social and health outcomes collected
The baseline data included sociodemographic characteristics, employment, pre-injury health (BMI and history of chronic disease), HRQOL, health status, lifestyle habits, pain, disability and functioning, psychological factors, health care utilisation, injury, crash related factors, work and social life (social satisfaction and participation). During 6 and 12-months follow up, work, social life, health status and HRQOL, compensation, disability and functioning, psychological factors and pain were assessed. All psychometric scales used have been shown to be valid and reliable measures. The Short Form Survey (SF12) has 12 questions from the SF36 survey, and has two domains: the physical component summary (PCS) and the mental component summary (MCS) [20]. Higher PCS and MCS scores indicate better physical and mental wellbeing. The telephone administered version of EQ-5D-3L measures health status across five dimensions including mobility, self-care, usual activities, pain or discomfort, and anxiety or depression [21]. Each dimension has three response options (e.g. no, some and major problems), from which an overall summary index can be derived, based on health state valuations of each possible health state where 1 represents full health, 0 represents dead, and negative values represent health states valued as worse than dead and have a lower bound of − 1 [22].
Orebro Musculoskeletal Pain Screening Questionnaire (OMPSQ, short form) is a screening tool that predicts failure to return to work following a soft tissue injury and includes 10 items with the total score ranging between 1 (lowest risk) to 100 (highest risk), with a score of > 50 indicating a higher risk for future work disability [23]. The World Health Organization Disability Assessment Schedule II 12-item version (WHODAS II) has six domains including cognition, mobility, self-care, getting along, life activities and participation, ranging from no disability (0) to full disability (100) [24]. The Depression, Anxiety and Stress Scale (DASS-21) is a 21-item scale that provides a general assessment of psychological distress, depressive mood, anxiety and stress [25]. Impact of Events Scale – Revised (IES-R) is based on 22 self-reported items assessing subjective distress following traumatic events [26]. Further details of the study protocol have been published in the protocol paper [17].
Statistical analyses
Two sets of analyses were developed. Our primary analyses examined the associations of hospital length of stay with the outcomes of disability and functioning (WHODAS II), health status and HRQOL (EQ-5D summary score and SF-12) and work status. Supplementary analyses examined the association of ISS with these outcomes. Hospital length of stay was used for the primary analysis because it was collected in a consistent manner for all FISH participants based on participant self-report. Statistical analyses were performed using SAS v 9.4 (SAS Institute Inc., Cary, NC), MPLUS Version 7.3 and DAGGITY statistical software.
First, we examined descriptive statistics on baseline characteristics by subgroups of injury severity using means (SD), frequencies and percentages with chi-square tests, t-tests and general linear model F tests (Step 1). Second, we examined descriptive statistics for a variety of longitudinal health-related and psychological outcomes from baseline to 12 months, again using means (SD), frequencies and percentages with chi-square tests, t-tests and general linear model F tests (Step 2).
Third, we evaluated differences in paid work, SF12, EQ-5D summary score and WHODAS outcomes longitudinally from baseline to 12 months between levels of injury severity after adjusting for relevant covariates, using linear mixed models for repeated measures with unstructured serial correlation between time points within individuals. Consideration was given to the roles of the following factors via directed acyclic graphs: age, sex, preinjury health (comorbidities), preinjury EQ. 5D, education, preinjury work, recruitment source, social satisfaction, preinjury history of anxiety or depression, crash role, perceived danger in crash, hospital admission (for models of ISS level, not in models of hospital stay as a proxy measure for injury severity), pain at baseline, pain catastrophising (pain catastrophising scale) at baseline, DASS-21 and IES-R scores at baseline, compulsory third party insurance (CTP) claimant status. DAGGITY software was used to construct and examine the directed acyclic graphs, while SAS was used to run the linear mixed models. Adjusted model results (beta coefficients, 95% confidence intervals and p values) are presented for two models. Model 1 adjusted for minimally sufficient adjustment factors for both exposure and outcome: age, sex, crash role, perceived danger in crash, preinjury health, preinjury EQ. 5D, and recruitment source. Model 2 adjusted for all factors hypothesised to underlie either exposure or outcome status, including post-injury factors. Additionally, these models were adjusted for education, preinjury work, social satisfaction, preinjury history of anxiety or depression, pain at baseline, pain catastrophising at baseline, DASS-21 and IES-R scores at baseline, and CTP claimant status. (Step 3).
Fourth, since post-injury factors including baseline SF12 scores, baseline pain, baseline psychological status and CTP claimant status are possible mediators of the impact of more severe injury on long-term outcomes, we evaluated the presence of these mediating pathways for SF-12 PCS and MCS using MPLUS Version 7.3 (Step 4). The direct effect is defined as the effect of exposure on the outcome without the mediator, whereas the indirect effect is when the effect of exposure works through the mediator on the outcome.