Study design
We undertook a secondary data analysis of cross-sectional surveys conducted among compensated transport accident victims in Victoria in 2010 and 2011, linked to compensation claims and payment records.
Ethical considerations
The Monash University Human Research Ethics granted exemption from ethical review as the study satisfies 5.1.22 of the National Statement on Ethical Conduct in Human Research ([18], p.40): “Institutions may choose to exempt from ethical review research that: is negligible risk research; and involves the use of existing collections of data or records that contain only non-identifiable data about human beings”.
Transport injury compensation system
In the state of Victoria, Australia, those injured in land-based transport accidents involving a car, motorcycle, tram, bus or train are eligible to claim compensation for treatment, income replacement, rehabilitation and long-term support services via the Transport Accident Commission (TAC), regardless of fault. In addition, the TAC provides compensation for injury and death for individuals travelling in a Victorian-registered motor vehicle in other Australian states. Injuries and death occurring on the road but not involving a motorised vehicle (e.g. a collision between a pedal cyclist and a pedestrian) are not eligible for compensation [19].
Data sources
Data were collected from the Client Outcomes Survey (COS). The TAC conducts an annual COS to measure the health and vocational status of its clients. The survey is designed to inform the TAC about the impact of its claims management practices and the design of the compensation scheme on the health and vocational outcomes of its clients. The survey includes standardised measures of vocational and health status prior to injury, current vocational status, current health status, including physical and mental health, pain, mobility and functional independence, access to and satisfaction with healthcare and satisfaction with the TAC. Data are collected via computer automated telephone interview (CATI) conducted by a third-party social research organisation. The questionnaire takes approximately 25 min to administer.
Data were also collected from the Compensation Research Database (CRD). The CRD is an administrative database held by the Institute for Safety, Compensation and Recovery Research (ISCRR) at Monash University. The database contains de-identified transport-related injury claims data from 1 January 1987 through to 31 December 2014. The database contains one record for every claim received by the TAC, and each record contains information necessary for the management of the compensation claim, including accident, demographic, and injury details, and payments for health and other services. For the purposes of this study, the CRD was linked to the COS via a unique claim identifier.
Study participants
In 2010 and 2011, a total of 2476 participants completed the COS, including 1649 (67 %) participants with MSI. The sample age ranged from 16–89 years. The sample included active and inactive claims. Active claims were defined as having a payment from the TAC within the last six months prior to being surveyed. Inactive claims were defined as having no payments in the last six months but at least one payment made within seven to 24 months prior to being surveyed. The sample comprised of minor to moderate injuries such as soft tissue or complex orthopaedic/multi-trauma, including mild and moderate brain injury. Catastrophic injuries such as spinal cord injury, severe traumatic brain injury, amputees and burns were excluded. In this study, the sample was limited to participants with MSI including sprains/strains, soft tissues, fractures and dislocations.
Measures
Predisposing characteristics
Predisposing characteristics included gender (female vs. male), age, country of birth (Australia vs. others), education (university level vs. less than university level), employment status at time of accident (yes vs. no), occupation, and socio-economic indexes for areas (SEIFA). Age was defined as the age of claimant at the time of the interview and was categorised into six groups: 16–24, 25–34, 35–44, 45–54, 55–64, and 65+ years. Among those working at the time of the accident, occupation was categorised into eight groups according to the Australian and New Zealand Standard Classification of Occupations (ANZSCO) [20]: managers, professionals, technicians and trade workers, community/personal service workers, clerical/administration workers, sales workers, machine operators/drivers and labourers. SEIFA is a measure of relative disadvantage and advantage based on a range of attributes such as a person’s residential location and income [21]. The distribution of scores was divided into ten equal deciles. A high decile reflects relative advantage. The deciles were recoded into two categories, where decile 1–5 reflected relative disadvantage and deciles 6–10 reflected relative advantage.
Need factors
Need factors included physical health, mental health, pre-injury health status (excellent, very good, good, fair and poor), injury types (soft tissue, sprains and strains, fractures, whiplash), time since injury, and hospitalisation (>1 day hospital stay vs. not hospitalised) which was used as a proxy for injury severity [22, 23]. Physical health and mental health were assessed by the Short-Form-12 Health Survey Version 2 (SF-12V2). The SF-12V2 is a validated international tool that consists of twelve questions [24]. The SF-12V2 measures eight concepts: physical functioning, role limitations due to physical health problems, bodily pain, general health, vitality (energy/fatigue), social functioning, role limitations due to emotional problems, and mental health (psychological distress and psychological well-being). Two summary scores were derived: the Physical Component Summary (PCS) and the Mental Component Summary (MCS). The PCS focuses mainly on limitations in physical functioning, role limitations due to physical health problems, bodily pain, and general health. The MCS focuses mainly on role limitations due to mental and emotional problems and social functioning. The scores were derived using Australian weights based on the Australian population norms [25]. Higher scores on the PCS and MCS indicated more positive physical and mental health. Time since injury was derived from the date of the interview and the accident date.
Enabling factors
Structure of social support included marital status and number of dependent children. Marital status was grouped into married/de facto relationship, widowed/separated/divorced and never married. Preliminary analysis found an association between marital status and number of dependent children; thus a family structure composite was created. The family composition was categorised into six groups: married/de facto relationship with children, married/de facto relationship with no children, widowed/separated/divorced with children, widowed/separated/divorced with no children, never married with children, never married with no children. Sources and functions of social support included accessing help from family and friends. For family and friends items, participants rated their level of agreement with the following question; ‘Can you get help from family members/friends if you need it?’ on a 4-point scale that ranged from 1 “yes, definitely” to 4 “no, not at all”.
Healthcare service use
Two categories of healthcare services were examined in the two year follow-up period from the date of the accident: allied and mental healthcare services. Allied healthcare services included services provided by physiotherapists, chiropractors, osteopaths, acupuncturists and occupational therapists. Mental healthcare service included services provided by psychiatrists, psychologists, general practitioners (restricted to mental health treatment plan only), social workers and vocational counsellors. Allied healthcare service use was measured as the number of services accessed in the two year follow-up period. Due to the small number of mental health care services accessed, mental health care services use was transformed into a binary variable, those who accessed mental health care services (yes) and those who did not (no).
Statistical analyses
Descriptive statistics and frequency distributions of key variables are presented. For the direct effect analyses, two types of models were conducted to examine an association between each source of social support and healthcare service use. Allied healthcare services use was analysed using zero-inflated negative binomial regression (ZINB) modelling. ZINB is a maximum-likelihood count regression analysis, designed for non-normal (i.e. skewed and over dispersed) count data with an excess of zero values [26]. The ZINB models the probability of being a non-user versus a user of healthcare services (i.e. the logistic model component) and weighs cases accordingly in order to determine the prediction of healthcare services use intensity (i.e. the negative binomial regression model component). Vuong tests were conducted to assess the appropriateness of a ZINB model against the standard negative binomial regression model. Mental healthcare services use was analysed using logistic regression modelling. Both models were adjusted for predisposing factors and need factors.
For the mediation analyses, we used the Karlson, Holme and Breen (KHB) method [27] to assess whether social support mediates the association between predisposing factors, need factors and each type of healthcare services use. This method provides unbiased decompositions of total effects into direct and indirect effects for both linear and nonlinear models. The decomposition is accomplished by comparing the estimated coefficients obtained from a reduced model (without mediator) to a full model (with mediator). The differences between these two sets of estimated coefficients provide an estimate of the indirect effect (i.e. the part of the total effect running through the mediating variable). However, the KHB method is currently not suitable for count models. Therefore, allied health care service use was transformed from a count variable into a binary variable - those who access allied health care services (yes) and those who did not (no).
Lastly, for the effect modification analyses, we tested interaction effects to see whether social support modifies the association between predisposing factors, need factors and healthcare service use. A ZINB model with interaction effect was used to analyse allied healthcare services use and a logistic regression model with interaction effect was used to analyse mental healthcare services use.
In all statistical models, the “not often” category in the sources of social support variables was used as the reference group instead of the “no, not at all” category as participants who rated not receiving any support may not be a homogenous group (e.g. participants who did not require any help, or did not have family living in the area). A p-value of less than 0.05 was considered significant in all analyses. Data analyses were conducted using STATA version 12 and SAS version 9.4.