Psychological distress following a motor vehicle crash: preliminary results of a randomised controlled trial investigating brief psychological interventions

The details of the protocol, such as recruitment, design, setting and aims for this randomised controlled trial (RCT), have been previously reported in the study’s published protocol paper [30]. A CONSORT checklist file shows how the recommendations for a clinical trial have been addressed (Additional file 1). Participants included MVC survivors who lodged a compensation claim in New South Wales (NSW) or Victoria, Australia, between July 2015 and May 2017. Subsequent to publication of the protocol paper [30], a third recruitment site was secured in NSW. Recruitment sites included: (1) Suncorp and (2) the National Roads and Motorists’ Association in NSW and (3) the Transport Accident Commission in Victoria. Figure 1 shows the flow for the study’s recruitment, intervention and assessment processes. The NSW compulsory third-party insurance system is fault based; therefore, none of the NSW people in the study were at fault in their crash, whereas Victoria participants could be either at fault or not at fault. The inclusion criteria were (1) adult (age > 18 years) survivor of an MVC who lodged a claim within 4 months of their MVC and (2) English speaking. Exclusion criteria included the presence of severe injuries, such as spinal cord injury, amputation, blindness, severe traumatic brain injury and other injuries requiring extended hospitalisation. Table 1 shows participant characteristics including socio-demographic, injury and intervention-related characteristics by group.

This RCT is ongoing, with participants being assessed across four time points: baseline assessment upon entry into the RCT (mean 12 weeks post-MVC), and 3, 6, and 12 months post-intervention. For this preliminary investigation, the trial was stopped at the midway point when 30 participants in each group had completed the baseline, the intervention and the 3-month post-intervention assessment.

Based on prior results for the CBT and HL interventions [7, 14, 15], the two interventions were assumed to have at least a small to moderate mean effect size of 0.25 (Cohen’s d) compared to the control group. Assuming α = 0.05 and a power of 80% for a three-group comparison analysis with four measures over time, a sample of 135 participants needed to be recruited to detect true differences [30]. It was, therefore, proposed to recruit at least 180 participants to accommodate loss to follow-up of approximately 25% based on similar research [31].

The preliminary analysis shows a total of 240 individuals received information about the trial, of whom 104 were randomised into one of three groups: CBT (n = 37), HL (n = 35) or active waiting list (control, n = 32). We included 104 participants in this analysis to ensure attrition did not impact on the 30/30/30 preliminary analysis. The CBT and HL groups received fortnightly modules in Microsoft PowerPoint presentation format with homework worksheets, self-monitoring templates and instructions for using CBT or HL skills. CBT and HL participants also received clinically focussed telephone support on alternate weeks to encourage and remind them to practise the skills and to read the module material. The control group received compensation-related material every second week and a phone call on the alternate week. The reading material for the control group was restricted to publicly accessible claim information, and telephone calls confirmed this information was received. Table 2 shows the content of each module for each intervention group. Preliminary findings for the first 90 participants who completed the baseline assessment, interventions and 3-month follow-up are reported here and Fig. 1 shows the flow of these participants from eligibility through to this preliminary analysis.

A selection of the measures employed in this study have been analysed for this preliminary investigation. These include socio-demographics, MVC and injury characteristics, pre-injury mental health, self-reported body mass index (BMI), pain intensity and aspects of acceptability and feasibility. Psychometric measures included the Depression, Anxiety and Stress Scale (DASS-21) and the Impact of Events Scale (Revised) (IES-R). Structured interviews based on criteria from the 5th Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) were used to diagnose MDD and PTSD.

DASS-21 is a 21-item self-report scale providing an assessment of the severity of psychological distress as a total score and in three domains: depressive mood, anxiety and stress [32, 33]. Participants completed 21 4-point Likert items (0–3) assessing self-reported distress over the past week. Higher scores indicate elevated distress. Total scores are calculated by summing all 21 items [33], and then, in accordance with the DASS-42, scores were multiplied by 2 [33]. DASS-21 has sound psychometric properties, including acceptable internal reliability and validity [32].

IES-R is a 22-item self-report measure of trauma-related distress [34], validated in people with traffic injuries [35]. Respondents indicate their degree of distress during the past 7 days related to their recent MVC. It is a 5-point scale ranging from 0 (not at all) to 4 (extremely) with subscales for avoidance (e.g. avoidance of feelings or situations), intrusion (e.g. distressing thoughts) and hyperarousal (e.g. irritability and hypervigilance). Domains are scored by determining the mean item score [34]. Higher scores indicate increased distress.

Participants were given an identification number to ensure anonymity, and an internet link to complete their assessments through Survey Gizmo, a professional online secure survey software tool. All groups completed identical assessments at baseline and post-intervention (approximately 3 months post-injury). The data analyst had no contact with participants and was blind to randomisation. The intervention program coordinator, who also provided clinically focussed telephone support, was blind to all assessment data.

The purpose of this preliminary analysis was to analyse participants diagnosed with MDD at baseline and determine whether cessation of recruitment to either intervention group was required. As part of this analysis, preliminary efficacy trends were investigated to provide some indication of acceptability and feasibility. Statistical significance at this point in the RCT analysis was not the major objective, nevertheless limited statistical analyses were performed using SPSS version 22 [40]. While it is understood these analyses were preliminary and underpowered, we believe it is important to examine efficacy trends. The analysis included descriptive statistics for the primary outcome measures (Table 3). General linear repeated measures MANCOVA was used to determine differences over time and between groups for the total scores from DASS-21 and IES-R. The total scores were used as primary outcomes as they provide maximum variance. As described in the original protocol [30], it was decided to adjust for variables believed to influence primary outcomes (e.g. the baseline value of a primary outcome) and those pre-specified in the protocol [41]. Therefore, as specified, physical health (the physical component summary of the baseline SF-12) was adjusted for in the DASS-21 and IES-R analyses. Further, baseline DASS-21 depressive mood was adjusted for in the DASS-21 total score analyses and IES-R intrusion was adjusted for in the IES-R total score analysis. These two baseline measures were chosen because they are related to the primary outcome and their inclusion as covariates that were adjusted for baseline severity improved the efficiency of the analyses. Baseline differences between groups were regarded as randomisation anomalies [42].

Table 1 shows that between groups, the participants did not differ significantly by age, sex, BMI, marital status, education, employment, role in the MVC, days since the MVC, injury type, perceived pain intensity or previous mental health history. Table 3 shows descriptive unadjusted data for DASS-21 and IES-R measures over time. No significant differences were found between groups for baseline DASS-21 and domains. For IES-R, the HL group had significantly higher levels of intrusion and total scores compared to the CBT group (P < .05). Table 4 shows the frequencies of diagnosed MDD and PTSD by groups over time. There was no significant difference in baseline rates of MDD, or over time between groups (χ²₂ = 1.4; P > .05). There was a significant difference in baseline rates of PTSD (χ²₂ = 9.5; P < .01; HL had a higher rate of PTSD). No group differences in rates of PTSD occurred post-intervention.

The primary sub-group analyses compared changes in the primary measures between groups and those meeting DSM-5 criteria for MDD (no separate sub-group PTSD analyses were conducted as all but one participant with PTSD met the criteria for MDD). For participants without MDD, Fig. 2 (DASS-21 total score) and Fig. 3 (IES-R total score) show significant reductions in distress over time for all groups as measured by DASS-21 (F(1,79) = 4.0; P < .05) and by IES-R (F(1,79) = 10.0; P < .01). For those diagnosed with MDD, Fig. 4 (DASS-21 total score) shows a non-significant interaction effect in which the CBT and control groups trend toward reduced distress over time compared to the HL group. For those with MDD, Fig. 5 (IES-R total score) shows a non-significant interaction effect in which the CBT group shows a trend toward reduced distress compared to the HL and control groups. Table 5 shows adjusted means for group by MDD status. Secondary analyses showed there was no main effect or interaction differences between total groups over time, though a significant reduction in distress occurred over time regardless of group for DASS-21 and IES-R (P < .01).

Acceptability of the interventions appears to be positive. In the CBT group, 18 of 30 participants (60.00%) were satisfied or /very satisfied with the program, 11 (36.66%) were neither satisfied nor dissatisfied and 1 (3.33%) was very dissatisfied. For the HL group (two had missing data), 17 of 30 participants (56.66%) were satisfied or very satisfied, 8 (26.66%) were neither satisfied nor dissatisfied and 3 (10%) were dissatisfied. Similar acceptability of the CBT and HL modules was found. The participants also believed the program was worthwhile, with 22 of 30 (73.33%) CBT participants and 18 (60.00%) of HL participants responding positively. Finally, 21 of 30 (70.00%) CBT participants would recommend the CBT program, whereas 20 of 30 (66.66%) HL participants would recommend the HL program. Attrition rates were less than 20% (see Fig. 1).

While recruitment into the RCT has been slow, the method and goal of recruitment are feasible. At this preliminary stage, 104 participants have been recruited with 14 dropouts, resulting in 30 per group. Slow recruitment is due in part to reluctance to participate in prospective research involving treatment soon after an MVC, especially in the context of sustaining a physical injury and being engaged in a potentially stressful compensation process. Figure 1 shows that at the time of analysis, 61% who met the inclusion criteria agreed to receive information about participation in the trial. Of those who received information, 65% consented to participate. Some participants gave a reason for non-consent: (1) no time to read modules, (2) excessive pain, (3) assistance not required and (4) legal advice against receiving assistance. These reflect the difficulty of recruiting through an insurer, and further, delays occurred due to frequent re-organisations of the insurance company and the limited time for case managers to introduce the research to potential participants. Recruitment within 4 months from MVC to claim lodgement proved problematic due to: (1) delays in achieving contact with participants and (2) the need for frequent reminders to provide consent and complete assessments. Table 1 shows that there were approximately 3 months (mean days = 82.09, standard deviation, SD = 57.08) from the MVC to completion of the baseline assessment. Completion of the post-intervention assessment often did not occur immediately after program completion, often taking several weeks and repeated phone calls (mean calls = 2.99, SD = 2.4). Further, the need for frequent reminders to complete assessments may have negatively influenced participants’ program feedback compared to their positive verbal feedback received during the intervention. Rapport ratings (Table 1) suggest more positive rapport was established with the CBT group than the HL group, with 28 of 30 (93.33%) positive ratings for CBT compared to 23 of 30 (76.67%) positive ratings for the HL group.

Finally, not all who met the inclusion criteria had access to email, resulting in some postal delivery of modules. Table 1 shows email was an efficient means of delivery with 84 (93.33%) of 90 participants receiving their modules by email, and only 6 (6.67%) having no access to email who received modules via the post.

The 104 (30/30/30 with attrition) participants are more than likely a biased sample, given that fewer than 50% of the participants approached by the insurers agreed to enter the trial. Reasons for this include approaching participants at a stressful time, a perception of no need for psychological help and the opt-in style used for recruitment. It is also noted that the sample appears highly educated and the impact of all these limitations on bias in the sample needs to be considered when drawing conclusions. A further limitation relates to pre-morbid mental health history. This information is not routinely collected by insurance companies. However, the information collected about prior treatment by a psychologist or psychiatrist and on whether participants are taking prescribed psychiatric medications do provide a proxy for pre-morbid mental health history, which has been shown to be a strong predictor of post-MVC psychological distress [44, 45]. However, Table 1 showed pre-morbid psychological distress was not significantly different between the groups. We intend to explore fully the relationship between pre-morbid psychological distress and distress outcomes at the completion of the trial.