The iBobbly Aboriginal and Torres Strait Islander app project: Study protocol for a randomised controlled trial

The following demographic variables will be ascertained: age, gender, Aboriginal or Torres Strait Islander status, usual form of accommodation (own residence, parent, boarding, shelter, no fixed address/homeless, other), employment status (full-time, part-time, part-time and looking for full-time work, unemployed/looking for work, not in labour force), usual occupation (free text), educational attainment (some primary, all primary, some secondary, 4 years secondary, 6 years secondary), post-secondary or tertiary schooling (none, trade/apprenticeship, other certificate, diploma, bachelor or higher degree), and relationship status (single, de facto, separated, married, divorced, widowed).

The Suicidal Ideation Attributes Scale (SIDAS) [24] is a brief measure of severity of suicidal ideation. The five-item scale assesses frequency (item 1), controllability (item 2), closeness to attempt (item 3), distress (item 4), and interference with daily activities (item 5) over the past month. Respondents answer on an 11-point scale (ranging from 0 to 10) with total scores ranging from 0 to 50. Higher scores indicate greater suicidal ideation. Item 2 (controllability) is reverse scored. Those who endorse a frequency of 0 for the first question skip the remaining items and are given a score of 10 (full control) for item 2 (controllability). The SIDAS is suitable for online use. The scale has demonstrated high internal consistency with Cronbach’s alpha of 0.91, and good convergent validity. In an online survey of community-based Australian adults (n = 1352) [24], a cut-off of 1 had sensitivity of 85.5% for suicide plans and 84.0% for suicide attempts (with 67.1% and 63.6% specificity, respectively). However, high specificity is needed to identify individuals most likely to engage in suicidal behaviour. At a cut-off score of 21, the SIDAS has 95.8% specificity for presence of plan in the past year and 94.9% specificity for presence of preparation for or making a suicide attempt in the past year. The suicidal intent and suicidal plan questions of the Quick Inventory of Depressive Symptomatology (QIDS) will be used to assess suicide plans.

The General Help-Seeking Questionnaire (GHSQ) is a brief measure of a person’s intentions to seek help for emotional or personal problems and suicidal thoughts [25]. The measure consists of two questions, the first asking the likelihood of seeking help for an emotional or personal problem and the second asking the likelihood of seeking help for suicidal thoughts. The two questions list 11 possible sources someone might ask for help, e.g. their friend, doctor, etc., and includes the statement “I would not seek help from anyone” as a source. Each source is rated on a seven-point scale from 1 (extremely unlikely) to 7 (extremely likely). The scale demonstrates good reliability (α = 0.83 for the suicidal ideation question and α = 0.70 for the emotional/personal problems question) and validity amongst a sample of Australian high school students [25].

The Resource Use Questionnaire (RUQ) is a modified version of the Client Socio-Demographic and Service Receipt Inventory—European Version (CSRI) [26]. The language in the questionnaire has been simplified to suit Indigenous Australians where English may not be their first language. The RUQ captures service use measures, including contact with mental and primary health care staff and medication usage. A comparison of GP records with data from the CSRI found no significant difference in the average number of GP visits recorded on GP case records (mean = 3.03; sd = 3.49) and on the RUQ (mean = 2.99; sd = 3.67), and the agreement between the two sources was high (concordance correlation coefficient 0.756) [27]. The measure was reviewed by Indigenous staff for appropriateness for economic evaluation within the Aboriginal community.

The Patient Health Questionnaire (PHQ-9) is the depression scale of the PHQ, and consists of nine items, each scored 0 (not at all) to 3 (nearly every day). Participants are asked to rate how problematic the items have been for them over the past 2 weeks, yielding an overall score ranging from 0 to 27. The scores are grouped into five categories: 0–4 representing no/minimal depression; 5–9 representing mild depression, 10–14 moderate, 15–19 moderately severe, and 20–27 severe depression. The PHQ-9 has demonstrated good internal validity in participants attending primary care clinics (α = 0.89, n = 6000) and obstetric patients (α = 0.86, n = 3000). It has an acceptable test–retest reliability of kappa = 0.84 [27]. An adaptation of PHQ-9 for Indigenous Australian communities was utilised in this study [28], and a validation study is currently underway [29].

The 15 items on the Distress Tolerance Scale (DTS) measure a person’s ability to experience and tolerate negative psychological states [30]. Scored on a five-point scale from strongly agree (1) to strongly disagree (5), a higher overall score on the scale indicates a higher perceived capacity to tolerance distress (range = 15 to 75). The DTS has been found to be a valid measure of distress tolerance in a non-clinical university student sample [30] and a wider community sample of smokers [31]. Cronbach’s alpha for the overall scale was found to be 0.91 [31].

The Interpersonal Needs Questionnaire (INQ) is comprised of 25 items that measure thwarted belongingness and perceived burdensomeness [32]. The respondent reports how they have been feeling recently on ten items that measure belongingness and 15 items that measure burdensomeness. The items are rated on a seven-point scale from 1 (not true at all) to 7 (very true). Higher overall scores indicate higher levels of thwarted belongingness and perceived burdensomeness (range = 18 to 126). The INQ has been found to be a reliable and valid measure of belongingness and burdensomeness in younger and older adults in clinical and non-clinical samples [32]. Comparable internal consistency coefficients were found for the belongingness (α = 0.85) and the perceived burdensomeness items (α = 0.89) and construct validity data were encouraging (e.g., thwarted belongingness, more so than perceived burdensomeness, correlates with a theoretically related interpersonal construct, i.e., loneliness) [32].

The AQoL-8D will be used to assess health related quality of life [33]. It is commonly used to calculate utility scores before and after intervention for use in economic evaluation studies. The AQoL-8D can be completed in 5 minutes and is comprised of 35 items from which eight dimensions and two ‘super-dimensions’ are derived. The 35 items may be reduced to a single utility score using the AQoL-8D algorithm. AQoL-8D score ranges from full health (1.00) to health states worse than death (− 0.04) [34]. In addition, the algorithm produces an index number for each of the eight dimensions and for the two ‘superdimensions’, ‘Physical super-dimension’ (PSD; independent living, pain, senses) and ‘Mental super-dimension’ (MSD; mental health, happiness, coping, relationships, self-worth). The AQoL-8D has been shown to have good criterion validity, with Cronbach’s Alpha of 0.97.

Primary analyses will be undertaken on an intent-to-treat basis, including all participants as randomised, regardless of treatment actually received or withdrawal from the study. Likelihood based methods (mixed-model repeated measures [MMRM]) methods will be used to analyse change in the primary outcome measure (SIDAS). An a priori planned comparison of change from baseline to the post-intervention endpoint (6 weeks) will be used to test the primary hypothesis. An unstructured variance–covariance matrix will be used to accommodate relationships between observations at different occasions. The stratification factor of gender will be included in these models and variables found to be substantially imbalanced (equating to a moderate effect size) between groups post-randomisation will be tentatively included but not retained to evaluate the robustness of conclusions reached. Similar analyses of continuous/scaled secondary measures will assess differential change due to intervention arm. Mathematical transformation or categorisation of raw scores into ordered or dichotomous classes may be undertaken to meet distributional assumptions and address any violation of assumptions attributable to outliers. MMRM assumes data are missing at random [35, 36]. As this assumption cannot be substantiated from observed data, sensitivity analyses will be undertaken to evaluate effects of missing data and the robustness of the trial’s conclusions. As appropriate, these will be based on plausible models of missingness informed, as much as possible, by the experience of those working closely with participants.

For dichotomous outcomes such as caseness, a comparable generalized mixed modelling approach will be used. Number needed to treat [37] to avoid high risk of suicidal behaviour will be derived from the SIDAS using a cut-off score of 21 [24]. Number needed to treat to avoid moderate depression will be derived from the PHQ-9 using a cut-off score of 10 [38]. Separate analyses will incorporate 6-month follow up data. At this stage it will be possible to estimate post-intervention change in the active group and also to estimate change in the wait-list group after use of the app; however, comparison with an untreated sample will not be possible. All tests of treatment effects will be conducted using a two-sided alpha level of 0.05 and 95% confidence intervals.

Relevant costs will be applied to identified resource use during the trial, including the costs of the intervention, other health services utilisation (including use of ambulance, visits to hospital, medications, and physician contacts) and out-of-pocket expenses (cost to the individual for mental health and social services care). The primary outcome for the CEA will be quality of life (QALY) as measured using AQoL-8D. Change in suicide ideation (SIDAS) will be used as a secondary outcome. An incremental cost-effectiveness ratio (ICER) will be calculated for both the QALY outcome and the suicidal ideation outcome. Confidence intervals will be estimated around the ICER using non-parametric bootstrapping, and a cost-effectiveness acceptability curve will be estimated.

Power to detect change in ideation is based on the expected effect of the intervention at the 6-week time point on the primary outcome measure (SIDAS) after completion of the intervention as delivered by the app (6 weeks post baseline). Expected effects were based on the findings from a recent Dutch trial using a modified CBT website in the general population [39] which reported an effect size of 0.28 (Cohen’s d). A smaller effect than this on suicidal ideation would mean the app would be of limited interest as an intervention. To be able to detect this effect size, assuming a correlation between baseline and post-intervention measures of 0.5, including an expected drop-out rate of 30%, 289 participants, yielding complete data from 202 people, will be needed in each condition (with alpha = 0.05 and power of 0.80). This will also result in good precision, yielding confidence intervals of width approximately ± 0.20 for the effect size. US data show that 33.4% of adolescents with suicidal ideation also report plans [40]. Assuming this rate of plan reporting in the control condition, a reduction of plans under the active intervention to 20.5% (relative risk 0.61) would be required to maintain 80% power using Fisher’s exact test. While the study is not explicitly powered to detect this outcome, it is plausible that reduction in rates of plan-making may be detected with the target sample size. It may be feasible to determine when plans were made, allowing the testing of hypotheses regarding plans using more clinically informative methods such as survival analysis.

A Data Safety Monitoring Board (DSMB) has been established in accordance with the Australian National Health and Medical Research Council Guidelines. The DSMB is independent of the funders, has no competing interests, and its members are not study investigators. The DSMB is comprised of three senior academics experienced in conducting randomised controlled trials in suicide prevention and mental health. One member is also a biostatistician. The DSMB will receive 6-month data reports and will be notified of adverse events, as required by the human research ethics committee (HREC). Their findings on adverse events will be reported to the HREC. The DSMB has the authority to stop the trial if interim analyses indicate that sufficient sample size has been achieved, or there is sufficient evidence that the intervention is causing harm. Site coordinators will report adverse events to the trial manager within 72 h of notification of the event. Police and postvention services in each site have been asked to advise the trial manager of any deaths by suicide within the region. The investigator team acts as the steering committee. The trial manager reports to the principal investigator, and the principal investigator is responsible for liaising with the DSMB and the HREC. Site coordinators report to the trial manager. Protocol amendments will be reported to the investigators and submitted to all ethics committees for their approval prior to implementation.