Study design and recruitment
An online cross-sectional survey was designed and advertised to a community-based sample of young Australians. Ethics approval for this study was obtained from the University of New South Wales Human Research Ethics Committee (HC200465).
Potential participants completed a short screening survey to determine eligibility based on the following criteria: aged 16 to 25 years, living in Australia, fluent in English, had experienced suicidal thoughts in the past 12 months, and had engaged with a mental health practitioner for any reason in the past 12 months. Participants aged 16 and 17 years could participate without parental consent provided they were deemed Gillick competent (Griffith, 2016). There were no study exclusion criteria.
The survey was published using the online survey platform Qualtrics on October 9th 2020, and data collection remained open until the recruitment objective was met (N = 513) (12th October 2020). During this period, the survey was promoted on social media via targeted Facebook advertisements (target parameters: ages 16–25, Australia, interests in National Suicide Prevention, Lifeline, Suicide prevention, Beyondblue, Headspace, Lifeline, R U OK Day, SANE). When a potential participant clicked on the study advertisement, they were directed to an online study portal and asked to provide individual digital consent before completing a five-item screening survey to determine eligibility. Participants who did not meet the inclusion criteria were directed to a webpage thanking them for their time and which provided relevant support contacts, such as Kids Helpline and Lifeline phone numbers and webchat addresses. Participants who met the study inclusion criteria were directed to complete a 20-minute online self-report structured questionnaire, which also included free text options for questions relating to reasons why they chose whether or not to disclose suicidal thoughts to their mental health practitioner, and what would motivate them to disclose to a mental health professional in future. Participants who completed the survey were emailed a $10 e-gift voucher as compensation for their time.
Measures
Demographic characteristics were measured to describe the sample, including age, sex, gender identity, sexual orientation, rural/remote or metropolitan location, relationship status, and mental health status.
Disclosure
of suicidal ideation (primary outcome) was measured with the question, “Have you told your mental health professional that you have suicidal thoughts?” (‘yes’ or ‘no’). A follow-up open ended question was asked based on the response: “What factors made you choose to tell them you have suicidal thoughts?” or “What factors made you choose not to tell them you have suicidal thoughts?”. Open ended questions included multiple choice response options based on common reasons observed in the literature and free-text options. We also asked participants, “Can you tell us what sort of things would make you more likely to tell a mental health professional that you have suicidal thoughts?”.
Suicidal ideation was measured using the Suicidal Ideation Attributes Scale (SIDAS, van Spijker et al., 2014). The SIDAS consists of 5 self-reported items rated on a 10-point scale (0 to 10). The scale provides a total score ranging from 0 to 50, with a higher score indicating greater suicidal ideation severity. Negatively worded items are reversed scored and scores of 21 or greater indicate a high risk for suicidal behaviour (attempt). The scale has demonstrated excellent internal consistency (α = .91, van Spijker et al., 2014).
Depressive symptoms were measured by the Patient Health Questionnaire Depression Scale (PHQ-9, Kroenke et al., 2001). This scale consists of 9 items rated on a 4-point scale, ranging from 0 (not at all) to 3 (nearly every day). Higher scores indicate the presence of more depressive symptoms, and the maximum total score is 27. The scale has demonstrated good internal consistency (α > .80, Kroenke et al., 2016).
Anxiety symptoms were measured by the Generalised Anxiety Disorder-7 Scale (GAD-7, Spitzer et al., 2006). The GAD-7 consists of 7 self-reported items rated on a 4-point scale (0 = not at all to 3 = nearly every day), with a total score ranging from 0 to 21 and higher scores indicating more severe anxiety symptoms. The scale has demonstrated good internal (α > .80, Kroenke et al., 2016)
Psychological distress was measured using the Distress Questionnaire-5 (DQ5, Batterham et al., 2016), which consists of 5 self-reported items rated on a 5-point scale (1 = never to 5 = always). The scale provides a total score ranging from 5 to 25, with a higher score indicating greater psychological distress. The scale has demonstrated good internal consistency (α = .86, Batterham et al., 2016).
Suicide attempt (lifetime history) was assessed with the question “Have you ever attempted suicide?” (‘no, never’, ‘yes, once’, or ‘yes, more than once’). For participants who answered more than once, an additional question asked them to specify the number of attempts.
Exposure to suicide loss was measured by asking, “Has anyone close to you died by suicide?” ‘yes’ or ‘no’. This question was a modified from a recent study by Maple and Sanford (2020).
Personal suicide stigma was measured using the Personal Suicide Stigma Questionnaire (PSSQ, Rimkeviciene et al., 2019), which consists of 16 self-reported items rated on a 5-point scale (1 = never to 5 = very often). The scale provides a total score ranging from 16 to 80, with a higher score indicating suicide-related stigma experiences. The scale has demonstrated excellent internal consistency (α = .96, Rimkeviciene et al., 2021).
Prioritisation of mental health issues was measured with the question, “In the following list of mental health problems, we’d like you to rank the top 3, according to how important these are to you when talking to your mental health professional.”. Respondents could rank a list of ICD-10 categories of mental health diagnoses, including suicidal thoughts, from 1 (most important) to 3 (less important). Prioritisation of suicidal ideation specifically (ranked as top 3; ‘yes’ or ‘no’), was used in regression analyses.
Therapeutic alliance was measured using the Revised Helping Alliance Questionnaire (HAq-II, Luborsky et al., 1996), which asks respondents to carefully consider their relationship with their most recent therapist and rate 19 items on 6-point scale according to how strongly they disagree (1) to agree (6) regarding the mutual collaboration and bond between client and therapist. The scale provides a total score ranging from 19 to 114, with a higher score indicating greater alliance with the therapist. Negatively worded items are reversed scored. The scale has demonstrated excellent internal consistency (α ≥ .90, Barber et al., 1999).
Statistical Analysis
To detect non-disclosure of suicidal ideation in the cohort (primary outcome), a total minimum sample size of n = 471 was needed. This estimate accounts for a youth population-level incidence of 12-month suicide ideation of 26% (Nock et al., 2008) and a 34% rate of suicidal ideation in a youth population engaged with mental health services (Bruffaerts et al., 2011), with power set at 90%, alpha set at 0.01.
Descriptive information was presented as proportions (%) and means (with standard deviation; SD). T-tests and chi-square (χ2) tests were conducted to establish whether the participants who did and did not disclose suicidal ideation differed significantly on demographic and clinical characteristics. Significant variables (with a more liberal cut-off of p < .10, see Ranganathan et al., 2017) were chosen to be entered into a subsequent binary logistic regression model to examine what factors are independently associated with disclosure of suicidal ideation to a mental health professional. In this model, disclosure (yes/no) as the dependent variable was used to estimate odds ratios (ORs) with 95% confidence intervals (CIs).
Quantitative data were analysed using SPSS version 25.0, alpha was set at p < .05 for interpreting significant effects. The following guidelines were used to interpret effect sizes: correlation coefficient (r) values of .10, .30, .50, and .70, correspond to small, medium, large, and very large effect sizes, respectively (Cohen, 2013); Phi (ϕ) values of .10, .20, .30, and .40, correspond to small, medium, large, and very large effect sizes, respectively (Funder & Ozer, 2019); ORs of 1.44, 2.48, and 4.27 correspond to small, medium, and large effect sizes, respectively (Cohen, 2013); raw means were calculated to estimate Cohen’s d, with values of .20, .50, and .80 corresponding to small, medium, and large effect sizes, respectively.
Free-text responses were qualitatively analysed using reflexive thematic analysis via NVivo version 12. Responses were thematically analysed, with similar responses grouped together by two authors (LM and DR) independently. Individual responses were then reviewed between the two authors and discrepancies were discussed until agreement was reached.