Target-D: a stratified individually randomized controlled trial of the diamond clinical prediction tool to triage and target treatment for depressive symptoms in general practice: study protocol for a randomized controlled trial

The study will be conducted in at least ten general practices in Victoria, Australia (see Additional file 3 for the location of study sites).

The follow-up phone call from an RA will involve a discussion about the treatment recommendation they received, using the results of the diamond CPT to tailor the discussion to the individual’s classification. To encourage treatment engagement, this discussion will use motivational interviewing, an approach to conversations that promotes collaboration and aims to strengthen the person’s motivation and commitment to making a change [41].

The recommended treatment for each of the three groups were selected based on a stepped care approach [22], with treatment intensity lowest in the minimal/mild group and highest in the severe group. To select the specific treatments offered to each level of intensity, we examined existing primary care data from the diamond cohort study to describe the characteristics, treatment, and service use of individuals stratified to each group. We also reviewed systematic reviews of the evidence relevant to each group and presented these findings to our investigator team to inform treatments offered. Another comprehensive description of the interventions delivered, using the TIDier checklist, will be included in the primary outcome paper as per Hoffmann et al. [46] and CONSORT guidelines.

Participants randomized to the comparison arm will access health services as usual. The choice of “usual care” as a comparator was made as the study aims to determine the extent to which the intervention improves (or worsens) patient outcomes relative to standard practice [56]. Participants in this arm will also receive some non-therapeutic attention from Target-D to control for any effect of contact with the study team following completion of the diamond CPT; thus, this study arm is referred to as “Usual care plus Target-D” or UC+. UC+ participants will be blinded to their depressive symptom severity group allocation and will not receive a tailored treatment recommendation. Instead, they will see a screen on the Target-D website advising them that they will be asked to provide feedback on: (1) their opinions on research in primary care; and (2) how they normally manage their emotional health and wellbeing. Similar to the procedure for intervention participants, those in the usual care arm will be contacted by phone by an RA within one week of diamond CPT completion. The RA will reiterate the importance of the participant’s involvement in the study, ask a series of structured questions about the participant’s views on research and the involvement of their general practice in research, and advise the participant that he or she will be contacted via email in 12 weeks.

The nature of the study interventions is such that no substantive modifications are anticipated. Patients in the minimal/mild and moderate groups may discontinue using the online program at any time and treatment for those in the severe group may be discontinued at patient request.

If any participant indicates high levels of suicidal ideation during contact with a member of the study team (as indicated by a response of “nearly every day” to question 9 on the PHQ-9: “thoughts that you would be better off dead or of hurting yourself in some way”), regardless of study arm allocation, a standardized suicidal ideation assessment used previously by the study team will be administered and the patient’s GP alerted. This will be reported as an adverse event but is unlikely to result in treatment discontinuation or modification. The assessment will determine if the adverse event was related directly to the intervention or other circumstances not intervention related.

In both the usual care and intervention arms, participants will be permitted to continue any treatment they were engaged with at entry to the trial. Concomitant care will be assessed via self-report questionnaire and routinely collected Government data (see below).

In the intervention arm, adherence to treatment in the minimal/mild and moderate depressive severity groups will be assessed using website analytics within myCompass and This Way Up (i.e. tracking individual log-ins, access of components, completion of modules and lessons). In the severe group, adherence to the treatment plan will be assessed by the Target-D CM as part of the planned follow-up schedule.

In the control arm, in order to compare “usual care” before and during participation in the trial, information about health service use will be collected at each study assessment (see below).

The primary outcome is the difference between the two treatment arms in mean depressive symptom severity at three months, controlling for baseline depressive symptom severity. Depressive symptom severity was selected as the outcome measure rather than a clinical diagnosis of major depressive disorder as it is more relevant to the design and delivery of stepped mental healthcare.

Secondary outcomes include difference between study arms in mean depressive symptom severity at 12 months and mean mental-health self-efficacy and anxiety at three and 12 months. The cost-effectiveness of the intervention over the study period will comprise an additional secondary outcome.

We will follow principles of good recruitment by engaging with all stakeholders, branding the Target-D trial, and using a well-developed engagement strategy [58]. We will recruit general practices via our Victorian Primary Care Practice-Based Research Network (VicReN), which has around 200 GP members located in Victoria, Australia. Practices will be contacted by phone and/or email to introduce the study and establish interest. One of the Target-D researchers will then visit interested practices to determine eligibility, provide detailed information about the study, and gain consent to participate. This process will continue until sufficient practices are recruited to obtain the required sample size.

To enhance GP and practice staff engagement in the trial and the activities necessary to make it function, we will be guided by the principles of Normalization Process Theory (NPT) [59]; namely, coherence (meaning of the trial to GPs and staff), cognitive participation (commitment and engagement), collective action (the work GPs and staff do to make the trial function), and reflexive monitoring (GP and staff appraisal of the trial). In each participating practice, GPs and staff will be given a training session clarifying the goals and activities of the trial, in order to instill a sense that the trial is a good idea and worth committing to. In addition, we will clearly outline the trial procedures and how they are likely to affect the work of the practice, with emphasis on how the trial fits with the overall goals of the general practice.

Randomizing individuals that were recruited from the same general practice, where the clinician is not blinded to the participant’s study arm or the intervention is implemented at the practice level, may have a greater risk of contamination between intervention and comparison arms than randomizing a group of individuals that do not belong to the same practice [60]. However, the risk of contamination in this trial is expected to be minimal because of several factors. First, recruitment of participants will be conducted in the waiting room by an RA who is not involved in delivering the study intervention and has no access to the allocation schedule. While patients in the waiting room or from the same family or friendship groups may share information, it is unlikely that this will impact the intervention effect as the intervention can only be accessed by permission of the study team. Second, GPs will only be informed of participants allocated to collaborative care treatment. Even if other patients inform their GP that they are participating in Target-D, GPs will not be informed of their treatment allocation nor be able to access study interventions for UC+ patients. Third, the intervention for minimal/mild groups will be via Internet-based programs delivered outside the practice, reducing the potential for practice-based contamination. We will assess the number of UC+ participants registering for these programs to measure the degree of potential contamination. Fourth, the risk that GPs may implement some of the intervention to patients predicted to have severe symptoms and allocated to the UC+ group is small. We anticipate that fewer than ten such participants will be recruited per practice and this small number of patients will be seen by different GPs. We have successfully used a similar approach in a previous RCT in general practice; data from this trial showed very low levels of interaction between comparison participants and the GP during the study time-frame [61].

Potential participants will be alerted to the study via posters and information pamphlets displayed in practice waiting rooms; an awareness raising strategy used by the research team previously. All study materials and procedures were developed and tested with focus groups and individual feedback to ensure they are engaging and user-friendly. Upon completion of the screening survey, eligible participants will enter their name, telephone number, and email address into an online form. They will then be presented with an electronic copy of the plain language statement and provide online consent to participate.

To achieve our required sample size, trained RAs will approach an average of 50 adults per practice per day, taking approximately 440 working days to approach our required 22,000 adults. These numbers are typical of RCTs in primary care and we have achieved similar numbers in previous successful studies [39, 62, 63]. While this number may seem ambitious, several factors contribute to it being achievable. First, we have designed the trial so that participants are recruited by RAs rather than expecting already time-pressured GPs and practice staff to take responsibility for recruitment. Second, the 22,000 patients RAs will approach include all adult patients in the waiting room; RAs or practice staff are not required to identify only those patients who are presenting to the GP for mental health reasons. Finally, we have tried this approach out in a clinic and shown that on weekdays alone an RA can approach at least 50 patients per day and invite 1100 patients per month to the study. Our pilot work has shown that an RA spends only 1 or 2 min with each patient and can comfortably approach 150 patients in a working day. Based on this experience and after accounting for weekend recruitment in some practices, we anticipate participant recruitment to take place over approximately 18 months. Recruitment of participants will continue until the numbers within each depressive symptom severity group have been met.

Consent and baseline measures will be collected prior to randomization to minimize reporting and selection bias. When the individual has completed the diamond CPT, he or she will be randomly assigned in a 1:1 ratio to the intervention or comparison arm. Randomization will be stratified by general practice and depressive symptom severity group. The allocation sequence will be computer-generated sequentially within stratum using a biased-coin algorithm [64] embedded within the Target-D website which is housed on the secure National eResearch Collaboration Tools and Resources (Nectar) cloud which provides computing infrastructure to Australian researchers. Using restricted randomization within the stratum ensures that number of individuals is balanced between study arms within stratum and the stratification factors will be balanced in each study arm. The randomization will be triggered automatically within the Target-D website, after a participant has completed baseline measurements and the diamond CPT, thus ensuring allocation concealment.

Due to the nature of the intervention, participants cannot be blinded to their treatment allocation. However, GPs will not be notified of their patients’ allocation to either intervention or comparison arm. No emergency unblinding of GPs is anticipated, including in the case of the research team alerting the GP to patient suicidality. As outcome assessment is conducted online, no blinding of outcome assessors is required. All study analyses will be conducted by a statistician blind to participants’ allocation; study arm allocation will be coded as A or B, with the code for the study arm revealed only after data are analyzed.

Depressive symptom severity will be assessed at each timepoint using the PHQ-9. The PHQ-9 assesses the nine DSM symptoms of depression over the last two weeks using a 4-point Likert scale (where 0 = “not at all” and 3 = “nearly every day”). Total scores are in the range of 0–27, with suggested cut points of 5, 10, and 15 indicating mild, moderate, and severe depression, respectively [40]. The PHQ-9 is a validated diagnostic measure in primary care [65], with demonstrated efficacy and sensitivity as an outcome measure for treatment trials with a recommended Reliable Change Index [51].

Self-efficacy will be measured using the Mental Health Self-Efficacy Scale (MHSES) [66]. The MHSES comprises six items that require respondents to rate on a 10-point Likert scale how confident they are in performing behaviors related to mental health self-care (from 1 = “not at all confident” to 10 = “totally confident”). Total scores are in the range of 6–60 and provide a unidimensional measure of self-efficacy; higher scores indicate greater levels of self-efficacy. The MHSES displays high internal consistency (Cronbach’s alpha = 0.91) and good construct validity, correlating well with measures of depression, anxiety, and functional impairment.

The seven-item Generalized Anxiety Disorder scale (GAD-7) will be used to assess anxiety [67]. The GAD-7 assesses the presence of anxiety symptoms over the past two weeks using a 4-point Likert scale. Scoring is similar to the PHQ-9; each item is scored from 0 to 3 (for a total possible score of 0–21), with cut points of 5, 10, and 15 corresponding to mild, moderate, and severe anxiety symptoms. The GAD-7 has excellent internal consistency (Cronbach’s alpha = 0.92) and test–retest reliability. Its construct, convergent, and discriminant validity are high; it correlates well with measures of depression and functioning (while assessing a distinct construct), as well as with other measures of anxiety.

Quality of life will be assessed at each time point using the Assessment of Quality of Life (AQoL-8D) [68]. This is a validated, reliable measure [69] that comprises eight dimensions (independent living, senses, pain, mental health, happiness, self-worth, coping, and relationships) that can be used to calculate quality-adjusted life years (QALYs) via a utility algorithm. The AQoL-8D has been shown to be sensitive to depressive symptom severity levels [69].

Cost-effectiveness of the intervention will be measured through assessment of health service use, effects on productivity, and calculation of QALYs. Health service use will be tracked using data extracted from the Australian Government Department of Health: the Medicare Benefits Schedule (MBS) that maintains information about visits to healthcare providers and diagnostic tests; and the Pharmaceutical Benefits Scheme (PBS) database of medications supplied on prescription. Participants will provide additional consent to access their MBS and PBS data. Other resource use not captured by these national databases, including the use of broader health and welfare services and effects on productivity (i.e. education and workforce participation), will be assessed via self-report using an adapted questionnaire developed by members of the research team and used in numerous other Australian mental-health intervention trials [70‐72].

Analyses will use an ITT approach, where participants will be analyzed in the study arm to which they were allocated [78]. In the first instance, we will implement strategies to minimize the missing outcome data, including the participant retention strategies outlined above. Reasons participants are lost to follow-up will be recorded. Sensitivity analysis will be used to assess the robustness of the missing data assumption.

Incremental cost-effectiveness ratios (ICERs) will be determined (cost of intervention – costs of comparison/outcome of intervention – outcome of comparison) using the AQoL-8D to determine QALYs. ICERs using other important study outcomes (such as cost per remitted case) will also be determined. Variation will be determined by bootstrap and regression analyses and results presented in cost-effectiveness planes and acceptability curves. Sensitivity analyses will also be used to determine the impact of important study parameters (such as unit cost price variation). Dependent on trial results, modeling may also be used to extrapolate beyond the trial time horizon.