Development of the Depression Medication Choice decision aid
We will use a novel practice-based, patient-centered approach for decision aid development based on design/participatory action research, developed and validated by the Knowledge and Evaluation Research Unit at Mayo Clinic (Rochester, MN, USA), and executed by a multidisciplinary team comprised of designers, patients, primary care and psychiatry clinicians, and decision making and health literacy scientists (Figure 1) [13, 14, 17].
This approach includes three main steps: 1. synthesis of clinical evidence regarding the risks and benefits of each option, and its approval by stakeholders; 2. observations of real-time clinical encounters involving clinicians and their patients with depression facing the decision to use antidepressants; and 3. prototyping of the decision aid through a series of iterations and field testing in the same context as the initial observations.
Synthesis of evidence
For the purpose of the development of the depression medication choice decision aid, we will synthesize the evidence regarding the risks and benefits of antidepressants based on the Agency for Healthcare Research and Quality (AHRQ) effective healthcare systematic review report on the comparative effectiveness of second generation antidepressants (07-EHC-007) [18]. We will then convene psychiatrists, primary care clinicians, patients and health plan representatives to meetings in which they will review the summary, and agree on its results and associated credibility and applicability.
Observation of encounters
Members of the research team led by a designer will observe approximately 10 to 15 usual clinical encounters, typically 3 to 5 without aids, and 3 to 5 with each iteration of the decision aid prototype. Using a predefined observation grid we developed and used in previous studies, we will look for patterns of patient-clinician conversations regarding antidepressant medications, as well as nonverbal behaviors and attitudes.
Prototyping and field testing
The construction of the Depression Medication Choice will be based on the summary of evidence and clinical observations, guided by the International Patient Decision Aid Standards (IPDAS) [19], which provides recommendations and standards for the development of decision aids. The initial prototype of the decision aid will result from discussions with members of the study team, clinicians, and patients from our patient advisory groups. They will be asked to comment on the content, format and ease of use of the prototype, on how the prototype aid captures current knowledge and addresses their concerns about antidepressants, and on to what extent the prototype helps bring to light patient context, values and preferences for decision making. Once developed, we will pilot this initial prototype in 3 to 5 clinical encounters, looking for patterns of the conversations, and documenting the issues, problems and challenges. When feasible, we will conduct a brief structured interview with the patient and the clinician after the encounter. The study team will then evaluate the quality of the conversations and the ability of the prototype to facilitate the decision making process. This process will be repeated with subsequent prototypes until the research team and stakeholders reach a consensus that the prototype is successful in involving patients in decision making (in our prior work it has taken up to seven iterations). This process should result in a discrete and highly intuitive decision aid, requiring minimal training to use, and designed to meet the needs of patients and their primary care clinicians working in fast-paced care settings.
Evaluation of the impact of the Depression Medication Choice decision aid
Study design and setting
We will conduct a clustered, practice-based, randomized trial of the Depression Medication Choice decision aid in ten rural, suburban and inner city primary care practices in the Midwest USA. The Mayo Clinic Institutional Review Board (IRB) has approved the study procedures described herein. We also obtained ethical approval from Hennepin County Medical Center Human Subjects Research Committee for the three participating practices not covered under the Mayo Clinic IRB. The trial design and protocol are registered in ClinicalTrials.gov (http://NCT01502891).
Study participants
Primary care practices were invited to participate in the study based on key criteria for a successful collaboration identified from past similar trials: sufficient number of clinicians interested in participating; an easily identifiable clinical champion; a large enough patient population meeting our inclusion criteria; and an on-site, or willingness to have an on-site, study coordinator. Practices that had concurrent research studies involving antidepressants were automatically excluded from participating in the study.
Primary care clinicians (physicians, nurse practitioners, physician assistants) from participating practices will be eligible for the study if they provide care to patients suffering from depression. Eligible clinicians will be invited to participate in the study through an initial meeting held by the investigators at the practice or individually by the on-site study coordinator.
Adult patients will be eligible for the study if they have a diagnosis of depression as judged by the clinician, a score on the Patient Health Questionnaire (PHQ-9) ≥10 [20], receive their depression care in the primary care setting, do not have bipolar disorder (as judged by the clinician), and have no major barrier (language, sensorial, cognitive) to providing written informed consent.
Eligible patients will be identified in advance from the appointment schedules of clinicians. All eligible patients who present to the practice for a depression visit during the study will be invited by an on-site study coordinator to participate in the study.
Allocation procedures
We will randomly allocate practices to the intervention or usual care arm. We will pair practices by size (the number of clinicians seeing adult patients with depression) and by whether they have implemented the DIAMOND (Depression Improvement Across Minnesota, Offering a New Direction) program, a practice redesign initiative to improve depression care through the use of care coordinators [21]. Within each pair, we will randomize practices to either: 1. the use of Depression Medication Choice within the clinical encounter; or 2. usual care. A study statistician will perform the randomization centrally after the practices have been enrolled and ensure concealment of allocation.
Practices, clinicians and investigators will not be masked to allocation. However, patients will remain unaware of the study’s main hypotheses. To prevent loss to follow-up and support the intention-to-treat principle, we will centrally follow patients, and ensure that patient surveys and pharmacy follow-ups are complete.
Intervention
The intervention will consist of the use of the Depression Medication Choice decision aid by patients and their primary care clinician during the clinical encounter. Training of clinicians will be provided prior to enrollment of their patients and will include a demonstration showing how to use the decision aid with patients. The focal points of the demonstration will be: 1. the decision aid serves as a guide for conversation rather than a script; 2. clinicians have flexibility in the manner in which they use the decision aid, including how and when they use it during the visit; and 3. clinicians may elect not to use the tool with certain enrolled patients, per their own judgment. Brief video clips and a storyboard to demonstrate the basic use of the decision aid will be available for clinicians to review at their convenience. A study team member will remain available to undertake one-to-one demonstrations after the initial group demonstration, as needed.
Usual care
For patients in the usual care arm, clinicians will manage the discussion about medication regimen as usual, without using the decision aid.
Data collection and analysis
We will collect patients’ data through: 1. self-reported questionnaires administered before and after the clinical encounter with their clinician, and at 3 and 6 months post-encounter; and pharmacy data 2. information about depression-related care included in medical records. Participating clinicians will be given brief questionnaires to complete once before and then immediately following each clinical encounter with a participating patient. In addition, we will videotape each clinical encounter where consent to do so has been provided by both the patient and clinician. Patient, clinician and practice socio-demographic information will be collected through questionnaires, and administrative or patient medical records.
Outcomes measures
The RE-AIM framework, developed to address how an intervention is implemented in a real-world setting, will guide this portion of the study and includes the following dimensions: Reach (how broadly is this intervention used within the practices), Effectiveness (what is the impact of the intervention on outcomes), Adoption (can this be adopted by new groups with ease and minimal modifications), Implementation (what are the special issues and barriers in implementation), and Maintenance (can the intervention be maintained and will the impact continue) [22, 23].
To assess the Reach of the decision aids, we will use a tracking log to record patients who are enrolled as well as those who declined the invitation to participate, and their respective characteristics. This will allow us to measure participation and representativeness.
To assess the Effectiveness of the decision aids, we will measure patient decisional conflict as the primary outcome, and decision making quality and processes (that is knowledge, satisfaction, preference in decision making style, patient involvement in decision making), depression outcomes, and medication adherence as secondary outcomes in the following ways:
-
1
Decisional conflict. Patients will complete the decisional conflict scale (DCS) immediately after the clinical encounter, the most commonly used outcome measure in decision aid trials [24, 25]. Psychometric properties include good internal consistency, test-retest reliability, and effect sizes for responsiveness to change ranging from 0.4 to 1.2. Clinicians will also complete a similar scale immediately after the clinical encounter [26].
-
2
Knowledge. Patients will complete a questionnaire addressing general and specific knowledge about depression care contained in the decision aids. This questionnaire will be developed according to prior recommendations and will be completed after the encounter, and at 3 and 6 months post-encounter.
-
3
Satisfaction. Patients and clinicians will be asked questions regarding the acceptability and satisfaction regarding the way information was shared during the encounter.
-
4
Preference. Patients will identify their preference in decision making style prior to and after the encounter [27]. Clinicians will also complete the same questionnaire at enrollment in the study and following each encounter with a participating patient.
-
5
Patient involvement in decision making. We will use the Observing PatienT Involvement in shared decisiON making (OPTION) scale to assess the extent to which clinicians seek to engage patients in decision making [28]. Our group has extended the use of this third-observer tool using video recordings of the encounters, with excellent inter-rater reliability (intraclass correlation coefficient >0.7) [14, 16, 17, 29].
-
6
Depression outcome. Patients will complete the nine items of the PHQ-9 [20], as a measure of the severity of depression symptoms, before their encounter, and at 3 and 6 months post-encounter.
-
7
Medication adherence. Patients’ measure of adherence and persistence to their depression medication will be derived from patient self-report and pharmacy records. We have been able to obtain complete pharmacy records for all of our participants in previous trials [14, 16, 17]. When mentioned the Adoption of the decision aids, we will estimate the proportion of practices or clinicians who adopt the intervention. Using clinician surveys and medical record review, we will compare the adoption rates across the intervention practices. We will select a random sample of participating patients’ medical records from each site to determine the extent to which the use of the decision aid is discernible in the records, and the extent of use and success when mentioned.
To assess the Implementation of the decision aid, we will seek to determine the extent to which the intervention (that is, use of the depression medication choice decision aid) is implemented as intended. Using the video recordings of the clinical encounters, we will assess the fidelity with which the decision aids are delivered and used as intended during these clinical encounters. The fidelity checklist is specific to the decision aid and setting, and will be developed based on previous checklists for other decision aids.
We will not be able to assess the Maintenance of the use of decision aid in practices as this requires a time period that is outside of the scope of the funding of the current study.
Socio-demographic and other patient characteristics
At entry in the study, patient characteristics collected will include age, gender, marital status, education, income, subjective health literacy [30, 31] and subjective numeracy [32]. We will collect the following clinician characteristics: gender, birth year, position, practice specialty and number of years in practice. Practice-level characteristics will include distribution of patient ethnicity, race, payment method, population of community, location of clinic (urban, suburban, rural), distribution of full-time equivalent across staff (clinicians, nursing staff, educators), average number of visits per week and average length of appointments.
Sample size
The Statin Choice randomized trial evaluated the decision quality by comparing the decision aid to usual care, and reported a 9.8 point difference in decision quality with the standard deviation (SD) of 16.9 and 14.1 for the usual care and decision aid groups, respectively [16]. Thus, to determine the statistical power of the current trial, we made the following assumptions: 1. variances will be as reported in the statin choice study; 2. the trial should detect a difference of 9.8 points or greater in decisional conflict between two groups at a significance level of 0.05, with a two-sided t-test; 3. a modest correlation of outcomes within practices represented by an intracluster correlation coefficient (ICC) (between cluster variance/total variance) of 0.05; and 4. a 10% attrition rate. We estimated the power using a standard calculation for a t-test, adjusted for clustering by a variance inflation factor (VIF) (1 + (n - 1) x ICC), where n is the number of patients per practice [33]. Under these assumptions, we estimated that we will have 90% power to detect at least a 9.8 difference in the DCS if we are able to recruit 30 patients per clinic for a total recruitment target of 300 patients. Assuming a similar ICC and attrition rate for other outcomes, this sample size will have 99% power to detect 1 SD difference in any continuous measure (for example approximately a 2-point difference in a 10-question knowledge scale), and 80% power to detect a 30% difference in 6-month adherence rates assuming a control adherence rate of 50%. Actual power is likely to be greater because we will adjust for baseline values and characteristics, and because of our conservative assumptions.
Analysis
Rather than assume that patient outcomes are independent of the clinician and practice, we will use techniques appropriate for clustered observations [33, 34]. For unadjusted comparisons, we will use cluster-adjusted t-tests and chi-square tests, and for all adjusted analyses we will use hierarchical generalized linear models (HGLMs) [35],]. These statistical methods address the ‘unit of analysis’ issue through terms for each level of grouping or clustering, and have a number of features important to this study. Specifically, these models account appropriately for clustering of patients within clinicians and practices, and allow us to deal with repeated observations of patients. Three types of outcomes will be compared between randomization groups, as classified by measurement period and unit of observation: 1. patient-level after the index visit; 2. patient-level after 6 months; and 3. clinician-level outcomes at end of the study period. All patient and clinician outcomes will be analyzed as randomized in adherence to the intention-to-treat principle. Patient characteristics will be compared between randomization groups, and all outcomes adjusted for any that are not balanced between groups.
We will measure the following outcomes at the patient-level after the index visit: 1. Decision to initiate antidepressants. This will allow us to estimate the differences in prescription of medication for the decision aid group compared to the control group. As a secondary analysis, we will conduct subgroup analyses by health literacy level to test for an interaction. We will use this test to determine whether the difference in uptake of depression medications in the decision aid and control arms differs for the low versus high health literacy groups; 2. Decisional conflict and decisional quality and processes. Through these analyses, we will evaluate the effect of decision aids on decisional quality, patient satisfaction with the encounter, and patient knowledge related to the decision.
We will evaluate the following outcomes at the patient-level using 3- and 6-month data: 1. Antidepressant medication persistence and adherence based on decision at the index visit. Medication persistence will be measured as continuation of depression medications at 6 months for patients who leave the index visit with prescriptions for antidepressants; this includes their first medication and changes in the prescriptions that occur in the 6 months following the index visit. Medication adherence will be measured using fill dates from the pharmacies, measured using the percentage of days covered approach; 2. Depression outcomes. Depression treatment outcomes will be measured at baseline and at 3 and 6 months using the PHQ-9 [20]. This is the measure used by practice-based interventions in primary care, such as the DIAMOND initiative, there are practice pathways to collect this information, its score is linked to clinical action and it is familiar to primary care clinicians [36].
Finally, we will evaluate the clinician-level outcomes also using random effect models. However, for these analyses we will only cluster within clinic and account for the repeated measures for satisfaction and visit length, while controlling clinician and practice characteristics.