Study design
We conducted a pre/post study guided by the Ottawa Decision Support Framework (ODSF) [14]. This framework is helpful for understanding a deliberative process for making a preference-sensitive decision, or a decision with more than one option, with unknown outcomes, or the known benefits and harms can be valued differently by patients. The ODSF posits that health decisions are negatively influenced by unmet decisional needs, such as feeling uninformed, having unclear values, and feeling unsupported, and unresolved decisional conflict. Quality decisions are more likely when patients are informed by the best available evidence and the patient’s values and preferences [14, 15]. In short, decisional needs can be addressed using decisional support interventions (decision coaching using a patient decision aid) to improve decisional outcomes and reduce decisional conflict [16]. The study was approved by the Research Ethics Boards at the Children’s Hospital of Eastern Ontario and the University of Ottawa.
Participants and setting
Youth and their parents were recruited from an ambulatory diabetes program at a tertiary academic pediatric hospital. The hospital serves an urban and rural population of 1.3 million in Eastern Ontario, Canada; the diabetes program provides care for 850 children and youth with T1D. At the time of the study in 2013–2015, MDI was rarely used by children or youth in our centre. Since then, MDI has become the usual insulin delivery method from diagnosis onwards.
We recruited youth and parents who had told either their pediatric endocrinologist or pediatric diabetes physician during their regular diabetes clinic visit that they were considering a change in insulin delivery method, were capable of participating in the decision making process and were scheduled for decision coaching by one of our diabetes social workers which is a step in the process for youth in our clinic who are considering a change in insulin delivery method. To be eligible for this study, youth had to be under 18 years old with type 1 diabetes duration of at least 10 months, and they and their parents had to be able to read and speak English or French. No lower age limit was set for youth participants, as required by our Research Ethics Boards, provided the youth and parent(s) could participate in the consent or assent process. Family dyads (youth and one parent) and family triads (youth and two parents) were included. The study was introduced to youth and parents being scheduled for decision coaching by the administrative assistant for the diabetes team. A research assistant contacted those who expressed interest in the study. This contact was by telephone to assess study eligibility and explain the study in detail. Youth and parents, regardless of the youth’s age, who agreed to participate provided written informed consent, and assent by the youth if necessary, prior to the decision coaching.
Procedures
Youth and parents were referred for decision coaching by their diabetes physician. Decision coaching sessions were conducted in private meeting rooms within the diabetes clinic and booked separately from regular diabetes clinic follow-up appointments. Data were independently collected from youth and parent participants at three time points: baseline (T1), immediately post decision coaching (T2), and 10 to 14 days post decision coaching (T3) by return mail. At T1 and T2, a research assistant was present to help the youth and parents’ work through the data collection form by explaining the forms and to answer any questions. Participants did not adopt a new insulin delivery method prior to T3.
Intervention to support shared decision making
Participants received decision coaching guided by the Ottawa Family Decision Guide (OFDG), a generic decision aid suitable for any difficult health or social decision [16]. The OFDG is a practical shared decision making tool that provides a structured approach for multiple stakeholders to work through the decision making process, which was available in both English and French. For this study, the OFDG was pre-populated in advance with evidence on benefits and harms of each of the three insulin delivery options based on the Canadian Diabetes Association Clinical Practice Guidelines (1) and reviewed by the local centre’s pediatric diabetes educators and pediatric endocrinologists. Options included: a) standard insulin therapy (2 or 3 injections/day) at the time of the study, b) multiple daily injections and, c) insulin pump therapy.
Two social workers, who were members of the diabetes team and trained in interprofessional shared decision making and decision coaching, provided tailored decision coaching sessions to family dyads and triads. The diabetes interprofessional team nominated social workers as decision coaches for this study because it was thought that decision coaching was a ‘natural’ fit in the social workers’ role in this clinic and the intervention could be incorporated into the clinical pathway with little disruption. Decision coaching training consisted of the Ottawa Decision Support Tutorial and a 3-h skills building workshop. This training intervention has been shown efficacious for improving healthcare providers’ decision support knowledge in randomized controlled [17]. The social workers followed a decision coaching protocol to guide the sessions, which emphasized encouraging the youth to speak first to minimize power-imbalances and discourage parents’ biasing the youth’s responses.
We made two changes from the pilot study [12], based on feedback from the decision coaches. The first was to change the introductory question from “Should I/my child get an insulin pump?” to “What is the best insulin delivery method for me/my child?” We also changed the order in which the options were presented, so that insulin pump therapy became the third instead of the first option presented in the coaching session. The rationale for these changes was to minimize perceived bias that may have been present in the pilot study related to positioning insulin pump therapy as the first, and therefore potentially preferred option, for consideration.
Outcome measures
Paper-based, self-administered questionnaires were used to measure participants’ decisional conflict (T1 and T3), choice predisposition (T1 and T2), and satisfaction with the decision coaching intervention (T3). The primary outcome was change in decisional conflict from T1 to T3. We used the low literacy version of the validated Decisional Conflict Scale (DCS) [18, 19]. We chose this version to accommodate various literacy levels when working with youth and parents. The low literacy DCS has good internal reliability with alpha coefficients ranging from 0.80 to 0.86 [18]. The 10-item low literary DCS evaluates four domains of decisional conflict: a) feeling informed (3 items), b) feeling clear about personal values (2 items questions), c) feeling supported in decision making (3 items), and d) feeling certain about best choice (2 items). Each question has three response categories (yes/no/unsure). The scale ranges from 0 (no decisional conflict) to 100 (extremely high decisional conflict). Scores equal to or exceeding 37.5 are associated with delaying decision making or feeling unsure; scores less than 25 are associated with implementing the decision [18].
Secondary outcomes were: change in choice predisposition from T1 to T2, dyad agreement about preferred choice, and satisfaction with the intervention. The validated Choice Predisposition Scale was used to measure the preferred option pre-intervention [20]. This scale has a test-retest coefficient > 0.90 [20, 21]. We compared choice predisposition at T1 to immediately after the intervention (T2) and compared the youth’s choice at T1 and T2 to parent(s). When two parents participated, we included both parents’ responses and compared their choices by randomly choosing 1 parent. Participants’ satisfaction with the intervention was measured using a questionnaire that combined a SDM satisfaction rating tool [22] and a modified version of the Genetic Counselling Satisfaction Scale which has excellent reliability with an alpha coefficient of 0.90 [22, 23].
Data analysis
We managed our data using REDCap software, a secure web-based data management application [24]. Descriptive statistics were used to summarize demographics outcomes. Means (with standard deviations) were calculated for continuous variables, and proportions for categorical variables. A paired t-test was used to compare the overall DCS score pre (T1) and post coaching (T3). Exploratory t-tests were also conducted for the DCS sub-scales pre and post. Cohen’s kappa coefficient was used to determine rater agreement about the preferred treatment option within each family (i.e., youth-mother, youth-father, mother-father) before and immediately after decision coaching. Overall treatment option agreement was evaluated using the Fleiss’ kappa coefficient, which accounts for agreement between multiple raters (e.g., youth-mother-father). Statistical significance was set at p < 0.05 and adjusted for the four DCS subscale analyses to make the significance level 0.0125. With decisional conflict as the primary measure, the sample size of 45 provided 80% power to detect an effect size of 0.4 with alpha of 0.05 allowing for 10% attrition rate.