Background
Patients with complex chronic diseases are presented with competing disease priorities; competing patient-physician priorities further complicate care. Effective shared decision making (SDM) tools have been adapted for use in chronic care including diabetes [
1,
2] and may enable prioritization of treatment options. With SDM, patients and clinicians establish an ongoing partnership, exchange information, deliberate on options, decide upon the priority for taking action, and then act on the decision [
3]. A systematic review of randomized controlled trials of SDM identified 11 studies [
4], five of which examined physical and psychological wellbeing. Two of these studies reported positive outcomes; these two studies examined long-term decisions in the setting of chronic disease, suggesting that SDM may play an important role in complex diabetes care. A more recent systematic review of SDM in older people identified 22 studies, two in patients with diabetes, and found that SDM increased knowledge, increased risk perception, reduced decisional conflict and enhanced participation in SDM [
5].
Despite this evidence supporting the role of SDM in complex diabetes care, the integration of SDM into clinical practice is limited. Specific to diabetes, patients have reported that patient/clinician power imbalance, health literacy, and denial of the condition were barriers to SDM. Provision of medical knowledge, validation of patient experiences, strong interpersonal skills, and clinician availability were facilitators of SDM [
6].
An interprofessional (IP) team approach may overcome these barriers to SDM. Interprofessional care, where professionals from different disciplines collaborate to provide an integrated approach to patient care [
7] is particularly appropriate for diabetes care. Participation by more than one profession, expanding roles, and adding new team members in diabetes care has been demonstrated to improve clinical outcomes [
8‐
10] and may increase uptake of SDM [
11].
Furthermore, SDM can be facilitated by the use of patient decision aids (PtDAs) [
12‐
14]. For example, an observational study by Corser et al. showed that using a PtDA to support goal-setting in diabetes care increased patient knowledge (
P = 0.001) as well as the number of documented diabetes goals (pre: 0.67 goals; post: 1.09 goals;
P < 0.001), though there was no change in glycemic control, weight, or diabetes empowerment score [
15]. However, this study did not examine an interprofessional approach nor specifically integrate principles of shared decision-making.
Building on this study, we hypothesized that a multi-component SDM toolkit (patient-directed, clinician-directed, and point-of-care tools) that individualized care priorities and incorporated an IP approach to SDM would help to prioritize complex guideline recommendations for patients with type 1 or type 2 diabetes and other comorbidities. The aim was to explore decision-making experiences of health professionals and patients with diabetes and co-morbidities, then develop an intervention employing user-centred design to facilitate IP-SDM and goal-setting.
Methods
This study was part of a larger study focused on developing and evaluating an SDM toolkit for goal setting in patients with diabetes and other comorbidities; the development and evaluation protocol is described in detail elsewhere [
16]. Briefly, we used the Knowledge to Action (KTA) Framework [
17] to design and evaluate a multi-component SDM toolkit. This toolkit consisted of a PtDA (
MyDiabetesPlan) and its accompanying implementation tools (such as how-to videos, and enabler cards with step-by-step instructions). It was developed to help prioritize guideline-based disease management in patients with multiple comorbidities, defined as those with type 1 or type 2 diabetes and two additional chronic conditions. At the outset, we engaged a multi-disciplinary research team, including people with diabetes, family physicians, nurses, and content experts with expertise in decision aid development and evaluation, qualitative and quantitative methodology. We also used the Medical Research Council framework for the development and evaluation of complex interventions [
18]; this combined model was particularly relevant to our intervention since it relies on multiple interacting components, behaviours and groups [
19].
Study overview
Development and evaluation of the SDM toolkit consisted of 4 phases: [
1] feasibility testing; [
2] toolkit development; [
3] heuristic evaluation; and [
4] usability testing. Throughout the development process, the toolkit was refined iteratively based on findings of each phase. We used the Consolidated Checklist for Reporting Qualitative Research (COREQ) checklist [
20] (Additional file
1: Table S1) to ensure reporting rigour and the International Patient Decision Aids Standards (IPDAS) checklist [
21] to ensure decision aid development rigour (Additional file
1: IPDAS checklist).
Phase 1: feasibility testing
Individual semi-structured interviews [
22] with clinicians and patients were used to explore their experiences with priority-setting and SDM including their facilitators and barriers, the utility of a PtDA and toolkit for priority-setting, what content to include in the PtDA and how to integrate these into practice. Participants then worked through a prototype of the PtDA while the interviewer provided scripted responses, acting as either the “patient” or “clinician”.
Participants
Using purposive sampling [
23], family physicians, as well as nurses, dietitians, and/or pharmacists (either certified diabetes educators or not) with varied profiles (age, gender) were recruited through family health teams in the academic and community settings in the Greater Toronto Area and surrounding regions.
Patients with type 1 or type 2 diabetes and two other comorbidities including: heart disease (e.g. ischemic, valvular, congestive, arrhythmic, congenital disease), stroke, hypertension, cancer (excluding non-melanoma skin cancer), chronic lung disease, arthritis, inflammatory bowel disorders, and urinary incontinence and with varied socio-demographic profiles (gender, educational attainment) in these family health teams were identified through their clinicians or chart review. Exclusion criteria were: inability to speak English or provide consent, pregnancy or considering pregnancy.
Data collection
Individual interviews with patients and clinicians were conducted using semi-structured interview guides developed by team members with expertise in SDM and qualitative methods. The interview guides were based on pre-existing literature regarding approaches to multiple comorbidities, barriers and facilitators to SDM and goal-setting [
6,
14], and the Theoretical Domains Framework [
24,
25]. We also explored barriers and facilitators to PtDA adoption and preferences for format and content (Additional file
1: Interview Guide).
Participants then used a web-based decision-aid prototype created by the principal investigator with content from evidence-based guidelines [
26] and International Patient Decision Aid Standards checklist [
21]. During this component of the session, the interviewer played the role of either the clinician or patient using scripted responses while the participant used the prototype (e.g. if the participant was a clinician, then the interviewer responded as if she was a person with diabetes, and vice versa). We chose to conduct role play with a prototype to mimic actual clinical use in order to identify facilitators and barriers that may arise with real use, as well as feedback regarding format and content. Face validity of this guide was assessed with team members (patient, nurse, family physician, endocrinologist) and refined as needed; in addition, it was tested in “practice” interviews on other team members. A trained interviewer conducted each interview at the clinician’s practice (when interviewing clinicians), or at our research site (when interviewing patients). All interviews were audiotaped then transcribed verbatim and annotated with field notes.
Analysis
Data analysis was conducted in conjunction with data collection, resulting in iterative refinement of the interview guide. Inductive thematic analytic techniques were used [
27‐
29]. All transcripts were coded for emergent themes [
22], reviewed independently by at least two team members and consensus on coding was reached through discussion during. Regular meetings and documented with memos. NVivo software (v.10) was used to organize and store the data.
Phase 2: MyDiabetesPlan and implementation strategy development
Overview
Building on facilitators of SDM adoption [
14], a goal-setting intervention [
15] and the findings of Phase 1,
MyDiabetesPlan was developed, using the IP-SDM framework [
30] following the International Patient Decision Aids Standards criteria [
21]. Specifically,
MyDiabetesPlan elicits the patient’s general care priorities, identifies his/her diabetes-specific goals and outcomes, outlines diabetes-specific therapies, and details population-specific benefits and risks to therapies. Evidence-based recommendations and patient values uncovered in Phase 1 were used to inform its content. The initial
MyDiabetesPlan was in English and targeted a Grade 8 literacy level [
21]. Although we started with a paper-based prototype, due to the number of required inputs and potential management options, as well as complex weighting algorithms to arrive at the tailored management option based on user input, we elected to use a web-based format for our decision aid. Use of
MyDiabetesPlan within the care team was dependent on the usual roles, responsibilities and processes of care, and the needs of the patient. For example, if the usual process of care was that the patient first saw the clinic nurse followed by the family physician, then this was adapted for the study.
Decision aid prioritization methods
The decision aid framework guided patients through the process of selecting goals (for example, avoiding stroke) and options for achieving those goals (for example, taking preventive medications on time). This framework allowed patients to select goals from an ordered list ranked based on clinical and patient-important factors (for example, wanting to preserve motor ability in order to continue participating in outdoor activities). Each of these factors was assigned a relative weighting based on importance, and goals were listed in descending order (with the highest total weighting appearing first). We only included clinical factors evaluated in risk-prediction algorithms based on large population studies [
31‐
38], and each of these factors was assigned a weighting based on the relative importance to complication prevention. The utility of relative weightings in determining a prioritized goal list was evaluated by content experts and further refined based on the iterative development process.
Options were prioritized based on a similar weighting method, and non-applicable options were removed (e.g. losing weight was eliminated for patients with a normal BMI). The goal (selected by the patient in the previous stage) was used to determine weighting for each of the options, with points allotted to those options, which were included as risk factors identified in the previously mentioned risk-prediction algorithms [
31‐
38]. Weightings for these options were adjusted based on the iterative development process.
The PtDA was reviewed by expert clinicians (family physicians, endocrinologists, geriatricians, nurses, dietitians, pharmacists) and patients not involved in its development. Specifically, each completed a report assessing the accuracy, comprehensiveness, balance of perspective, and ease of understanding of MyDiabetesPlan. Data from these reports were analyzed, and any discordant responses between PtDA reviewers were discussed and resolved by the research team. Based on this feedback, revisions to MyDiabetesPlan were made.
Phase 3: heuristic evaluation
This and the next phase of the study involved usability evaluation of the tool. The United States Food and Drug Administration recommends incorporating “usability engineering processes during the development of medical devices, focusing specifically on the user interface [ …]” . The goal is to ensure that the device user interface has been designed such that the user errors that could cause harm or degrade medical treatment are either eliminated or reduced to the extent possible [
39].
Heuristic evaluation was the first of the usability evaluations undertaken during this study as its objective is to identify weaknesses in the design, especially when use error could lead to harm [
39]. This review can be completed by usability experts, thus providing an opportunity to address major usability issues before the end-users interact with the user interface.
Usability issues were identified, listed, and then categorized by severity as minor, moderate, major, or catastrophic or “show-stoppers” by a human factors engineer; severity estimates were based on frequency, impact, and persistence of errors [
40]. In addition, extensive quality assurance was also conducted by a member of the human factors engineering team using various clinical scenarios to confirm that the program produced the expected result.
The user interface was refined in response to the usability issues that were identified prior to proceeding to the Phase 4.
Phase 4: usability testing
Cognitive task analysis [
41] was conducted in 45-min sessions; users were asked to “think aloud” [
42] as they performed specific tasks to cover the major functionalities of the SDM toolkit (clinician enabler,
MyDiabetesPlan, and patient workbook).
Participants
A total of 11 patient-clinician dyads were invited to participate (3 usability cycles of 4, 4, and 3 dyads respectively). Clinicians were first recruited; each clinician then identified a patient with diabetes and 2 other comorbidities. These patients were invited by the research coordinator to participate in the usability testing. Research has shown that up to 80% of usability issues can be identified through 5 to 8 participants [
41].
Data collection
A research team member with expertise in human factors engineering conducted each session in the primary care setting using the live website, a structured interview guide, and predefined task (completing the MyDiabetesPlan). The following data were documented: navigation choices, errors made, when and where they encountered confusion or frustration, task completion rate, and time spent on the PtDA toolkit or individual tasks within it. Participants were then interviewed regarding satisfaction, strengths and weaknesses of the toolkit, helpful/not helpful/missing content, the quality of decision support, and general comments. All interviews were audio- and video -recorded, and field notes were kept of all sessions as a further source of data.
Analysis
Data analysis was conducted as described in Phase 1.
Throughout Phases 1 to 4, MyDiabetesPlan was refined iteratively by study team members including a graphics designer and computer programmer.
Discussion
Following principles of user-centred design [
47] and adhering to IPDAS criteria [
21], we developed a diabetes-focused, goal-setting PtDA to facilitate SDM by interprofessional teams, as well as a strategy with which to implement it in clinical practice. Working with patients, clinicians, and a human factors engineer, we found that people living with diabetes used diverse approaches to decision-making with a preference for SDM. Dialog and a trusting relationship with their clinicians were vital prerequisites to SDM, which together promoted patient empowerment. They viewed goal-setting as a dynamic process, distinct from goal achievement. Clinicians working with complex patients employed both holistic and disease-specific approaches in order to prioritize concerns through negotiation and compromise and achieve a tailored plan. They expressed frustration when patient priorities were discordant with their own, but sought to bridge this disconnect by further eliciting the patients’ perspectives. Clinicians’ approaches to decision-making were closely tied to their professional identity, ranging from that of an educator to that of a decision-maker. Though the approach they used was tailored to the individual patient, they recognized that the patient had the final say. While they also recognized the critical importance of SDM, clinicians highlighted again the challenge of conflicting clinician and patient agendas, the resolution of which depended on an appropriate knowledge-base on the part of the patient, and active listening on the part of the clinician, all within the context of a longitudinal relationship.
Both patients and clinicians reported “disconnects” between goal-setting and goal achievement, and “discordant” agendas, as challenges to SDM. Our finding regarding the disconnect between
attaining a certain identity (i.e. goal-setting) yet not
performing the necessary actions to get there (i.e. goal achievement) highlights the important objective of any decision-making process – that of reaching a compatible “middle” ground of congruence, mirroring the process described by Image Theory [
48]. Image Theory postulates that individuals base their decision on three images: value image (their principles or “be” goals), trajectory image (their agenda or “do” goals) and finally, strategic image (their action plan). Potential approaches to bridging this gap include incorporating Value Theory-based strategies into the decision-making process such as compatibility testing (screening out incompatible options), followed by probability testing (deliberating between the remaining options) [
49]. We integrated these processes into
MyDiabetesPlan. In addition, we incorporated specific strategies cited by our participants to bridge this gap, such as action planning. This finding validates previous studies [
48] and highlights the critical importance of including a clear action plan as part of the SDM toolkit.
Our finding regarding conflicting patient-clinician agendas, moderated by negotiation and compromise, highlights a fundamental philosophical issue regarding roles of patients and clinicians in decision-making. In SDM, the patients are the authorities on their own experience of living with an illness, the burden of disease on their lives, and how treatment plans may best suit their needs, while the physician is the expert in the evidence-based medicine and evolution of disease [
50]. By its very definition, in the paradigm of SDM, the decision is a shared responsibility of both the patient and the physician, albeit with different responsibilities: one contributing expertise for themselves, and the other contributing expertise for the disease. When conflict arises between these priorities, SDM seeks to resolve this through deliberation to reach a common understanding [
30], echoing the comments of our participants. In this way, the conflict between patient and physician autonomy can be resolved, with the anticipated outcome of patient beneficence. However, what happens in practice? The literature suggests a spectrum of behavior. On one end of the spectrum, physicians report an intent to use SDM, followed by its application in practice. However, they often employed subversive tactics to steer both patient and family towards the decision they perceived as correct [
51]. On the other end of the spectrum, a questionnaire-based study found that nearly half (47%) of patients preferred the clinicians to make the decision without their participation and only 19% wished to share the decision equally with their physician, with 3% wishing to make the decision independently [
52]. A systemic review of patient decision role preferences identified 115 eligible studies, which demonstrated increasing preference for shared decision-making with time: the majority of respondents preferred shared decision-making in 50% of the studies conducted before 2000, increasing to 71% of studies conducted after 2000 [
53]. This is corroborated in a qualitative study exploring 51 patients’ preferences for decision-making approaches, which found that while some patients preferred full engagement in treatment decision-making, others preferred partial or minimal involvement, deferring to their physician due to clinical expertise and a trusting relationship [
54]. Our findings – from both patients and clinicians – reflect this diversity of preference when describing their approach to decision-making, suggesting that a flexibility of decision-making approaches is an important facilitator of patient-centred care; that is, patient-centredness does not mean sharing all decisions, but rather taking into account and responding to the patient’s desire for sharing decision-making [
55]. Together, these findings underline the importance of training clinicians on tailoring to patients’ needs and preferences, presenting information in an unbiased manner, empowering patients with knowledge about their disease, and listening and learning about the patient perspective, in our decision aid in order to bridge potential gaps between patient and clinician agendas.
Our study’s strengths include the systematic development process, as well as rigorous qualitative methodology. We employed evidence-based and user-centered development and implementation, which was facilitated by an interdisciplinary team and early engagement of knowledge-users. The team included expertise in shared decision-making, knowledge translation, information technology, primary diabetes care, and qualitative and quantitative research methods, as well as key knowledge users – primary care providers and people with diabetes. We ensured analytic rigour through the use of at least two coders, interview-by-interview validation, as well as field notes and meeting minutes documenting analytic decisions.
Study limitations included potential volunteer and recruitment bias.
Acknowledgements
We thank our knowledge users Eva Butler, Heather Whetstone, Art Scrannage, Terry and Pauline Wijeyasekara. We thank our design team Joyce Hui, Imran Somji, and Judit Balog. We thank Marc Settino and Amy Hoang-Kim for helping us with study coordination.
IPSDM Team Contributors:
Site leads: Paul Cantarutti, Southlake Academic Family Health Team, Newmarket, Canada; Karen Chu, Bridgepoint Family Heath Team, Toronto, Canada; Paul Frydrych, Mount Dennis Weston Health Centre, Toronto, Canada; Noah Ivers, Family Practice Health Centre -Women’s College Hospital & University of Toronto, Toronto, Canada; David Kaplan, North York Family Health Team, Toronto, Canada; Fok-Han Leung, St. Michael’s Hospital, Toronto, Canada; John Maxted, Markham Family Medicine Teaching Unit, Markham, Canada; Jeremy Rezmovitz, Sunnybrook Academic Family Health Team, Toronto, Canada; Sumeet Sodhi, Toronto Western Family Health Team – University Health Network, Toronto, Canada; Deanna Telner, South East Toronto Family Health Team, Toronto Canada.
Content experts: Joanna Sale, Li Ka Shing Knowledge Institute of St. Michael’s Hospital, Toronto, Canada; Dawn Stacey, School of Nursing - University of Ottawa, Ottawa, Canada; Sharon Straus, Institute of Health Policy, Management and Evaluation - University of Toronto, Toronto, Canada.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.