Addressing the gaps in diabetes care in first nations communities with the reorganizing the approach to diabetes through the application of registries (RADAR): the project protocol

The RADAR model incorporates two distinct elements: 1) an EHR, integrated with a population-based diabetes registry and analytics platform – CARE platform; and, 2) a centralized care coordinator. The CARE platform was designed by OKAKI Health Intelligence Inc. to meet the specific requirements of nursing-driven health programs currently being delivered on reserve in Canadian First Nations communities. Major limitations of existing EHRs, which are largely designed for physician offices, are: misalignment with programs, reporting and professional practice needs of service providers on reserve; no population-based registry that supports sharing between facilities and population (as well as individual) measurement and chronic disease management; and cost [18, 19].

The first element of RADAR, the CARE platform, specifically integrates a patient registry and electronic clinical chart for patients across three traditionally distinct and segregated programs responsible for care on reserves (Home and Community Care, Aboriginal Diabetes Initiative, and Community Health); thereby, improving coordination of care between service providers on reserve. The CARE platform includes a diabetes registry, enabling the systematic, standardized capture of patient data on key diabetes outcomes and quality of care indicators (e.g., A1c, BP, cholesterol, foot and eye exams); and incorporates performance-based measurement and reporting on clinical care standards and targets at both the population and individual level. The system also provides real-time recognition and recall of patients needing follow-up care. Moreover, the CARE platform uses a simple interface and can be accessed through a web-based portal allowing for remote monitoring and evaluation of care and reporting.

In our ongoing discussions with community leaders, most diabetes care tends to be reactive to clients seeking care rather than proactively in accordance with clinical practice guidelines. Many clients do not receive ongoing diabetes care and do not have a regular general practitioner who coordinates care. Thus, we have combined our CARE platform with a key facilitator – a centralized care coordinator (CC). The CC is ideally a First Nations registered nurse and certified diabetes educator with significant experience with care on reserve. The CC will work remotely with the local healthcare workers on reserve to improve quality of care for people with diabetes, facilitated through shared use of the CARE platform and web conferencing technologies. The CC supports systematic review of patients and case conferencing activities to support and prioritize care, problem-solves system or patient-level care issues, and provides education to local healthcare workers. In addition, the CC acts as a bridge between the First Nations communities and other service providers to help coordinate patient care on and off reserve. Together, both RADAR elements address the 5Rs in organizing diabetes care, that is, recognize, register, relay, recall, and resource, in the First Nations context. In addition, managers and researchers can use the analysis of population-level data for epidemiological assessments, quality-improvement, and decision-making.

RADAR involves collaboration with several First Nations communities in Alberta, representing separate treaty areas across the province (continuous expansion of the RADAR model to additional First Nations communities is ongoing). We estimate that approximately 1000 people with type-2 diabetes currently reside in the pilot communities. All communities involved are several hours of travel from a major urban center, where the majority of diabetes specialist care services are located. Some communities have local family physician clinics, located just off reserve that also provide care for patients with diabetes; other communities have visiting family physicians on one or more days a week; while other communities have no local physician services, although services are available in neighboring rural communities. As a result, local, on-reserve, healthcare workers are the primary service providers delivering diabetes care and general health services for these populations. Prior to implementing RADAR, all communities were familiar with the CARE platform for the management of patient care within their respective communities.

The primary outcome will be a 10% improvement in measures over the baseline control periods in any one of the following: A1c, BP, cholesterol. This is consistent with our previous Canadian work that considers a 10% improvement in any of these measures to be the minimally clinically important difference [20]. For patients who are already at ‘target’ for any or all three measures at baseline, we will consider persistence at target (i.e., maintain values within 10% of baseline) as also achieving the primary outcome. Secondary endpoints will address other areas of diabetes care and process indicators compared to baseline including: the proportion of clinical measures and tasks completed in accordance with Canadian Diabetes Association Clinical Practice guidelines (e.g., A1c, BP, cholesterol, foot and eye examination, receipt of vaccinations, smoking cessation counseling); the number of patients registered in CARE; and the proportion of patients linked to a health services provider. In addition, we are collecting detailed information to conduct a comprehensive economic evaluation, assessing the cost-effectiveness of RADAR specific to these communities. Our approach to the economic analysis will be analogous to methods previously published by our group [21, 22].

We will also conduct a qualitative assessment as part of our comprehensive evaluation to supplement the primary and secondary quantitative assessments. The purpose of the qualitative component is to provide contextual information, such as the quality/usability of the CARE platform and the impact/satisfaction of local healthcare workers with the CC model.

Participation in RADAR is dictated by the Health Director and leadership within each First Nations community health facility. After agreeing to participate, local healthcare workers (e.g., Health Center, Healing Center) will actively register First Nations patients with diabetes into the CARE platform. All patients in the community are eligible to be included in the CARE platform; however, patients must meet the following criteria to be evaluated within RADAR:

The implementation and evaluation of RADAR is segregated into two distinct parts. OKAKI Health Intelligence Inc., who has a long history of working with First Nations communities in the province, will implement RADAR, including the CARE platform and care coordination. Researchers from the University of Alberta will independently evaluate the effectiveness of RADAR. By separating the implementation from the evaluation, we believe a more robust evaluation will be achieved. To ensure a robust evaluation, a modified stepped wedge controlled implementation trial design will be employed. This design is a highly valid cross-over design, whereby each community serves as its own control before ‘switching’ to the intervention at different time points [23, 24]. In essence, RADAR will be implemented sequentially to the communities in consecutive 4–6 month periods, based on the communities ‘readiness’. By the end of the evaluation period, RADAR will have been implemented in all communities (Fig. 1).

After completion of the Baseline Phase, each community will transition into the intervention phase for a minimum of 1 year. The intervention phase will conclude on December 31, 2020, for all communities and length of follow-up is dependent on timing of initiation of RADAR within the community. During the intervention phase, the CC will help the local healthcare workers to proactively manage patients with type-2 diabetes, including monitoring and recall of patients, relaying clinical information and coordinating care, facilitated through shared use of the CARE platform and web conferencing technologies. In addition, the CC will provide support on the use of the CARE platform and training and education for the local healthcare workers on ‘best practices’ for the management of type-2 diabetes. It is expected that the local healthcare workers’ reliance on the CC will diminish over time, as they become more confident and competent using the CARE platform and in their diabetes knowledge and ability to manage patients, leading to sustainability.

We will conduct a comprehensive evaluation of RADAR, using the RE-AIM framework [25]. We have used this framework to evaluate other diabetes-related, quality-improvement interventions [26]. RE-AIM consists of five dimensions: Reach into the target population; Effectiveness of the intervention; Adoption by targeted end-users; Implementation, including consistency and cost of delivery; and Maintenance of intervention effects over time [25] Table 1.

Thus, our comprehensive evaluation will include both quantitative and qualitative components. We will use the quantitative data for the clinical outcomes mentioned above as part of our overall assessment of the dimension of effectiveness. The primary source of qualitative data includes interviews during the baseline, intervention and post intervention phases. We will use purposeful sampling to identify key informants with expert knowledge of the implementation of RADAR in each community (e.g. the CC, local healthcare workers, and health directors). All interviews will take place at community health facilities or via videoconferencing, using semi-structured interview guides. The purpose of the interviews is to document the organization of existing type-2 diabetes care within each community, implementation facilitators, barriers, and lessons learned, perceived appropriateness of RADAR for the communities, and the overall usability and acceptance of the technology (i.e., CARE platform). In addition, the software vender will also capture changes made to the software over time as requested by the communities. The interviews will be digitally recorded for subsequent analysis, transcribed verbatim by an independent transcriptionist, and verified for accuracy. All qualitative data sources (e.g. validated baseline assessments and interview transcripts) will be compiled and managed using Atlas.ti 7 software.

For quantitative data, all relative changes will be analyzed using the last known recorded value from the CARE platform from baseline to the end of the intervention phase. We will employ an intention-to-treat analysis, whereby all patients who are enrolled and have baseline data will be included in our primary analysis. The primary endpoint (i.e., 10% improvement in any of A1c, BP, cholesterol values from baseline or those maintaining values within 10% of baseline if at any or all three targets at baseline) will be evaluated using a conventional intention to treat analysis with logistic regression. However, if the observed intracluster correlation coefficient (ICC) is non-ignorable, we will use generalized estimating equations (GEE) that accounts for the clustered and hierarchical nature of our data [24]. Several a-priori subgroup analyses are planned for the primary endpoint. First, we will restrict analyses to patients not at target for any of the 3 primary outcome measures. Second, we will restrict analyses to those patients who were at target for each primary outcome measure (A1c, BP, cholesterol) to fully examine persistence at target over time. Additional subgroups of interest include analyses stratified by sex, and analyses stratified by age (<65 vs > =65 years). Secondary endpoints include: the proportion of clinical measures completed in accordance with guidelines (e.g., A1c, BP, cholesterol, foot and eye examination, receipt of vaccinations, smoking cessation counseling); the number of patients registered in CARE; and the proportion of patients linked to a health services provider.

We anticipate not having complete quantitative data on all endpoints of interest, given the real-world implementation of the project. Several approaches will be used for missing clinical data. First, in alignment with the intention-to-treat approach, we will use the baseline observation carried forward technique for our primary outcome. The stability of the results will be additionally evaluated using the multiple imputation method as a sensitivity analysis.

For qualitative data analysis, we will use an integrated approach. First, we will use the RE-AIM framework as the initial coding structure. Second, we will use an inductive approach [27] to identify emerging codes and concepts within each dimension using conventional approaches [28‐30]. We will employ several verification strategies to ensure the reliability and validity of the qualitative results including methodological coherence, appropriate sampling, collecting and analyzing data concurrently [31], and debriefing [30].

In prior studies, about 40% of diabetes patients will have at least a 10% improvement in A1c, BP, and cholesterol over time when studied under ‘usual care’[32]. To detect at least an additional 10% improvement in the proportion of patients achieving our primary endpoint under the RADAR model (i.e., move the population from 40% achieving endpoint to at least 50% achieving endpoint), at an alpha =0.05, an ICC of 0.01, we require 1153 total patients to detect a clinically important difference with 80% power. Assuming a 10% loss to follow-up, we require approximately 1250 total patients among our communities.