Primary care practices’ choice of implementation strategy for continuous glucose monitoring for patients with diabetes: a multiple methods study within a larger hybrid type-3 effectiveness-implementation study

Abstract

Background

Most diabetes care occurs in primary care. Continuous glucose monitoring (CGM) is associated with clinical, behavioral, and psychosocial benefits. While CGM uptake in primary care is increasing, understanding models to support CGM use in diverse primary care practices is needed. The PREPARE 4 CGM study evaluated strategies to implement CGM in primary care. We compared characteristics among practices choosing a practice-led, self-paced CGM implementation strategy or referral to a virtual CGM implementation service that provided patients and their referring primary care practices CGM initiation and data interpretation support for at least six months.

Methods

Colorado PC practices interested in implementing CGM enrolled and chose to use the American Academy of Family Physicians Transformation in Practice Series (TIPS): CGM implementation modules or refer patients to a virtual CGM initiation and education service designed and staffed by a primary care multi-disciplinary team. In this multiple methods study, baseline practice characteristics were compared across study arms using chi-square and t-tests. Semi-structured interviews with practice members provided additional context to explain study arm selection.

Results

Of 76 practices enrolled, 46 chose self-paced implementation using TIPS modules, 16 of which (35%) had a diabetes care and education specialist (DCES) in the practice; of the 30 that chose the virtual CGM initiation service, none (0%) had a DCES, X²(1, N = 62) = 11.046, p <.001. Aside from having a DCES, no differences in 37 other practice characteristics were seen between groups.

Conclusions

Primary care practices were eager to implement CGM. All practices with a DCES chose to implement CGM on their own; of the practices without a DCES, implementation method selection was evenly split (half chose to implement on their own, half chose virCIS). DCESs may have potential as diabetes technology champions in primary care practices. Referral to the virtual CGM implementation service allowed access to a certified DCES and multidisciplinary team for practices without them. As many practices without a DCES also chose to implement CGM on their own, multiple models may be necessary to foster CGM implementation in primary care.

Trial registration

This project was reviewed and approved by the Colorado Multiple Institutional Review Board (COMIRB; Protocol 21-4269) and registered with ClinicalTrials.gov on March 23, 2022 (NCT05336214).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12875-025-02903-0.

Introduction

Over 38 million people (11.6% of the U.S. population) had diabetes in 2021, approximately half of whom struggled to meet glycemic targets (47.4% had an A1C value of 7.0% or higher) [1, 2]. Continuous glucose monitoring (CGM) technology is associated with reduced A1C among people with type 1 or type 2 diabetes [3–6]. CGM is also associated with improved diabetes knowledge, diabetes distress, understanding of the impact of diet on diabetes management, and quality of life [7, 8]. CGM uses a sensor inserted subcutaneously for 7 to 15 days to continuously measure interstitial glucose levels, which reduces or eliminates the need for fingerstick glucose checks [9, 10]. In the U.S., CGM use in primary care is growing rapidly [11], but fewer than half (39–44%) of primary care clinicians have ever prescribed CGM [12, 13]. Past prescribing experience predicts favorability toward future prescribing, suggesting that obtaining some experience using CGM with their patients can increase acceptability of this technology for primary care clinicians [13].

The American Diabetes Association (ADA) recommends CGM for people with type 1 or type 2 diabetes [14], yet relatively slow uptake of CGM in primary care limits access to CGM for people who could benefit from it, as in the U.S. 90% of people with type 2 diabetes and 50% of adults with type 1 diabetes are managed by primary care clinicians [15–17]. Uptake of CGM in the U.S. has occurred much more rapidly in endocrinology settings compared to primary care, but these specialty care offices are less accessible to many patients due to travel distance (especially in rural areas, as 75% of counties in the U.S. have no endocrinologists) [18], long wait times, and lack of endocrinologists who participate with patients’ insurance [19]. In a U.S. national study of over 500 primary care clinicians, 41% had experienced challenges referring patients with type 1 diabetes to endocrinology and 45% had experienced challenges referring patients with type 2 diabetes who take both long- and short-acting insulin to endocrinology [13]. The greatest barriers to prescribing CGM that face primary care clinicians are limited insurance coverage from many payers, the resulting high cost of CGM for patients, a great amount of required documentation (e.g. prior authorization) for insurance to approve CGM for patients with coverage, gaps in clinician knowledge and experience, and a lack of access to diabetes resources, support staff, and subspecialists such as Certified Diabetes Care and Education Specialists (certified DCES) [20–24].

Implementation support can help primary care practices address these barriers and bring this evidence-based strategy to more people who could benefit from it [25]. One strategy to support practices to implement evidence-based interventions is practice facilitation. Trained practice facilitators use a variety of approaches including educational tools, audit and feedback, and quality improvement (QI) strategies to support primary care practices’ implementation of evidence-based guidelines and assist practices in building capacity for and implementing organizational changes [26–29]. Specific facilitation strategies include tailoring QI work, guiding practices through the change process and addressing resistance and barriers to change in the practice [29]. Practice facilitator professional backgrounds vary and include previous work as nurses, dieticians, and practice managers [26]. In a systematic review, primary care practices that used practice facilitation were 2.76 times more likely than those without facilitation to adopt evidence-based guidelines [26]. An additional implementation strategy was developed based on findings from a national survey of primary care clinicians which found that 72% of clinicians would be moderately or very likely to prescribe CGM if they had access to CGM education training/workshops and 75% of respondents indicated that they found websites, training modules, or other online resources moderately or very effective as information channels [13]. To address these barriers, authors Oser and Oser worked with the American Academy of Family Physicians (AAFP) to develop AAFP Transformation in Practice Series (TIPS) Continuous Glucose Monitoring (CGM): Enhancing Diabetes Care, Workflows, Education, and Payment. Topics of AAFP TIPS CGM modules include an overview of CGM and ADA Standards of Medical Care in Diabetes, identifying patients who may benefit from CGM, educating patients about CGM, shared decision-making, practice workflow and integration, documentation, and billing [30]. Though evidence is limited, promising strategies for CGM implementation include online team training using AAFP TIPS CGM [30] and practice transformation efforts consisting of specialist support and streamlined workflows [31]. To support practices seeking specialist support for initiating and managing CGM for their patients, our team developed a virtual CGM initiation service (virCIS) led by a certified DCES and two family physicians. DCESs use knowledge, skills, and experience related to diabetes care to provide person-centered, comprehensive care and education to people with diabetes [32]. DCESs consist of a wide range of professionals such as dietitians, nurses, pharmacists, physicians, and behavioral health providers [33]. Certified DCESs, such as the professional on our virCIS implementation strategy team, are a subset of the larger group of DCESs, distinguished by having received formal certification including training and passing a certification examination.

The effectiveness of single or combinations of strategies to implement evidence-based interventions in primary care across different organizational settings is unclear [34]. Understanding how practice characteristics may influence choice of implementation strategy can help healthcare organizations select and tailor approaches to CGM implementation. This study addresses this gap by examining primary care practice characteristics associated with selection of self-paced online education and workflow tools (AAFP TIPS) with the potential for added practice facilitation through study randomization or referral to a certified DCES-led virtual service for CGM implementation.

Methods

This paper presents findings from a larger study with the primary aim of determining the impact of three implementation strategies for CGM in primary care, called PRimary Care Education & Practice Adoption Resource Evaluation for Continuous Glucose Monitoring (PREPARE 4 CGM) [35]. PREPARE 4 CGM is a hybrid type 3 effectiveness-implementation study [36]; it was approved by the Colorado Multiple Institutional Review Board as minimal risk (#21-4269) and is registered with ClinicalTrials.gov (NCT05336214).

Intervention

Practices selected one of two approaches to CGM implementation (their preferred arm): referring patients to a virtual CGM initiation service (virCIS) or clinician and staff education through completion of self-paced online educational modules (AAFP TIPS CGM) [30]. Practices that selected the AAFP TIPS CGM online educational modules for implementation were then randomly assigned to one of two study arms: completion of online educational modules only (TIPS) and completion of online educational modules with support from a practice facilitator (TIPS + Practice Facilitation). Practices in the TIPS + Practice Facilitation study arm received six sessions of practice facilitation over 6 to 12 months. Practices in the virCIS study arm referred patients to the virCIS virtual service for CGM initiation and three CGM interpretation visits over six to nine months conducted via Zoom. VirCIS followed the model of “Primary Care Helping Primary Care” and included a primary care registered dietitian nutritionist (RDN)/certified DCES and two primary care clinical pharmacists, with oversight from two primary care physicians [37]. We used the term DCES in the survey and when referring to survey results, as we intended to capture exposure to a diabetes care and education specialist: any staff on the healthcare team (e.g., nurse, dietitian, pharmacist, others) who specifically provide diabetes education and self-management support, regardless of whether they had obtained certification. This study did not explicitly ask practices to indicate whether their DCES had gone through the process to become a certified DCES.

Participants & setting

Our team used a broad communication strategy to recruit Colorado primary care practices to participate in the PREPARE 4 CGM study. We aimed to recruit diverse practices in terms of size, geographic location, ownership structure, and patient demographics. We used email newsletters distributed though University and professional organization listservs and presentations to practices and health systems to reach large audiences of primary care practices.

Data collection and processing

Practice enrollment took place from April 2022 through February 2023. During this time, a practice manager or clinician completed an application to participate in the study to gather information about practice characteristics, including each practice’s ownership structure (e.g. clinician-owned, hospital- or health system-owned, federally qualified health center [FQHC] or rural health clinic [RHC]), medical specialty (e.g. family medicine, internal medicine), average number of patient visits per week, preferred study arm (self-paced instruction through use of AAFP TIPS CGM with or without PF or referral to virCIS), and practice zip code. During data processing, an indicator variable was created to identify family medicine practices, and visits per week was transformed from a continuous variable to a categorical variable as a proxy for practice size (small, medium, and large) at tertiles (100, 300). Ownership structure was collapsed into three categories: clinician/independent, FQHC/RHC, and hospital- or health system-owned. We used practice zip code alignment with Rural-Urban Commuting Area (RUCA) codes to assess geographic location, assigning practices with ZIP codes corresponding with RUCA codes 1 to 3 as “Metro” and 4 to 10 as “Nonmetro” [38]. After study enrollment, practices completed a baseline practice survey to document number of clinicians and types of staff (e.g. registered nurse [RN], medical assistant [MA], practice manager, DCES), payer mix, patient demographic characteristics, number of patients with diabetes (type 1, type 2), number of patients who currently use CGM, and reasons for study arm selection (e.g. our practice wanted formal education on CGM, our practice does not have time to take part in educational modules). During data processing, indicator variables were created to identify practices with five or more providers, ten or more staff, and endorsement of each staff role. We performed quantitative data collection using REDCap electronic data capture tools [39].

We conducted semi-structured interviews with clinicians and staff from a sample of participating practices near the end of their participation to understand implementation processes, including additional context to understand study arm selection and participation. Qualitative design, data collection, and analysis were led by a PhD-level co-Investigator with more than 10 years of experience conducting qualitative primary care research (TH) with assistance from another study co-Investigator with master’s-level training in anthropology and more than 25 years of experience with qualitative methods (DF). The interview items were developed for this study and have not been previously published elsewhere. Interview guide questions were based on RE-AIM, a framework to understand implementation and impact of evidence-based interventions [40]. We made minor adjustments to the guide throughout the interview process by reordering questions to facilitate flow and removing questions that elicited little useful detail, leaving more time for items that elicited richer detail. For information-rich interviews across multiple types of practices, we used purposive sampling [41, 42] to select practices varying in ownership structure, size, rural/urban geographic location, and study arm for interviews. We conducted multiple rounds of targeted interview recruitment to expand variation in sampled practices, continuing until reaching saturation, when each interview yielded few to no new topics or insights about the implementation process and related outcomes. Interviewers completed memos to document main topics and reflections following each interview, which the team referenced to assess topic saturation and inform further interview sampling. We invited the lead clinician and site contact (usually a clinician or practice manager) to participate in interviews using an initial email from study principal investigators. A study team member (TH) sent up to two additional invitation emails. Participants were asked to identify other clinicians and staff members with experience in CGM implementation in their practice with the goal of speaking with up to two participants per site from a variety of professional roles. Interviews were conducted by two study team members (TH, JC). Interviews were conducted near the end of or shortly following the intervention period between August 2023 and August 2024 and lasted 25 to 55 min. Interviews were audio recorded via Zoom with participants’ permission and professionally transcribed. Participants received a $50 gift card for participating.

Analysis

We used enrollment and practice survey data to describe practice characteristics and examine their relationships with study arm selection. Descriptive statistics (frequency distributions, proportions, means, standard deviations) were used to summarize results. The primary outcome for analysis was study arm selection. Bivariate associations between arm selection and dichotomous or polytomous variables (e.g. ownership type, staffing indicator variables, RUCA classification) were examined using chi-square and Fisher’s exact tests. Bivariate associations with continuous variables (e.g., number of patients with type 1 diabetes, number of patients with type 2 diabetes) were examined using two sample t-tests assuming unequal variance. Associations between dependent variables and arm selection were adjusted for multiple comparisons using the Bonferroni correction and considered significant at < 0.001. All analyses were performed using SAS 9.4 (Cary, NC).

Upon completion of data collection, we loaded interview transcripts into ATLAS.ti software (Version 23, ATLAS.ti Scientific Software Development, GmbH) for qualitative analysis. We used a grounded theory approach incorporating editing and template coding styles to analyze practice member interviews [43, 44]. A priori template codes were based on domains of the RE-AIM framework (e.g. adoption, effectiveness) [40, 45] and specific research questions (e.g. study arm selection). Emergent codes were based on other common topics arising during review of transcripts (e.g. cost challenges, insurance and equipment supplier challenges). For the present analysis, a study team member (TH) used a template organizing style of analysis to organize interview data into main categories of research interest [44], coding all text related to practices’ study arm selection for further review. That team member then repeatedly reviewed all text related to a practice’s study arm selection choice to document and categorize similar reasons. A second analyst (DF) reviewed quotation texts and summaries for confirmation and additional detail. Select interview results related to study arm selection are reported in this publication; broader interview findings will be reported separately.

Results

We exceeded our initial recruitment goal of 60 practices by recruiting 87 within 48 h. Of the 87 practices that applied to the study, 76 enrolled; eight chose not to start due to competing priorities in the practice, staffing challenges, or other inability to support project objectives; and three were deemed ineligible (did not provide primary care, did not serve adults). Over half of enrolled practices (N = 46, 61%) specialized in family medicine only; others included internal medicine (N = 6, 8%), multiple primary care specialties (with a combination of primary care specialties represented) (N = 20, 26%) and Nurse Practitioner (NP)-led primary care (N = 4, 5%). Practices represented a variety of ownership structures, including clinician- or independently-owned (N = 38, 50%), FQHC or RHC (N = 24, 32%), and hospital- or health system-owned (N = 14, 18%).

Of 76 practices enrolled, 46 chose self-paced implementation through AAFP TIPS CGM, 16 of which (35%) had a DCES in the practice; of the 30 that chose virCIS, none (0%) had a DCES, X²(1, N = 62) = 11.046, p <.001. Aside from having a DCES, no differences in 37 other practice characteristics were seen between groups. See Table 1 for practice characteristics and tests of comparison of characteristics across study arms.

Table 1.

Baseline practice characteristics and tests of comparison of characteristics across study arms

Characteristic	N	Overall	TIPS, N = 46	virCIS, N = 30	p-value**
Categorical Variables, n (%)		n	n (% [row-based])	n (% [row-based])
Specialty	76				0.3765
Family Medicine		46	26 (56.5%)	20 (43.5%)
Other		30	20 (66.7%)	10 (33.3%)
Organization Type	76				0.6332
Clinician/Independent		38	25 (65.8%)	13 (34.2%)
FQHC or RHC		24	13 (54.2%)	11 (45.8%)
Hospital- or health system-owned		14	8 (57.1%)	6 (42.9%)
Staff total	62				0.0674
Greater than or equal to 10		47	34 (72.3%)	13 (27.7%)
Less than 10		15	7 (46.67%)	8 (53.3%)
Provider total	63				0.8190
Greater than or equal to five		27	18 (66.7%)	9 (33.3%)
Less than five		36	23 (63.9%)	13 (36.1%)
Clinic reported staff in the following roles	62
Medical assistant (MA)		61	40 (65.6%)	21 (34.4%)	1.0000
Front Desk		57	38 (66.7%)	19 (33.3%)	1.0000
Administrator		47	33 (70.2%)	14 (29.8%)	0.2291
Behavioral health provider		32	21 (65.6%)	11 (34.4%)	0.9310
Registered nurse (RN)		28	19 (67.9%)	9 (32.1%)	0.7942
Care Manager		24	16 (66.7%)	8 (33.3%)	0.9433
Diabetes care and education specialist (DCES)		16	16 (100%)	0 (0%)	0.0003
Pharmacist		13	6 (46.2%)	7 (53.8%)	0.1074
Community Health Worker (CHW)		11	7 (63.6%)	4 (36.4%)	1.0000
Social Worker		11	6 (54.5%)	5 (45.5%)	0.4852
Licensed Professional Nurse (LPN)		10	6 (60%)	4 (40%)	0.7219
Certified Nursing Assistant (CNA)		4	4 (100%)	0 (0%)	0.2900
RUCA Classification	76				0.4532
Metro		47	30 (63.8%)	17 (36.2%)
Nonmetro		29	16 (55.2%)	13 (44.58%)
Visits per Week	76				0.1995
Small (less than 100)		21	11 (52.4%)	10 (47.6%)
Medium (100–299)		28	15 (53.6%)	13 (46.4%)
Large (300+)		27	20 (74.1%)	7 (25.9%)
Continuous Variables, mean (SD)	N	Mean (SD)	Mean (SD)	Mean (SD)	p-value**
Number of Patients with Type 1 Diabetes	62	41.1 (76.72)	49.32 (91.75)	25 (26.6)	0.1226
Number of Patients with Type 2 Diabetes	61	290.87 (357.77)	307.7 (402.3)	258.9 (258.3)	0.5686
Number of Patients with Type 1 Diabetes who use CGM at Baseline	61	13.64 (21.83)	13.46 (21.55)	14.00 (22.96)	0.9291
Number of Patients with Type 2 Diabetes who use CGM at Baseline	62	24.31 (50.44)	28.84 (60.26)	15.48 (19.27)	0.2004
Percent Payer Mix	62
Medicare		22.17 (16.35)	21.56 (17.46)	23.36 (14.25)	0.6851
Medicaid		24.11 (18.88)	26.01 (18.97)	20.39 (18.59)	0.2710
Dual		6.46 (9.37)	7.34 (10.66)	4.76 (5.96)	0.2283
Private		37.54 (22.57)	34.7 (21.38)	43.09 (24.30)	0.1678
Other		2.41 (6.25)	2.3 (4.7)	2.7 (8.7)	0.8488
Uninsured		7.32 (10.08)	8.14 (9.82)	5.73 (10.64)	0.3761
Percent Patient Race
White	62	71.51 (21.40)	71.13 (24.01)	72.25 (15.61)	0.8271
Black	59	5.57 (5.95)	4.72 (5.50)	7.23 (6.57)	0.1258
Indian Native	58	1.31 (2.03)	0.87 (0.96)	2.15 (3.07)	0.0830
Asian	58	4.72 (9.59)	4.71 (11.28)	4.74 (5.28)	0.9885
Native Hawaiian or other Pacific Islander	58	0.65 (1.15)	0.39 (0.52)	1.14 (1.75)	0.0758
Other	58	8.5 (14.23)	9.02 (16.07)	7.51 (10.15)	0.6636
Unknown	58	7.59 (14.25)	9.29 (16.93)	4.37 (5.87)	0.1123
Percent Patient Gender
Female	62	55.43 (8.95)	55.02 (6.94)	56.22 (12.12)	0.6799
Male	62	43.33 (9)	44.2 (7.13)	41.64 (11.86)	0.3716
Other	49	1.24 (5.18)	0.78 (2.07)	2.15 (8.49)	0.4753

* SD: Standard Deviation

** Fisher’s exact, chi-square, or independent t-tests (assuming unequal variance). After adjusting for multiple comparisons, the p-value for significance is 0.0012. BOLD p-values represents statistical significance

Sixty-two practices completed a baseline practice survey (TIPS: N = 41, virCIS: N = 21). In this survey, practices selected from a list of reasons for study arm selection. TIPS practices most commonly selected “We prefer to provide services to patients directly in our practice (versus sending them to an external resource or site)” (N = 26, 63%) or “We felt that the TIPS arm would make it more likely for our practice to sustain CGM use in the practice even after the project/study ends,” (N = 26, 63%). virCIS practices most commonly selected “My practice does not have enough resources (i.e. staff, time) for CGM initiation” (N = 11, 52%). See Table 2 for frequencies for all reasons.

Table 2.

Reason for selecting study arm (items appear in descending order of endorsement rate)

Reasons for Study Arm Choice– TIPS arm	Practices, n (%) (N = 41)*
We prefer to provide services to patients directly in our practice (versus sending them to an external resource or site).	26 (63%)
We felt that the TIPS arm would make it more likely for our practice to sustain CGM use in the practice even after the project/study ends.	26 (63%)
Our practice wanted formal education on CGM.	25 (61%)
We felt that the education/implementation process of the TIPS arm would be better suited to our practice.	21 (51%)
Our practice has worked with practice facilitators on previous projects and found the resource helpful.	11 (27%)
We felt that the TIPS arm would be better because of the size of our practice.	6 (15%)
Other “We think pt would benefit from learning about CGM” (sic.)	1 (2%)
Reasons for study arm choice– virCIS arm	Practices, n (%) (N = 21)*
My practice does not have enough resources (i.e. staff, time) for CGM initiation.	11 (52%)
We felt that the effort required of our practice would be better suited to the virCIS arm.	10 (48%)
We felt that the education/implementation process of the Refer arm would be better suited to our practice.	10 (48%)
We felt that the virCIS arm would make it more likely for our practice to sustain CGM use in the practice even after the project/study ends.	10 (48%)
We felt that the virCIS arm would be better because of the size of our practice.	5 (24%)
Our practice does not have time to take part in educational modules.	4 (19%)
Other: “Administration decision” and blank	2 (10%)

^* Percentages add to more than 100% as respondents could select more than one reason

Interview results

We conducted interviews with 28 clinicians and staff from 21 practices. Interview participants most commonly represented clinician-owned solo or group practices (N = 9, 42.9%) or hospital-owned, health system-owned, or academic health centers (N = 9, 42.9%). Most practices represented metro (vs. non-metro) geographic areas (N = 18, 85.7%). There was balanced representation across study arms (TIPS: N = 7, 33.3%; TIPS + Practice Facilitation: N = 7, 33.3%; virCIS: N = 7, 33.3%). Nearly half of individual interview participants were physicians (N = 13, 46.4%). Other participant roles included NP or Physician Assistant (N = 5, 17.9%), clinical staff (RN, MA) (N = 4, 14.3%), practice managers (N = 2, 7.1%), certified or non-certified DCESs (N = 2, 7.1%), and case managers (N = 2, 7.1%). See Table 3 for interview participant characteristics.

Table 3.

Interview participant characteristics

Individual Participant Characteristics	Participants, n (%) (N = 28)
Role
Physician	13 (46.4)
Nurse Practitioner or Physician Assistant	5 (17.9)
Clinical staff (RN, MA)	4 (14.3)
Practice Manager	2 (7.1)
Diabetes Care and Education Specialist	2 (7.1)
Case manager	2 (7.1)
Practice Characteristics	Practices, n (%) (N = 21)
Organizational Structure
Clinician-owned solo or group practice	9 (42.9)
Hospital/health system-owned and/or academic health center	9 (42.9)
Federally qualified health center	2 (9.5)
Other	1 (4.8)
Number of clinicians
Less than 5	9 (42.9)
5 or more	12 (57.1)
Geographic Region
Metro	18 (85.7)
Nonmetro	3 (14.3)
Study Arm
TIPS	7 (33.3)
TIPS + Practice Facilitation	7 (33.3)
virCIS	7 (33.3)

One of the most common reasons cited in interviews for selecting self-paced instruction using AAFP TIPS CGM was that team members felt that this implementation strategy fit better with their practice because they already had capacity for diabetes education. Clinicians and staff from these practices explained that their existing infrastructure allowed them to integrate patient education on CGM into their workflows. Similarly, many clinicians in the TIPS arms described a desire to treat patients with diabetes in-house rather than referring them out. Several clinicians described this as a desire to be “hands on” with patients. Additionally, clinicians and staff wanted to learn more about CGM and how to incorporate it into their practice and felt that the TIPS study arms would provide that. Practices in the virtual CGM initiation service study arm cited staffing and time constraints as the main reasons for selecting the virtual service over the TIPS arms. These practices described high staff workloads, frequently changing clinical schedules, or frequent staff turnover as barriers to learning about and implementing CGM. Several indicated that it was appealing to have assistance with insurance processes and patient education as part of this study arm. See Table 4 for practice interview quotations illustrating reasons for study arm selection; these responses are based on the interview guide (see Supplementary file 1).

Table 4.

Illustrative quotations from practice interviews describing reasons for study arm selection

Explanation for study arm selection	Study arm	Quotations
Felt that implementation strategy aligned with practice due to existing capacity for diabetes education	TIPS	“We also have a diabetic education team. [We have] some nutritionists, some RNs that actually work with our diabetic patients to provide education and resources, and since we kind of already had that infrastructure set up, I think it made the most sense to try to figure out how to utilize our resources and our teams more efficiently as opposed to referring it out.” -Physician, practice #6
	TIPS	“I think we refer out for a lot of stuff that isn’t in our purview, and diabetes is bread and butter for family medicine, and I feel like we already have relationships with our patients, and I think that we could easily do that ourselves. And so, that also made it so that we learned more about all of this, and that we could keep it in-house rather than further referring it out.” -Physician, practice #7
Staffing and time constraints hindered CGM education and implementation	virCIS	“Part of it has to do with staffing, so as far as the people who handle most of that, the nurse, the MA, and then the front-desk folks are already pretty well maxed out on what they’ve been doing, so being able to have that virtually taken care of that actually helps gets those things rolling quicker for us.” -Physician, practice #10
	virCIS	“One, the [education] is done by people who know what they’re talking about with the patient, and second, obviously, our staff support is always in flux. So, people know that they have stable support to be able to consistently provide that education support and everything that is required by that.” -Physician, practice #17

Discussion

We examined 38 primary care practice characteristics to identify which predicted selection of implementation strategy for CGM, and only one characteristic differed across study arms: presence of a DCES on staff. Practices with a DCES were more likely than those without one to select self-paced CGM implementation using AAFP TIPS CGM online educational modules with or without practice facilitation. Survey responses and clinician and staff interviews bolstered this finding, suggesting that practices with existing diabetes education capacity, such as a DCES, preferred to keep this type of patient care in-house and felt that they could integrate it into their workflow. In contrast, practices without diabetes education infrastructure cited staffing and time constraints as the basis for their selection of the virtual service. Addressing practice context, practice-specific barriers, and perceived resources for change, such as numbers of staff and key practice members, is an important element of intervention tailoring [46, 47]. Previous studies have called for increased attention to the context within implementation settings to promote effective and impactful implementation [48–50]. Finding that practices’ capacity for diabetes education is related to their selection of CGM implementation strategy offers insight for healthcare organizations working to adopt innovative, evidence-based interventions to improve diabetes care. Practices’ level of diabetes education capacity should be considered when designing and tailoring such interventions. Known barriers and facilitators to practices participating in research may also be reflected in the results. Practices with DCES capacity and focus on diabetes care in the clinic may have perceived benefits of increased knowledge, better clinical care, and better teamwork as motivating factors to select the education arm [51, 52]. Importantly, the virCIS option also offered practices an opportunity to participate in research while addressing a known barrier to research participation: insufficient time for research [51].

The association between presence of a DCES and use of online education for CGM implementation suggests that DCESs could have potential as diabetes technology champions in primary care practices. Notably, this study suggests that DCESs, regardless of certification, may influence selection of appropriate implementation strategies for diabetes-related interventions. DCESs may give primary care practices additional capacity for diabetes education to support patients to use CGM. This aligns with previous descriptions of how DCESs can serve as technology champions in their practices [53]. Previous research identified a lack of access to diabetes resources, support staff, and subspecialists as barriers to prescribing CGM in primary care [20]. In the current study, presence of a DCES predicted choice of self-paced CGM implementation through online educational modules, adding to evidence that diabetes resources are a key influence on practices’ ability to implement CGM in practice. Interview and survey data support these interpretations, as practices that selected the virtual arm suggested that staffing and time constraints hindered their ability to provide CGM education in-house. Increasing primary care practice capacity for diabetes education through DCES access or staff training could help increase primary care teams’ comfort with and ability to offer CGM without the need for subspecialists. The recent inclusion of the CGM-derived Glucose Management Indicator (GMI) as a Healthcare Effectiveness Data and Information Set (HEDIS) quality measure by the National Committee for Quality Assurance [54] could potentially lead to increased revenue through value-based payments that could support expansion of DCESs in primary care. Importantly, referral to a virtual CGM initiation service allowed practice teams without formal diabetes resources to collaborate with a certified DCES and multidisciplinary team through a service staffed by primary care. This could be a promising strategy for those primary care practices in which it is infeasible to employ a DCES/certified DCES or other diabetes support.

It is important to note that practices choosing to use the self-paced educational implementation module included many practices without a DCES, even though this characteristic significantly predicted selection of this arm. Multiple models are therefore necessary to foster CGM implementation in primary care.

Strengths & limitations

This study’s multiple strengths include representation from a diverse sample of primary care practices distributed widely across the state of Colorado and both qualitative and quantitative exploration of reasons driving selection of implementation strategies. Further, this study explored numerous characteristics for potential association with study arm selection. However, this study also has several limitations. This analysis of cross-sectional survey data does not allow for causal conclusions. Additional limitations include limited geographical region (only sites in Colorado), relatively small sample size at the practice level (limited power to detect smaller practice-level differences), and that there may be other reasons for study arm choice that our data did not reveal.

Conclusions

To decrease disparities in diabetes, all people with diabetes should have access to evidence-based treatments and technologies regardless of where they obtain care. Primary care practices are diverse, with varied staffing models and resources. Primary care practices’ available resources for diabetes education, such as a DCES, may influence their preference for implementation strategy for new diabetes technologies and interventions. There may be potential for DCESs to serve as diabetes technology champions in primary care practices, which may increase CGM use. Creation of virtual CGM Initiation Services may expand access to resources not available locally in the practice. Primary care organizations should consider practice context related to diabetes resources when selecting strategies to increase CGM implementation as there is not one implementation strategy that will benefit all primary care practices.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

AAFP

American academy of family physicians

CGM

Continuous glucose monitor

CHW

Community health worker

CNA

Certified nursing assistant

DCES

Diabetes care and education specialist

FQHC

Federally qualified health center

GMI

Glucose management indicator

HEDIS

Healthcare effectiveness data and information set

LPN

Licensed professional nurse

MA

Medical assistant

NP

Nurse practitioner

QI

Quality improvement

RDN

Registered dietician nutritionist

RHC

Rural health clinic

RN

Registered nurse

RUCA

Rural-urban commuting area

TIPS

Transformation in practice series

Author contributions

KW analyzed and interpreted quantitative data (survey responses) and contributed to the conceptualization, writing, and revisions of the paper. TH contributed to the research design, provided input in the quantitative analysis, analyzed and interpreted qualitative data (survey and interview responses) and contributed to the conceptualization, writing, and revisions of the paper. MD contributed to the paper conceptualization, research design, data collection, analysis, and interpretation of the data. PD contributed to the paper conceptualization, research design, data analysis, interpretation of the data and substantively revised the paper. DF, BJ, JP, and CS substantively revised the paper. TO was a study principal investigator, conceived of the research design and contributed to the conceptualization, writing, and revisions of the paper. SO was a study principal investigator, conceived of the research design, contributed to the quantitative analysis, and contributed to the conceptualization, writing, and revisions of the paper. All authors read and approved the final manuscript.

Funding

This study was funded by a research grant from the Leona M. and Harry B. Helmsley Charitable Trust, with support for REDCap data systems from NIH/NCATS Colorado CTSA (UL1 TR002535).

Data availability

The datasets used and/or analysed during the current study are available (de-identified) from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This project was reviewed and approved by the Colorado Multiple Institutional Review Board (COMIRB; Protocol 21-4269) and registered with ClinicalTrials.gov (NCT05336214). Individual participants provided written informed consent to participate. This study adhered to the ethical principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

TO and SO have served as Advisory Board Consultants (fees paid to the University of Colorado) for Dexcom, MedScape (more than 12 months ago), and Blue Circle Health. They have received research grants and contracts (through the University of Colorado) from NINR, NIDDK, the Helmsley Charitable Trust, Abbott Diabetes, Dexcom, and Insulet. They do not own stocks in any device or pharmaceutical company. All other authors report none.