A conceptual model of the continuous glucose monitoring integration process for older adults with diabetes developed using participatory systems science methods

Abstract

Aims:

Continuous glucose monitoring (CGM) use remains low in older adults. We aimed to develop a conceptual model of CGM integration among older adults with type 1 and type 2 diabetes.

Methods:

We previously engaged older adults with type 1 diabetes using participatory system science methods to develop a model of the system of factors that shape CGM integration. To validate and expand the model, we conducted semi-structured interviews with 17 older adults with type 1 and type 2 diabetes and 3 caregivers. Vignettes representing each integration phase were used to elicit outcomes and strategies to support CGM use. Data were analyzed using team-based causal loop diagraming.

Results:

The model includes six phases spanning (1) CGM uptake; (2) device set-up; acquisition of (3) belief in oneself to use CGM effectively; (4) belief that CGM is preferable to blood glucose monitoring; (5) belief in future CGM benefits CGM; and (6) development of a sense of reliance on CGM. Causal loop diagrams visualize factors and feedback loops shaping outcomes at each phase. Participants proposed support strategies spanning clinical, educational, and behavioral interventions.

Conclusions:

The model underscores the complex transition of learning new technology and provides opportunities for tailored support for older adults.

1. Introduction

Continuous glucose monitoring (CGM) has become standard of care for people with diabetes who require insulin, including the growing number of older adults living with diabetes [1]. Yet, despite known clinical benefits [1-4] and expanding eligibility criteria for insurance coverage [5], there are known disparities in adoption of CGM, with lower use in older adults compared to younger age groups [6]. Between 2016 and 2021, 27% of US adults over 60 with type 1 diabetes and 3% of adults with type 2 diabetes used CGM [7]. Lower uptake of CGM by older adults reflects, at least in part, that the integration of device use is highly complex with numerous potential barriers. Not only do older adults perceive greater burdens with CGM incorporation [8,9], but there are also known barriers to medical technology use more broadly among older adults, including lower technologic literacy as well as functional, visual, and cognitive impairments [9,10]. When considering new health technology, older adults also may also experience unique challenges disrupting self-management routines built over decades of having diabetes [11].

Effective, relevant, and age-specific interventions to bridge evidence-based diabetes technology to the growing population of older adults with diabetes are needed; yet a patient-oriented and more widely applicable model to describe the process by which diverse populations of older adults integrate CGM into routine management is lacking. Our research team previously engaged older adults living with type 1 diabetes and their caregivers to generate a qualitative model of the hypothesized complex system of factors that interact to shape patterns of CGM utilization and related outcomes [12]. Our objective in the present study was to extend our previous work and develop a model of the CGM integration process and underlying drivers of effective use at each phase that reflects experiences had by a broader population of older adults with type 1 diabetes and type 2 diabetes.

2. Subjects

Inclusion criteria included having type 1 diabetes or type 2 diabetes, being 65 years of age or older, insulin use, familiarity with CGM even if not currently using (ascertained via question “You mentioned that you never used CGM, but I was wondering if you could tell me about why you think someone who has diabetes may decide to use CGM?”), and English proficiency. Given the structure of the interview, we excluded people with significant cognitive or psychiatric conditions, including a clinical diagnosis of dementia. We used a purposive sampling framework to identify individuals with characteristics not well-represented in our first model, perspectives less represented in the literature, and factors likely to lead to variability in CGM integration: diabetes type, race and ethnicity, care setting, CGM use, living situation, caregiver support, and geography (Supplemental table 1). Participants were offered the opportunity to include a caregiver in the research study, which was defined as someone who provided emotional, informational, or practical support to manage diabetes. We adjusted recruitment strategies throughout data collection to meet purposive sampling goals.

We prescreened prospective participants by reviewing electronic health records (EHR) of primary and specialty clinics at the University of North Carolina Health System. We used a two-step recruitment process involving emails and phone calls. Via phone, all participants completed the purposive sampling questions and scheduled an interview (in-person or over Zoom videoconferencing). After the initial scheduling call, all potential participants had a follow-up confirmation call or email to confirm the interview location and answer any questions. All participants received a $50 gift card incentive upon completion of the interview.

3. Materials and methods

3.1. Study design and data sources

This study uses semi-structured interviews to extend a preliminary model of CGM integration developed using data from older adults with type 1 diabetes. In the previous study, we used a participatory systems science method called group model building with 33 older adults with type 1 diabetes and their caregivers recruited from a single academic endocrinology clinic [12]. We used a system mapping technique called causal loop diagraming to develop a qualitative model of the hypothesized complex system of factors that interact to shape patterns of CGM utilization and related outcomes [12,14]. The causal loop diagram functions as a system map to visualize drivers, barriers, and feedback loops (i.e., closed chains of cause-and-effect relationships that reinforce or counteract changes in a system over time) that serve both as potential problems or leverage points for increasing CGM use among older adults with type 1 diabetes [12]. To generate evidence of validity and extend our model to be reflective of a more representative and heterogenous population of older adults with type 1 diabetes and type 2 diabetes, we developed an interview guide to elicit focused feedback on the overall structure of the preliminary model.

3.2. Ethics and IRB

All participants provided informed consent. HIPAA authorization was obtained for patient participants only for data extraction from the EHR system. All study procedures were approved by the Institutional Review Board at the University of North Carolina at Chapel Hill (#22-3293).

3.3. Reflexivity statement

The team consists of people who research, provide care for, and live with diabetes. ARK and LY bring expertise in clinical aspects of CGM therapy. The core analysis team (CS, ACS, KHL, ARK) each bring eight or more years of experience conducting and analyzing qualitative research studies. ARK and KHL previously worked together to adapt group model building, a participatory systems science method, with older adults with type 1 diabetes. [12] ARK and CS have experience in qualitative data collection from older adults with diabetes.

3.4. Interview guide and material development

Interview materials consisted of an interview guide and accompanying visuals to facilitate discussion (Appendix A and B). The visual presentation included a didactic component on systems thinking that was adapted from our previous work with older adults [15] and an overview of earlier research findings, including special emphasis on our preliminary conceptual model of the CGM learning process.

We used vignettes, or stories related to different potential CGM-related outcomes and their underlying feedback loops, to validate and extend the system maps at each phase of CGM integration. Using vignettes in qualitative research offers opportunities for participants to engage with standardized and accessible stories [16]. These vignettes can be especially helpful for structuring conversations on sensitive or value-laden topics (e.g., how one “should” take care of their diabetes), in addition to conversations where participants may have good insight into emotions surrounding a challenge, even if they have not directly experienced it (e.g. difficulties with CGM sensor insertion) [17]. We designed vignettes to reflect the most saturated and influential feedback loops in the preliminary model. We designed vignettes to represent both reinforcing and balancing loops (e.g. an alarm for hypoglycemia wakes an older adult up in the middle of the night and prevents a severe episode during sleep, resulting in a positive experience to promote ongoing use, or conversely, repeating alarms for hyperglycemia contribute to alarm fatigue and diabetes burnout resulting in a negative experience that decreases CGM use over time).

A subset of vignettes was used with each participant. For each, participants were asked about strategies that may help participants troubleshoot the challenges associated with the CGM learning process. For vignettes where the example older adult had a positive experience with CGM, participants were asked about what factors enabled that outcome. In addition to the vignettes, participants were given the opportunity to add any other suggestions or changes to our conceptual model.

3.5. Data collection

Individual self-reported characteristics of patient participants were collected through a purposive sampling questionnaire (Appendix C). Clinical data such as hospitalizations and average HbA1c over the last 6 months were extracted from the electronic medical record system via chart review. For caregivers, relationship with the patient was collected.

All interviews occurred in a private and secure setting, facilitated by a trained interviewer (CS). The conversation was initiated by briefly redescribing the study, answering any questions, and collecting informed consent and verbal HIPAA authorization. Permission to record the conversation was solicited.

After all interviews, the interviewer completed a memo focused on reflexivity and emerging themes, in addition to a saturation checklist (Appendix D and E) to systematically guide the conduct of and content of subsequent interviews and rigorously assess when saturation was reached for each vignette. [18] Saturation was achieved when no new strategies were added to the saturation checklist after three consecutive interviews. This was complemented by regular team meetings to discuss emerging themes, opportunities for improvement, and early adjustments to the conceptual model.

3.6. Analysis

Participant characteristics were summarized using descriptive statistics.

Interviews were transcribed and reviewed by the study team in addition to interview recordings.

There were three components of the analysis: (1) developing a revised conceptual model for the CGM integration progress; (2) using system mapping to diagram the drivers and feedback loops dominant at each phase of CGM integration using causal loop diagramming; and (3) summarizing the strategies participants suggested to promote successful CGM integration among older adults.

The conceptual model was iterated through active diagramming and discussion between CS and ARK, with feedback from ACS and LY, using transcript data analyzed with Atlas.ti software V23.2.0 and Vensim visualizations V9.4.0 for causal loop diagramming. Six system maps representing each phase of CGM integration from the conceptual model were adapted to include new variables, add relationships between variables, and delete or modify concepts that were not dominant overarching themes after considering the new data arising from the phased vignettes (Supplemental Fig. 1). Areas where study participants expressed diverse views and individualized needs relative to other participants were denoted.

The support strategies were tracked as part of the saturation checklist after every interview and transcribed verbatim for additional review (Appendix D). Transcribed data were sorted into each distinct phase and reviewed to ensure all saturated strategies were included.

3.7. Role of the funding source

The funding source had no role in study design, in the collection, analysis, and interpretation of data, in the writing of the report, or in the decision to submit the paper for publication.

4. Results

4.1. Participant and interview characteristics

Twenty total participants completed our study, 17 of whom live with diabetes and three of whom were caregivers (an adult child, adult grandchild, and a friend). Characteristics of patient participants are shown in Table 1. Briefly, ten of the patient participants were female (58.8%), 13 had type 2 diabetes (76.5%), nine identified as African American or Black (52.9%) and eight identified as White (47.1%), two identified as Hispanic or Latino (11.8%), and average age was 72.9 (range 66–81) with seven participants (41.2%) aged 75 years or older. Three lived in a community for older adults (17.6%) and four (23.5%) lived alone. Twelve (70.6%) received their diabetes management through primary care, and nine (52.9%) had diabetes caregivers. Six (35.3%) currently use CGM, one previously used but was not actively using CGM (5.9%), and ten (58.8%) have never used it.

Table 1.

Characteristics of participants living with diabetes (N = 17).

Characteristic, n (%) or mean (SD, range)	Participants living with diabetes
Personal characteristics
Age, years	72.9 (4.2, 66–81)
Participants 75+	7 (41.2%)
Sex
Female	10 (58.8%)
Male	7 (41.2%)
Race
Asian or Pacific Islander	0
Black or African American	9 (52.9%)
Multiracial or biracial	0
Native American or Alaskan Native	0
White	8 (47.1%)
Ethnicity
Hispanic or Latino	2 (11.8%)1
Not Hispanic or Latino	15 (88.2%)
Lives in a rural county	4 (23.5%)
Highest level of education2
High school, GED, or less than 12th grade	3 (17.6%)
Some college but no degree	2 (11.8%)
Associate’s degree	4 (23.5%)
Bachelor’s degree	3 (17.6%)
Master’s, professional, or doctorate degree	4 (23.5%)
Employment status2
Retired	16 (94.1%)
Working	4 (23.5%)
Informal work (volunteering or caretaking)	5 (29.4%)
Living situation
Lives alone	4 (23.5%)
Lives with at least 1 other person	10 (58.8%)
Lives in a community for older adults	3 (17.6%)
Form(s) of health insurance on file3
Medicare (traditional and Medicare Advantage)	17 (100%)
Medicaid	5 (29.4%)
Other supplemental (including Tricare)	5 (29.4%)
No insurance	0
Health characteristics4
Diabetes type
Type 1 diabetes	4 (23.5%)
Type 2 diabetes	13 (76.5%)
Years since living with diagnosis
Type 1 diabetes	27.3 (22.7, 10–60)5
Type 2 diabetes	26.5 (11.4, 8–45)
Use of bolus insulin
Uses bolus insulin	7 (41.2%)
Does not use bolus insulin	9 (52.9%)
Unsure about bolus insulin use	1 (5.9%)
Insulin pump use
Currently uses insulin pump	3 (17.6%)
Has never used insulin pump	14 (82.4%)
Average number of insulin injections per day6	1.9 (1.4, 1–4)
CGM use
Currently uses CGM7	6 (35.3%)
Previously used but not actively using CGM	1 (5.9%)8
Has never used CGM	10 (58.8%)9
Uses a closed-loop system	3 (17.6%)
Primary diabetes management team
Primary care	12 (70.6%)
Endocrinology or specialty care	5 (29.4%)
Has diabetes caregiver(s)
Yes	9 (52.9%)
No	8 (47.1%)
Mean HbA1c over last 6 months10
HbA1c %	7.3% (1.6, 5.4–11.5%)
HbA1c mmol/mol	56.3 (17.2, 35.3–102.2)
History of severe hypoglycemic episode in medical record	1 (5.9%)
Complications from diabetes11
Cardiovascular disease12	8 (47.1%)
Retinopathy	3 (17.6%)
Nephropathy, DKD, CKD secondary to diabetes	9 (52.9%)
Neuropathy secondary to diabetes	4 (23.5%)
Healthcare utilization in last 12 months11
Number of diabetes-specific clinic visits	2.5 (1.8, 0–6)
Any emergency room visits	5 (29.4%)
Any hospitalizations	5 (29.4%)

¹One participant identified as Puerto Rican in addition to Hispanic/Latino.

²Education and work history were missing for one participant (5.9% missing).

³Insurance coverage was verified in the electronic medical record (EMR). Participants could have more than one form of insurance.

⁴Health characteristics are self-reported unless otherwise specified.

⁵Two participants had late-onset Type 1 or Type 1.5 diagnoses.

⁶Among those who do not use insulin pump.

⁷All people who reported CGM use used it all or almost all the time. 6 participants used Dexcom and one person could not remember their device type.

⁸The participant who previously used but was not currently using CGM had lost a piece and was waiting for a replacement. They planned to restart CGM as soon as the equipment arrived.

⁹Two participants described previously trying to get access to CGM through their doctor but were determined to be ineligible at the time.

¹⁰Determined through review of the EMR. If participants did not have a result within the last 6 months, the most recent value was used. One participant did not have updated medical records (5.9% missing).

¹¹Determined by chart review of the EMR. One participant did not have updated medical records (5.9% missing).

¹²Acute coronary syndrome, cerebrovascular accident, congestive heart failure; hypertension excluded.

Eight participants (40%) were interviewed remotely, and 12 (60%) joined in person. Interviews lasted a mean of 72 min (range 36–89 min).

4.2. Conceptual model of the CGM integration process for older adults with diabetes

The conceptual model (Fig. 1) includes six progressive phases of longitudinal CGM integration, including P0) Uptake of CGM: Factors that shape someone’s decision and ability to initially use CGM. Of note, access to CGM supplies remains a relevant driver of CGM use and thus component of the model throughout the entire process. Following the initial Uptake phase of CGM, the integration process progresses through five dynamic phases: P1) CGM Device Set Up Properly: Someone can insert the sensor and transmitter, and display glucose information on nearby device; P2) Belief in oneself to use CGM effectively: Someone comes to believe they are capable of device set-up and interpreting/using CGM data as part of their diabetes management; P3) Belief that CGM is better than fingerstick blood glucose monitoring (BGM): Someone comes to believe that CGM is a better approach relative to previous glucose monitoring approaches; P4) Belief in future benefits from using CGM: Someone comes to believe that CGM will have continued future benefits, and P5) Reliance on CGM: Someone comes to develop a sense of reliance on their CGM for glucose monitoring and broader diabetes management.

Fig. 1.

Conceptual model of the CGM integration process for older adults with diabetes.

4.3. Phase-specific system maps

System maps to visualize the factors, relationships, and feedback loops that underlie each the key phases are shown in Fig. 2. A description of major changes made to the original conceptual model and supporting quotes can be found in Supplemental table 2. The subset of topics and vignettes where study participants reported significant heterogeneity in views on key variables, their relationships, and feedback loops are shown in Supplemental table 3. These include the implications of “losing privacy” in diabetes self-management due to CGM alarms and sharing glucose data with caregivers. Three reinforcing loops occurred between device set-up (P1) and belief in oneself (P2); belief that CGM is better than BGM (P3) and belief in future benefits from CGM (P4); and reliance on CGM (P5) with long-term CGM use, demonstrating how positive or negative experiences in one phase can reinforce progress in another. The system map of P0 (Uptake), denotes many factors that influence CGM uptake; access to CGM supplies remains relevant throughout the process and feeds into each phase.

Fig. 2.

Phase-specific system maps.

4.4. Participant-proposed, phased strategies to support CGM integration

Strategies to support the CGM integration process are shown in Fig. 3. Throughout all phases, participants recommended encouragement and emotional support from clinical teams and loved ones. They also highlighted peers with diabetes as an important resource for education and support.

Fig. 3.

Participant-proposed, phased strategies to support CGM integration.

5. Discussion

Our study provides a patient-oriented conceptual model for the process of integrating CGM into both type 1 diabetes and type 2 diabetes management as an older adult, reflecting patient and caregiver perspectives collected over two successive participatory system science studies. The model delineates six progressive phases that span the longitudinal acquisition of new technologic and behavioral skills, while the system maps at each phase visualize the subset of factors and feedback loops that are important for shaping outcomes. Additionally, for each phase, older adults suggested strategies to support their success. As the health benefits of CGM use in older adults continue to grow along with the expanding insurance coverage of CGM making this tool more accessible [1,5], there is a growing need to ensure that older adults are able to effectively integrate CGM into routine diabetes management. We employed intentional efforts to build a model more generalizable to the population of older adults with diabetes for whom CGM may be clinically indicated through the use of purposive sampling . Our study collected data from older adults with and without an identifiable caregiver, and with diversity across a range of other characteristics (i.e., current CGM use, diabetes type, race and ethnicity).

A recent systematic review of qualitative studies on patient experiences with CGM found several overarching themes: “gaining control and convenience,” “motivating self-management,” “providing reassurance and freedom,” “developing confidence,” “burdened with device complexities,” and “excluded by barriers to access” [19]. Our model incorporates nearly all of these themes and provides strategies to address them, while also adapting them to an older adult population. Furthermore, our model illustrates that while the CGM integration process may proceed through the same six major phases influenced by the themes described above, an individual’s specific journey will be shaped by different factors and feedback loops. The longitudinal model also illuminates ways in the current model for CGM training and education may not provide full support for older adults to integrate new technology, specifically as they encounter technical or emotional challenges in weeks and months following initial training. Our data suggest that setting up the device and instilling initial user confidence—P1 and P2—are potentially vulnerable phases when participants may decide to stop using the device if they began having negative experiences early in their trajectory of CGM integration. Participants who described their own experiences moving into P3 and P4 (or beyond) highlighted they would continue using CGM unless they found it to become untrustworthy (impacting P3, or the relative usefulness as compared to prior technology) or if it became too expensive, limiting access.

Our model highlights the role for several foundational concepts in health behavior and technology adoption, including technologic literacy self-efficacy [20,21]. For example, P2 centers on building a belief in oneself to use CGM successfully, where accumulating self-confidence and diabetes management self-efficacy interacts in a reinforcing feedback loop with P1 (CGM device set-up) to help propel older adults through the integration process. P3 and P4, which focus on comparisons to use of manual blood glucose monitoring and belief in future benefits from CGM, demonstrate how an increasing perceived usefulness of the new technology—a key component of the Technologic Acceptance Model [22]—can build upon one another and propel long-term CGM use. Since there are known inequities not only in rates and complications of diabetes across race and ethnicity [23], income level [24], and rurality [24], but also in CGM uptake [25] and health technology literacy [26], it is critical to engage older adults of diverse backgrounds to develop a tool that can be reflected across social and clinical context. By including supply and access-related factors that continuously impact CGM uptake and use, our system map highlights the critical importance of attention to equitable CGM access among older adults throughout all stages of integration. Addressing these factors will require engagement beyond clinical, social, and behavioral approaches to address additional systems-level barriers.

In terms of the strategies proposed by older adults, participants proposed phase-specific interventions strategies spanning clinical, educational, and behavioral domains. Implementing these supports will require coordination across different groups, including healthcare workers, social support from friends, family, and peers with diabetes, CGM manufacturers, and insurance/Medicare. There are limited studies that have tested such multilevel interventions for CGM use in older adults. A ten-site quality improvement project increased CGM use among young adults by engaging with multiple stakeholders through patient coaching, clinical staff education, and coordination between vendors and payers [27]. This type of strategic approach that gains buy-in from multiple parties—while also adapting to specific institutional and individual needs—may be used as a model for how to implement the strategies recommended by older adults in the future, informed by feedback from older adults throughout the implementation process.

In our analysis, we noted where older adults may differ in their perspectives on CGM use and may benefit from tailored interventions at each phase. Similar to other studies, we found a wide range of comfort related to diabetes disclosure [28,29], including scenarios in which people were upset by public alarms and not interested in sharing their health data with caregivers. Discussions during early phases of the integration process about data sharing and alarm settings can help older adults determine how they would best like to customize their CGM experience. Respecting that participants may have vastly different perspectives on the degree of privacy they hope to achieve with their diabetes—as well as recognizing the reality of CGM alarm fatigue [30]—is foundational to tailoring support and interventions that support participants’ overall quality of life and long-term health. Additionally, developing personalized alarms to support individualized glycemic management targets [31] may improve experience with the device. Previous research suggests that CGM alarm thresholds have impacts on overall glycemic outcomes, underscoring the importance of setting up the CGM in an individualized manner [32]. Linking older adults to resources that can help them customize their CGM to align with management goals, such as online tutorials or in-office support, was shared as a strategy that could help with the integration process.

Future research should test participant-provided strategies and patient-centered interventions that acknowledge the effects of individual characteristics and social influence, as well as test the appropriate timing of interventions. Development and testing of specific constructs put forth by the conceptual model is critical to enable implementation and evaluation. Research that identifies high yield interventions (e.g., have large impacts or help a high proportion of patients)—while also paying close attention to heterogeneity, personalization, and equity considerations within those learning to use CGM—will help the ideas our participants shared with us become reality. Finally, future studies may utilize this approach to facilitate access to and integration of CGM in low-resource settings; similar models may elucidate opportunities to increase uptake and tailor support strategies for device integration based on local context and in partnership with community members.

Our study includes several limitations. We did not include participants with dementia or other cognitive limitations, and future work is needed to provide evidence of validity among much older adults (e.g. >80), Asian and Hispanic/Latino participants, and those in skilled nursing facilities and long-term care settings. Our sample is from one healthcare system in a single geographic region. We also did not collect data on technologic and health literacy. Strengths of the study include the use of a rigorous, qualitative approach bridging methods from the fields of system science with thematic analysis methods represent an innovative way to elicit insight from participants that honors the complexity of lived experiences while simultaneously structuring insights into actionable steps. One of the greatest assets of this study is that the proposed strategies are actionable and timed to appropriate phases, meaning they can be adapted for older adults at different phases of the integration process. The importance of this conceptual model will only continue to grow as diabetes technology evolves and the population of older adults with DM continues to expand.