Comprehensive Telehealth Model to Support Diabetes Self-Management

This cohort study examines glycemic outcome data for adults with type 1 or type 2 diabetes who participated in a virtual endocrinology clinic that supported continuous glucose monitoring and provided education and behavioral health support.


Introduction
There are approximately 37.3 million individuals with diabetes in the United States, and the number continues to increase. 1 Despite this increasing number, many patients do not have ready access to endocrinology care.In 2022, the number of endocrinologists in the United States was estimated at 8524 (a ratio of individuals with diabetes or other endocrine disorders to endocrinologist of 4375:1). 2 The great majority of diabetes care takes place in the primary care setting. 3][6][7][8][9][10] Despite evidence showing clinical benefits, CGM use is not yet widespread, particularly for patients whose diabetes is managed in a primary care setting.Despite primary care clinicians' interest in CGM use for their patients with diabetes, limited knowledge, technology readiness, and lack of resources are limiting factors in diabetes technology implementation and expansion. 11Navigating insurance coverage remains a key barrier, as well as access to CGM training support and education to improve understanding of data.
Telehealth offers clinicians caring for patients with diabetes an opportunity to use CGM technology to remotely monitor glucose levels and make shared decisions with patients about therapy modification without the need for an in-person visit. 12Evidence of the feasibility and effectiveness of remote CGM initiation through telehealth as a means of expanding access was reported in a pilot study of 34 adults with type 1 or type 2 diabetes, where reduction of HbA 1c value, increase in TIR, decrease in diabetes distress, and increased satisfaction with glucose monitoring were observed after 12 weeks of CGM use. 13Similar benefits were reported in the ONBOARD study in adults with type 1 diabetes already on CGM using a multicomponent telehealth intervention 14 and in the Onduo Virtual Diabetes Clinic for adults with type 2 diabetes where participants were evaluated remotely by an endocrinologist and provided CGM for intermittent use without in-office training. 15owing evidence of the potential for successful remote diabetes care and CGM technology use prompted this study to evaluate a model that could minimize adoption-limiting factors such as geography and access to specialty care for adults with type 1 or type 2 diabetes.The Virtual Diabetes Specialty Clinic (VDiSC) study was designed to assess the feasibility and efficacy of establishing a virtual endocrinology clinic to facilitate comprehensive diabetes care, support CGM integration into diabetes self-management, and provide behavioral health support for diabetes-related issues.
Decision support technology use was also evaluated within this virtual clinic model.

Methods
The study was conducted virtually from August 24, 2020, to May 26, 2022.Participants were recruited through an insurance company (CVS Health Clinical Trial Services) and through primary care and endocrinology clinic referrals from 2020 through 2021 and were followed up for as long as 26 weeks.Potentially eligible participants received a letter, email, and/or flyer with study information,

JAMA Network Open | Diabetes and Endocrinology
Comprehensive Telehealth Model to Support Diabetes Self-Management Eligible participants were US residents 18 years and older with type 1 or type 2 diabetes who used either an insulin pump or multiple daily injections of insulin.Participants were required to have a smartphone and access to a computer with internet that could be used for virtual visits.Exclusions included use of an automated insulin delivery system, pregnancy, or kidney dialysis.Participants were eligible if they were either not currently using CGM or if they were using CGM and their TIR was less than 60% or their time with values lower than 54 mg/dL was more than 1%.

Study Procedures
Enrolled participants who were not currently using CGM completed an initial phase with a blinded This clinical decision support was previously validated in patients with type 1 diabetes using insulin pumps and CGM, 16 but final recommendations to participants were determined by the CDCES in consultation with the study endocrinologist when needed.
A capillary blood collection kit was sent to participants at baseline and at 12 and 26 weeks for fingerstick blood sample collection returned by mail to the Advanced Research and Diagnostic Laboratory (University of Minnesota) for HbA 1c measurement using the Tosch Automated Glycohemoglobin Analyzer HLC-723G8.Accuracy of this procedure for capillary blood sample collection and HbA 1c measurement is comparable with that of venous blood HbA 1c measurements.weeks, and 26 weeks; results will be separately analyzed and reported.
Participants were asked monthly to report whether they experienced severe hypoglycemia (defined as an event that required assistance from another person because of altered consciousness), diabetic ketoacidosis, and hospitalizations.Available CGM glucose values were assessed to confirm self-reported severe hypoglycemia events, and the medical monitor reviewed the self-reported event description.

Outcome Measures
Efficacy outcomes included CGM use; CGM metrics for hypoglycemia (<54 and <70 mg/dL), hyperglycemia (>180 and >250 mg/dL), and TIR for 70 to 180 mg/dL; mean glucose level, and glycemic variability (coefficient of variation), HbA 1c value, and participant-reported outcomes, including psychosocial and diabetes treatment satisfaction questionnaires.Safety outcomes included severe hypoglycemia, diabetic ketoacidosis, hospitalizations, and emergency department visits.

Statistical Analysis
To be included in the analyses, it was necessary for a participant to initiate real-time (unblinded) CGM

Results
Among the 234 participants included in the analyses, 160 had type 1 diabetes and 74 had type 2 diabetes.The mean (SD) age was 47 ( 14) years, 123 (53%) were female, and median diabetes duration was 20 years.The 234 participants included in the analyses were a subset of the 341 who signed the study consent and were screened; 75 were ineligible, 31 dropped out before initiating the study, and 1 dropped out immediately after the CDCES sessions without obtaining any CGM data (eFigure 1 in Supplement 1).

JAMA Network Open | Diabetes and Endocrinology
Comprehensive Telehealth Model to Support Diabetes Self-Management b Two participants were missing baseline CGM data.
c For metrics with a skewed distribution, robust means using an M-estimator were calculated to downweight outliers.
d Target range was 70 to 180 mg/dL.
e The model includes all available data, imputations were done when mean CGM glucose values during the previous month were available using the glucose management indicator formula, 18 still missing data were handled using direct-likelihood method, and P value is for 6-month values vs baseline.

JAMA Network Open | Diabetes and Endocrinology
Comprehensive Telehealth Model to Support Diabetes Self-Management (56%) with at least 1 mental health visit; 82 (92%) had an initial visit following a positive alert on a screening questionnaire, while 7 were self-referred.Overall, there were 298 mental health visits.
On the 6-month questionnaire, nearly all participants (146 of 147) reported that CGM use helped them manage their diabetes better.A majority either strongly agreed (71%) or agreed (22%) that working with the virtual clinic helped with diabetes management (eTable 3 in Supplement 1).
Twenty-three participants (14%) self-reported a total of 30 adverse events, including 16 severe hypoglycemia events in 15 participants (9%), 1 of whom was diagnosed with COVID-19 and subsequently hospitalized, and 13 other hospitalizations in 10 participants (6%).As hospitalizations were self-reported, lack of hospital records precluded the ability to determine if any hospitalizations also represented diabetic ketoacidosis.The rates of severe hypoglycemia and hospitalization events were 18 and 15 events per 100 person-years, respectively.
The median (IQR) number of follow-up contacts per participant after completion of the 3 remote training sessions was 8 (6-11).Follow-up contact formats included video training sessions (74%), telephone calls (13%), emails (8%), and texts (5%).There were 32 participants (43%) with at least 1 mental health visit; all were referred for an initial visit following a positive alert on a screening questionnaire.Overall, there were 74 mental health visits.
All 61 participants who responded to the 6-month questionnaire reported that CGM use helped them manage their diabetes better and either strongly agreed (85%) or agreed (15%) that working with the virtual clinic helped with diabetes management (eTable 3 in Supplement 1 Several models of virtual diabetes care have been developed and described in the last few years, in part related to the COVID-19 pandemic creating the need to establish remote management options for chronic diseases.Successful remote management of type 1 diabetes, which in 52% of the patients included CGM initiation, was described by Eilan and Drincic, 19 where endocrinologist-based telehealth visits to rural clinics showed decreases in HbA 1c values.Even though this model showed success in improving type 1 diabetes outcomes, it still required a rural clinic location where individuals were remotely evaluated and treated.
Another virtual model, the Onduo Virtual Diabetes Clinic, is designed to support diabetes management in the primary care setting through interim endocrinology visits for individuals with type 2 diabetes. 15,20,21Features include a mobile app that connects wirelessly to glucose meters and CGM devices and allows for medication tracking, remote lifestyle coaching, and clinical support.
Preliminary data for 740 participants suggested improvement in HbA 1c values. 20Similarly, a subsequent prospective single-arm study with 55 participants with type 2 diabetes and intermittent CGM use over 4 months also showed a decrease in HbA 1c values of 1.2% from a baseline of 8.9% and an increase in TIR of 10.2% from a baseline of 65.4%. 21ile intermittent use of CGM has been shown to have benefits, 22 daily use of CGM provides the greatest benefits. 10,23Another virtual diabetes care model, Steady Health, 24 incorporates CGM and a multidisciplinary approach with person-centered diabetes care.Their real-world retrospective case series explored early glycemic outcomes of participants with TIR less than 70% from baseline CGM data.Fifty-three participants met criteria for analysis, and an increase in median (IQR) TIR of 16.6% (6.0%-27.9%)was observed when comparing the last available 4 weeks of CGM data to baseline (mean duration of care, 11 months).There was an HbA 1c reduction of 1.2% in the 27 participants with data for both baseline and follow-up HbA 1c .While these data lend support for the benefits of potential virtual diabetes care, data were retrospective, and all participants were already on CGM when they entered the study.

JAMA Network Open | Diabetes and Endocrinology
Comprehensive Telehealth Model to Support Diabetes Self-Management outcomes and CGM metrics, that also offered expanded specialty care for diabetes-related issues.
Although it is always recommended that patients with diabetes have annual in-person evaluations with their primary clinicians to assess possible long-term complications, such as foot and eye examinations, this virtual model has the potential to expand access to care, reducing health disparities for patients unable to access diabetes care in person, and to facilitate the adoption of technologies and support targeted diabetes self-management education.This study model was designed before the COVID-19 pandemic, but the results demonstrated how diabetes telehealth services, made more readily available during the pandemic, can be successful and should be allowed to continue.Longer-duration studies and cost analyses are needed to support expansion of virtual care models.
use, complete at least 1 training session, and have at least 168 hours of CGM data during follow-up.A linear mixed model with a random-participant effect was used to test mean change from baseline to follow-up.For metrics with a skewed distribution, robust means were calculated using an M-estimator to down-weight outliers.Descriptive statistics included means with SDs or medians with IQRs depending on the distribution of data.Missing HbA 1c values were imputed using the Glucose Management Indicator (GMI) formula18 to convert CGM-measured mean glucose into an HbA 1c equivalent (n = 43 at 6 months).Missing HbA 1c values without a corresponding CGM mean glucose available (n = 8 at 6 months) and questionnaire values at baseline, 3 months, and 6 months were handled using direct likelihood in the mixed model.For the glycemic outcomes, 2 participants missing baseline CGM data were excluded from the pre/post comparison.A mixed model was used to assess potential risk factors for the change in HbA 1c value from baseline to 6 months.The false discovery rate was used to adjust for multiple comparisons.P < .05 was used to define statistical significance.Analyses were performed separately by diabetes type using SAS version 9.4 (SAS Institute).

Figure .
Figure.Boxplots of Hemoglobin A 1c (HbA 1c ) Values for Study Participants With 6 Months of Follow-Up

Table 2 ,
Figure, A, and eFigure 2A in Supplement 1).These improvements were consistent across levels of diabetes type, prior CGM use, diabetes duration, presence of a chronic health condition, self-reported amount of exercise, and healthy eating (eTable 1

Table 2 .
Glycemic Results by Diabetes Type a P values were adjusted for multiple comparisons using the 2-stage Benjamini-Hochberg adaptive false discovery rate procedure.
years, 31 (42%) were female, and median diabetes duration was 19 years.All but 1 participant used multiple daily injections of insulin, and none were using CGM at enrollment.
11scussionCare for patients with diabetes is affected by the low number of endocrinologists compared with the growing number of patients with diabetes and often limited access to diabetes education in primary care and even in endocrinologists' practices.This may prevent patients with diabetes from adopting devices with established clinical benefits such as CGM.Primary care clinicians frequently lack resources to implement use of diabetes technology.In a cross-sectional web-based survey to assess CGM prescribing behaviors, Oser et al11noted that the most needed resources to support CGM in primary care were CGM training and workshops, education, or consultations on insurance coverage.
The VDiSC study was designed before the COVID-19 pandemic to address challenges of access to care through a telehealth care model that could be individualized to include education, technology, and behavioral health support.The 6-month follow-up results of this prospective study showed substantial reduction of HbA 1c values and improvement in TIR.