This was a prospective study that investigated the impact of the Livongo for Diabetes program on HbA_1c and the proportion of BG checks in range for patients with diabetes mellitus at the University of South Florida Diabetes Home for Healthy Living (USF DHHL). The Livongo for Diabetes program is a digital chronic condition management program that combines: (1) a Food and Drug Administration–cleared, cellular-enabled, two-way messaging glucometer that measures blood glucose and delivers personalized digital coaching messages (see Multimedia Appendix 1); (2) free unlimited blood glucose test strips; and (3) unlimited access to CDEs for goal setting and behavioral and lifestyle education based on the American Diabetes Association’s (ADA) Standards of Medical Care and the American Association of Diabetes Educator’s (AADE) Diabetes Education Prompt Deck and Educator Guide [19,20].

Personalized digital coaching methods are delivered algorithmically on the meter to members based on diabetes type, medication use, and clinical guidelines. Immediately following each BG check, members receive context-specific feedback based on the BG value measured, as well as BG trends and patterns established with repeated meter usage. This feedback is delivered through messages less than 140 characters in length and based on ADA and AADE recommendations.

The CDEs also provided 24 hours a day, 7 days a week, 365 days a year call support for members with BG readings of <50 mg/dL or >400 mg/dL, within 3 minutes of transmitted blood glucose, to provide ADA-recommended, nonmedication-related interventions to effect their BG (ie, “drink 8 ounces of orange juice to bring your BG values up and recheck BG in 15 minutes”).

In addition to the Livongo glucometer and access to CDEs, participants had access to a mobile phone application on iOS and Android, and a web portal available through traditional web browsers that tracked historical BG readings, provided reminders for BG checking, and allowed members to send Health Summary Reports of BG readings to care providers, family members, and friends (see Multimedia Appendix 2).

A convenience sample of participants were enrolled from the USF DHHL clinic from February 2015 to February 2016. All participants recruited for the study were established patients of the clinic with elevated HbA_1c who were receiving their usual care. Patients were eligible if they were at least 18 years of age, diagnosed with type 1 or type 2 diabetes mellitus, and proficient in English. Patients were excluded from the study if they did not have a baseline HbA_1c value and at least one other HbA_1c value within the study period for comparison, did not have a follow-up visit, and never activated the device (Figure 1). In addition, patients who died during the study were also excluded due to unavailable health information because of closed medical records.

The study protocol was approved by the University of South Florida Institutional Review Board (Protocol #PRO00016476). Verbal and written informed consent were obtained prior to participant’s enrollment in the study. Study procedures were conducted in compliance with the Declaration of Helsinki.

Summarizing statistics were computed for demographic characteristics. Outcome variables were computed between baseline and subsequent clinic visits. The nonparametric Wilcoxon rank sum test was used to compare continuous variables, and Fisher’s exact test was used for categorical data comparisons.

Baseline characteristics are presented in Table 1. Nearly half of the participants were female, with a mean age of 50 (SD 15) years, and were diagnosed with type 2 diabetes (56%; 48/86).

Mean HbA_1c improved from baseline in all participants throughout the intervention, and within each diabetes type. Statistically significant improvements were seen in all participants from baseline to 3 months (0.8%; P=.02). Additionally, insulin users, whether with type 1 or type 2 diabetes, experienced a greater decrease in HbA_1c than noninsulin users, at both 3 months (0.8%; P=.04) and 6 months (1.0%; P=.05). While HbA_1c improved from baseline to 12-months, it was not statistically significant at any time point for participants with type 1 diabetes, type 2 diabetes, or participants with no insulin usage, whether with type 1 or type 2 diabetes. Further details about changes in HbA_1c over the 12-month intervention by subgroups are reported in Figures 2-6.

Blood glucose values were analyzed by BG range categories, and by diabetes type, for all participants during the study period. Median BG checking frequency ranged from 0.8 to 1.0 times per day depending on diabetes type and insulin use. Participants with type 2 diabetes using insulin had the highest BG checking frequency at approximately 1.0 (SD 1.2) times per day, while participants with type 2 diabetes not using insulin checked an average of 0.8 (SD 1.0) times per day and participants with type 1 diabetes were at 0.8 (SD 1.8) checks per day.

Patients with type 2 diabetes not using insulin had the highest percentage of BG checks within the target range of 70-180mg/dL (89.0%; SD 15.9), compared to participants with type 2 diabetes on insulin (68.1%; SD 28.1) and type 1 diabetes (57.9%; SD 22.5) throughout the 12 months.

When comparing percentage of BG checks in range in the first three months of the study versus the last three months, all participants decreased their percentage of BG checks that were <70 mg/dL from 4.9% to 4.1% (P=.56; see Figure 7). Though not statistically significant, participants with type 1 diabetes experienced a slight decrease in percentage of BG checks over 400 mg/dL from 1.8% to 1.5% (P=.81). Similarly, participants with type 2 diabetes using insulin had a nonsignificant decrease in percentage of BG checks greater than 180 mg/dL from 33.5% to 25.9% (P=.43). Participants with type 2 diabetes not using insulin showed no significant improvements for BG time in range. Further comparison of BG checks in range by diabetes type from 0-3 months to 9-12 months is shown in Multimedia Appendix 3.

Over the intervention period, there was a statistically significant decrease in mean PHQ-2 score (P=.04) among all participants. Average baseline PHQ-2 score (N=40) was 0.83. Postintervention score (N=19) was 0.26. While both baseline and postintervention scores were <3, showing an unlikelihood for depression symptoms, participants still showed a statistically significant decrease from baseline to 12 months. When analyzing PHQ-2 score by diabetes type and insulin usage, participants using insulin showed a statistically significant decline in PHQ-2 score (1.03 to 0.19; P=.01), while type 1, type 2, and participants not using insulin saw a nonstatistically significant improvement in scores.

The results of this study in an outpatient diabetes clinic provide evidence that access to a cellular-enabled BG meter connected to CDEs with real-time personalized recommendations can improve HbA_1c. This improvement was significant since our study showed that with lower HbA_1c, participants also had increased BG values within target range, with a decrease in hypoglycemic events at 12 months. Furthermore, study participants had improved depressive symptom scores as measured by PHQ-2 surveys. Overall, a connected BG meter with personalized feedback and access to CDEs improved diabetes care at 12 months.

The ADA and WHO recognize digital health and technology advances can support and enhance the delivery of health services [6,19]. Specifically, the ADA’s 2019 Standards of Medical Care includes recommendations for diabetes technology recognizing digital self-management solutions for improvement in HbA_1c, especially when paired with a health care team, individualized feedback, patient generated historical health data, and education [19]. Additionally, the ADA recommends patients receive ongoing education and evaluation of glucose data to adjust therapy and self-care in relation to individual needs [19,22]. With the growing global epidemic of diabetes linked to a lack of knowledge around BG monitoring, self-management, education, and episodic communication with health care providers, testing a cellular-enabled BG meter with BG checking reminders, personalized digital coaching, access to CDEs, and historical BG reporting was critical to understand the benefit of including digital diabetes solutions related directly to issues increasing diabetes prevalence, in addition to a diabetes clinic’s standard care [1-5].

Using a cellular-connected BG meter provides health care professionals with instant access to patient-generated BG readings, allowing for faster change in care plan, education, and outcomes. In addition, a patient’s care team can provide a more personalized and proactive plan, with tailored education through the system’s generated insights of historical BG data and two-way communication with CDEs available through the Livongo meter [23]. Viewing a patient’s BG values within low, normal, or high ranges over a week or month allows for a timelier response in condition management versus waiting 3 months for a change in HbA_1c. In addition, two individuals with the same HbA_1c could have very different time in BG ranges, which would impact desired treatment plan. Without the ability to view historical BG readings in a timely manner, an important aspect of an individual’s personalized care plan could be overlooked as it would not be reflected in HbA_1c. As such, our study supports previous findings that access to a cellular-connected glucometer and CDE coaching decreases hypoglycemic episodes and leads to a decrease in HbA_1c up to 1% in 3 months [11,13,15]. Also, having access to a program like Livongo for Diabetes can provide continued BG monitoring, education, and coaching for individuals with diabetes who choose to not follow up with their health care team as recommended.

Combining digital health tools with human coaching for individuals with diabetes has also improved depressive symptoms, as measured by PHQ scores [17,18,24]. By incorporating technology and access to real-time support from CDEs into diabetes care, the challenges of self-management that can increase depressive symptoms, such as support, emotional burden, and access to education and management, are addressed in a more timely manner focused on the patient’s personalized needs [17,25]. The addition of coaching offers reinforcement of education, accountability, and creation of problem-solving skills to overcome behavioral and cognitive barriers for successful self-management [25].

The Livongo for Diabetes program has also been shown to provide cost savings to its users [26,27]. In 2019, Livongo users had a 21.9% decrease in spending compared to nonusers, translating to $88 per month. Specifically, a 10.7% reduction was observed in diabetes-related medical spending, and a 24.6% reduction in spending for office-based services [27]. While historically offering ongoing human coaching can be costly, the Livongo for Diabetes program has provided a return on investment for its users while improving clinical outcomes.

The limitations of this study include the lack of a control group and the small sample size for subgroup comparisons. Authors assume the dramatic decline in sample size is related to a physician leaving the clinic, resulting in a lack of follow up from that physician’s patients since patients come to the clinic to see specific providers. In addition, participants for this study were recruited at a Diabetes Center of Excellence, which provides access to the highest level of diabetes care. The BG checking frequency was unexpectedly low in the type 1 diabetes population. This finding may be a result of continuous glucose monitor use in this population, which requires two BG checks for calibration. Neither this information nor the use of non-Livongo BG meters were captured as part of the study.

Finally, changes in medication use, weight, knowledge of diabetes self-management, coaching interactions, and other factors that might influence BG control were not captured as part of this study. Further investigations will be required to see if findings would be applicable to the general population and to better understand the drivers of improved glucose control.

Despite the small sample size, this study provides a glimpse of how adding a new product into the market, or with standard care, can improve patient outcomes even in centers of excellence. This is an important contribution to the literature and for larger population studies in the future.

Participants provided with a cellular-enabled BG meter with real-time feedback and access to CDE coaching in a diabetes center of excellence experienced a reduction in HbA_1c, fewer hypoglycemic episodes, and a significant reduction in PHQ-2 scores. These results support evidence that the addition of diabetes digital health solutions can improve diabetes care. Further studies should be conducted to assess a larger population with the addition of coaching interactions, medication use, education, and self-management behaviors.