Long-Term Results of a Digital Diabetes Self-Management and Education Support Program Among Adults With Type 2 Diabetes: A Retrospective Cohort Study

Abstract

Purpose:

The purpose of this study is to examine the long-term impact of a digital diabetes self-management education and support (DSMES) program on A1C among adults with type 2 diabetes (T2DM).

Methods:

Data analyzed were from a retrospective cohort of commercially insured members with T2DM enrolled in the Omada for Diabetes program between January 1, 2019, and January 31, 2022 (n = 1,322). Linear mixed models measured changes in A1C and weight across 12 months (collected at baseline and every 3 months over 1 year) overall and stratified by A1C at baseline (≥8% vs <8%).

Results:

On average, members were 53.5 years old, 56.9% female, and 71.5% White, with a mean baseline body mass index (BMI) of 36.9 and A1C of 7.6%. Members with baseline A1C ≥8% demonstrated clinically and statistically significant adjusted mean reductions in A1C during follow-up, from 9.48% at baseline to 7.33%, 7.57%, 7.59%, and 7.47% at 3, 6, 9, and 12 months, respectively. Those with A1C <8% maintained glycemic stability (6.73%, 6.50%, 6.54%, 6.62%, and 6.51%, respectively). Collectively, members experienced a −1.17 kg/m² mean reduction in BMI over 12 months.

Conclusions:

This study provides real-world evidence that members with elevated baseline A1C (≥8%) enrolled in a digital DSMES program experienced clinically meaningful and statistically significant reductions in A1C. Those with baseline A1C within goal treatment range (<8%) maintained glycemic stability over 1 year. The findings support existing evidence that scalable digital DSMES solutions can help individuals with T2DM manage their condition.

Diabetes mellitus is characterized by the body’s inability to effectively regulate and uptake glucose from the bloodstream due to insulin deficiency and/or resistance, leading to chronic hyperglycemia. The Centers for Disease Control and Prevention estimates that 37.3 million people in the United States are living with diabetes, representing 11.3% of the population.¹ Diabetes is diagnosed when A1C is ≥6.5%, which is an indication of glycemic instability over a 3-month period, and the treatment target for managing diabetes is 7% to 8% for most people with the condition, depending on multiple factors, including the severity of disease and their risk for complications.^2,3 Although clinical guidelines recommend that individuals diagnosed with diabetes should aim to maintain an A1C of <7%, A1C targets should be individualized and reassessed regularly.^4,5 Particularly, a target A1C <8% may be appropriate for individuals susceptible to significant hypoglycemia or other adverse effects from treatment.⁴ Persistent A1C elevation ≥8% is associated with macrovascular and microvascular complications, including an increased risk for cardiovascular disease and renal failure.^6,7 Type 2 diabetes (T2DM) accounts for 90% to 95% of diabetes cases and is associated with lifestyle behaviors and obesity.^1,3 As the incidence of T2DM continues to trend upward, it stands as a leading cause of death and a major public health concern that calls for innovative solutions.^2,8

In addition to the epidemiological impact, managing diabetes comes with a significant socioeconomic burden. Per the American Diabetes Association, the total cost associated with the condition is $327 billion, which is a 26% increase from their previous estimate in 2012.⁹ On average, people with diagnosed diabetes have medical expenditures approximately 2.3 times higher than what expenditures would be in the absence of the condition.⁹ Poor care coordination among adults with chronic illness in the United States further encumbers outcomes.^10-12 To reduce costs and fragmentation of care, increased collaboration between providers and a high level of self-management among patients with diabetes is considered ideal.

Hence, diabetes self-management education and support (DSMES) programs have a strong evidence base and are particularly effective in managing the condition.¹³ DSMES program participation has many benefits, including improved A1C levels, improved blood pressure stability, higher rates of medication adherence, fewer or less severe complications, healthier lifestyle behaviors (eg, optimal diet, increased physical activity, use of clinician services), enhanced self-efficacy, and decreased health care costs (eg, fewer hospital admissions).^13,14 Participating in a DSMES program is critical for those living with diabetes because 98% of the needed care typically requires self-management.¹⁵

Despite this strong evidence, DSMES programs are underutilized. However, advances in technology have ushered in an era of virtual DSMES programs, helping to close the gaps in care through increased access via digital health platforms. A recent meta-analysis revealed that virtual health care services are generally more effective for diabetes care management than in-person care, with the greatest benefits seen in adults over 40 years old.¹⁶ As such, people participating in digital diabetes care programs have seen improved clinical outcomes (eg, A1C reduction) and reduced medical expenditures.^17-20 However, studies evaluating a number of digital health programs have either been short-term (<1 year), focused solely on A1C outcomes, implemented a restrictive diet, and/or not included an extensive behavioral health component.^21-27

Omada for Diabetes is a digital care solution that provides comprehensive diabetes management through an integrated cardiometabolic and behavioral health service model. This program has both automated and human-led touch points, which increases accessibility and personalized care for people with diabetes. In a trial assessing the program, T2DM participants with high A1C at baseline (mean 8.9%) experienced a significant reduction in A1C at their 4-month follow-up in addition to significant improvements in diabetes distress and medication adherence.²⁸ A recent analysis also revealed that participation in this program significantly improves A1C and body weight within 6 months, with the potential to slow onset of diabetes sequelae and reduce medical expenditures.²⁹ Although it is evident that participation in Omada for Diabetes has clinical and cost-saving benefits, research assessing real-world evidence outcomes over a longer period of time is warranted. Therefore, the primary objective of this study is to evaluate the impact of Omada for Diabetes on A1C across 12 months, overall and stratified by A1C at baseline (≥8% vs <8%). In addition, this study assessed the program’s impact on weight and body mass index (BMI) throughout the 12-month study period.

Methods

Study Design

This was a nonrandomized, real-world retrospective cohort study evaluating clinical outcomes over a 12-month time period among commercial health plan members with T2DM enrolled in the Omada for Diabetes program. People must meet the following criteria to enroll in this program: (1) have coverage from their employer or health insurance plan for the benefit, (2) be at least 18 years old, (3) not have any medical contraindications, and 4) have a self-reported diagnosis of diabetes. There was no compensation provided for participation in the program, and all members self-enrolled.

For inclusion in this analysis, members had to meet the following criteria: (1) previously or currently enrolled in Omada for Diabetes; (2) a program start date between January 1, 2019, and January 31, 2022; (3) a baseline A1C value and 1 or more follow-up A1C values; and (4) a self-reported diagnosis of T2DM (Figure 1). This was a secondary analysis of de-identified data previously collected for commercial purposes; we submitted the protocol to the WCG Institutional Review Board (IRB) for review, and they deemed it exempt from IRB approval and oversight (Confirmation ID No. 45104379).

Figure 1.

Member enrollment and study participation flow chart.

Program Description

The Omada for Diabetes program is a digitally delivered DSMES program that pairs asynchronous human support through health coaches and diabetes education specialists with a virtual platform that is accessed either through a website or through an app available on web-enabled devices (eg, smartphone, tablet). The program enhances diabetes self-management skills utilizing behavior change techniques, diabetes education, lifestyle modification (eg, individual health goals), and feedback on self-monitoring data (eg, meal, activity, weight, and glucose tracking) while encouraging ongoing communication with established health care provider(s) outside of the program. Each member is paired with an Omada for Diabetes care team, which consists of a health coach and a specialist. Specialists are certified diabetes care and education specialists (CDCESs) who are also either a registered dietitian or registered nurse.

The Omada for Diabetes program is made available to members through their employer or health plan. The program is designed to include 104 lessons, in which members receive new lessons each week for a period of 2 years. During those 2 years and beyond, should they continue to have coverage via their employer/health plan, the member can continue to engage with their health coach, specialist, peer group, communities, connected devices, and lessons.

The interactive curriculum lessons can be accessed through a computer or mobile device, allowing members to engage at the times and frequency that they choose and with the tools and resources that they find most useful. Recommendations made by the Omada care team are complementary to and in alignment with the members’ care plan created by their external providers. Members can also connect with peers through a closed online forum to discuss related topics and offer social support to each other. The smartphone app also allows members to view and enter personal health data, such as their daily step counts, weigh-ins, and glucose readings. Members can read lessons, revisit program resources, enter meals and snacks into their food tracker, and manage health goals from the program app.

Firmly grounded by evidenced-based guidelines, the Omada for Diabetes program serving T2DM members has achieved accreditation from multiple governing bodies, including the National Committee for Quality Assurance’s Population Health Program Accreditation and the Association of Diabetes Care and Educations Specialists.³⁰ Weekly lessons embody these standards by delivering education on a variety of diabetes-related topics from reviewing carbohydrate basics to leveraging exercise to manage glucose levels. Overall health topics, such as stress management and sleep hygiene, are also covered. Peer interaction is encouraged via a private group board and more broadly with larger community boards focused on topics like healthy cooking.

Eligible members receive cellularly connected devices, such as a blood glucose monitor (BGM) device (Telcare, LLC, Concord, MA, USA or Greater Goods, LLC, St Louis, MO, USA) with strip and lancet refills and a digital scale (Greater Goods, LLC, St Louis, MO, USA or BodyTrace, Inc, Palo Alto, CA, USA), at no cost to them. Additionally, a unique offering of the program is 2 complementary continuous glucose monitor (CGM) sensors (Abbott Diabetes Care, Inc, Alameda, CA, USA) for use at program start and 6 months later. Members who enter the program already using an eligible CGM have the ability to connect their CGM to the program app. This allows the care team and members to review CGM data together directly in the app.

Measures

A1C

A1C (%) was measured via self-report or through the use of an at-home nonfasting AccuBase A1C Test Kit (DTI Laboratories, Inc). These A1C values served as a member’s baseline or follow-up A1C value, as applicable. Self-report values were entered into the program app by either members during account setup or their program care team. When upcoming doctor visits were known to the specialists, they aimed to follow up via private message with the member after the visit to inquire about topics such as regimen changes and A1C values. Results from A1C test kits were transferred directly to the Omada database from the testing lab. Specialists notified members when an A1C kit was sent out and shared a detailed instructional video to help with proper use of the kit.

Baseline A1C was defined as the recorded value closest to program start date and within a window of 2 months prior to and 1 month after program start date. Follow-up A1C data were defined by the following: (1) 3 months: value closest to 3 months postprogram start date and within a window of 3 to 6 months postprogram start date; (2) 6 months: value closest to 6 months postprogram start date and within a window of 6 to 9 months postprogram start date; (3) 9 months: value closest to 9 months postprogram start date and within a window of 9 to 12 months postprogram start date; and (4) 12 months: value closest to 12 months postprogram start date and within a window of 12 to 15 months postprogram start date. At baseline, members were classified as A1C ≥8% versus A1C <8%.³ A clinically meaningful change in A1C over the study period was defined as a reduction of 1%.^31,32

Body Weight

Body weight (lbs) was collected from a cellularly connected digital scale provided as part of the program. Baseline weight was calculated as the average of all weight measurements on the day closest to program start date with at least 1 weight measurement but within the window of 7 days prior to and 30 days after the program start date. The follow-ups at 3, 6, 9, and 12 months were calculated as the average of weight measurements on the day with at least 1 weight measurement closest to the program start date but within the windows of 2 and 4 months, 5 to 7 months, 8 to 10 months, and 11 to 13 months postprogram start date, respectively. Absolute change in body weight (follow-up – baseline) and percent weight loss ([follow-up – baseline] / baseline × 100) were calculated. BMI was calculated from height and weight and categorized into an obesity indicator variable (obesity ≥30 kg/m² vs without obesity <30 kg/m²).

Member Characteristics and Program Engagement

Upon enrollment in the program, members self-reported demographic information, including age (years), sex (male/female), race/ethnicity (White, Black, Hispanic, Asian, other), annual income (>$50k, ≤$50k), and educational attainment (≥college, <college).

Program engagement was measured by counts of actions taken within the program app within each week in the program and included 6 components: (1) number of conversations started on a group discussion board, (2) number of comments (replies to conversations) made on the group discussion board, (3) number of hearts (likes) shared on the group discussion board, (4) number of messages sent to their care team, (5) number of meals tracked, and (6) number of lessons completed. The median of the average weekly actions for all members was used as the cut point for high versus low engagement. T2DM engagement and self-monitoring was assessed by the average times per week a member used their BGM device to assess glucose levels or manually entered glucose readings into the app. Additional self-monitoring behaviors were captured by the average number of times per week members recorded a physical activity and the average number of times per week members recorded a weight.

Statistical Analysis

Differences in baseline sample characteristics between members with baseline A1C ≥8% and <8% were assessed by chi-square and t tests. Changes in clinical outcomes (A1C, weight, percent weight loss, BMI) from baseline to each follow-up time point (3, 6, 9, and 12 months) was tested using paired t tests. Linear mixed-effects regression models were used to evaluate change in A1C (%) as the primary outcome. Unadjusted models included random intercepts for repeated measures within members, random slopes for study time points, and fixed effects for time. Covariates were introduced in a stepwise fashion with age, sex, and race/ethnicity added as fixed effects, followed by time-dependent program engagement, A1C source type, and weight. Member demographics were kept in final models, and additional covariates remained if found to be significant at the P < .05 level. All mixed models were assessed with the overall sample and were then stratified by A1C at baseline and repeated. Model fit and residual plots were examined, and an unstructured covariance matrix was specified in all models.

For secondary outcomes, linear mixed effects regression models were used to model the interaction of time by A1C at baseline on weight and BMI over the study period. Additionally, we conducted sensitivity analyses to ensure the robustness of the findings. We repeated our A1C model analyses using last observation carried forward (LOCF), an imputation method used for repeated measures analyses whereby the last observed nonmissing value is used to fill in future missing values. Also, to understand if there were differences for those with missing data versus not, we compared (1) members with baseline A1C values to those without a baseline A1C value and (2) members with baseline and at least 1 follow-up A1C value versus those with only a baseline A1C value using chi-square and t tests. All analyses used 2-sided hypothesis testing and were conducted in R 4.1 and Stata 17.0.

Results

Sample Characteristics

The final sample size for T2DM members with baseline A1C data and 1 or more follow-ups over the analysis period was 1322, including 411 members with baseline A1C ≥8% and 911 with A1C <8% (Figure 1). Demographic and clinical characteristics overall and by baseline A1C are shown in Table 1. The members were majority White (71.5%) and educated (62.2% with at least a college degree) with a mean age of 53.5 years, and 56.9% were female. About three-fourths of the cohort (73.8%) were classified as obese, and the mean A1C was 7.6% (Table 1). Stratified by A1C at baseline, those with A1C ≥8% were significantly more likely to be younger and have lower incomes, higher mean weight, and be classified as obese (Table 1).

Table 1.

Baseline Characteristics Overall and Based on Starting A1C at Baseline

Characteristic	Full Analytic Sample (N = 1,322)	BaselineA1C ≥8%(n = 411; 31.1%)	BaselineA1C <8%(n = 911; 68.9%)	P Value
Demographics
Age (y), mean ± SD	53.5 ± 8.9	51.5 ± 9.0	54.4 ± 8.7	<.001
Age distribution, n (%)				.007
18-44 y	214 (16.2)	85 (20.7)	129 (14.2)
45-64 y	1,014 (76.7)	308 (74.9)	706 (77.5)
≥65 y	94 (7.1)	18 (4.4)	76 (8.3)
Female, n (%)	752 (56.9)	227 (55.2)	525 (57.6)	.42
Race/ethnicity, n (%)				.03
White	944 (71.5)	284 (69.1)	660 (72.6)
Black	156 (11.8)	46 (11.2)	110 (12.1)
Hispanic	114 (8.6)	50 (12.2)	64 (7.0)
Asian	49 (3.7)	12 (2.9)	37 (4.1)
Other	57 (4.3)	19 (4.6)	38 (4.2)
Education, n (%)				.37
≥College	700 (62.2)	208 (60.3)	492 (63.1)
<College	425 (37.8)	137 (39.7)	288 (36.9)
Income, n (%)				.01
>$50 000	717 (74.2)	210 (69.1)	507 (76.6)
≤$50 000	249 (25.8)	94 (30.9)	155 (23.4)
Clinical characteristics
Weight (lbs), mean ± SD	230.1 ± 53.0	235.7 ± 52.5	227.6 ± 53.1	.01
BMI (kg/m²), mean ± SD	36.9 ± 9.8	37.7 ± 10.0	36.6 ± 9.7	.08
Obesity status, n (%)				.01
BMI ≥30 kg/m²	923 (73.8)	307 (78.3)	616 (71.7)
BMI <30 kg/m²	328 (26.2)	85 (21.7)	243 (28.3)
A1C source type, n (%)				.006
Lab kit	143 (10.8)	30 (7.3)	113 (12.4)
Self-report	1,179 (89.2)	381 (92.7)	798 (87.6)
A1C (%), mean ± SD	7.6 ± 1.6	9.5 ± 1.5	6.7 ± 0.7	<.001

Abbreviation: BMI, body mass index.

Program Engagement

Overall, 50.8% (n = 671) of members were classified as “highly engaged” with the program (median weekly actions ≥6; mean engagement = 15.0 actions per week) versus 49.2% (n = 651) as “less engaged” (weekly actions <6; mean engagement = 2.9 actions per week). Meals tracked (80%) and messages sent to a care team member (9.3%) accounted for about 90% of the mean weekly engagement metric. There was a significant difference in program engagement between members with A1C ≥8% and <8% at baseline (P = .007), with 45.3% of those with A1C ≥8% at baseline considered to be highly engaged compared to 53.2% of those with A1C <8%. Members tracked their physical activity on average 3.9 times per week, weighed in on average 4.1 times per week, and used their BGM or self-reported glucose values on average 4.8 times per week with no significant difference between A1C groups at baseline (P = .71).

Clinical Outcomes

On average, months to follow-up from program start date for A1C data at 3, 6, 9, and 12 months were 4.2 months, 7.3 months, 10.5 months, and 13.2 months, respectively. Overall, members demonstrated significant unadjusted mean reductions in A1C from baseline to each follow-up time point (Table 2). Those with A1C ≥8% at baseline experienced a significant and clinically meaningful reduction in A1C at 3 months (−2.31%) and maintained significant reductions throughout the 12-month study period. Those with A1C ≥8% had significantly larger mean reductions compared to those with A1C <8% at each time point (Table 2).

Table 2.

Unadjusted Mean (95% CI) Change From Baseline in Clinical Outcomes Over the Study Period Overall and by A1C at Baseline^a

Mean (95% CI)	3-mo Change	6-mo Change	9-mo Change	12-mo Change
A1C (%)
Overall	−0.97 (−1.08 to −.86)n = 776	−0.79 (−0.92 to −.65)n = 463	−0.67 (−0.83 to −.51)n = 396	−0.65 (−0.79 to −.51)n = 405
A1C ≥ 8%	−2.31 (−2.53 to −2.08)n = 267	−2.06 (−2.38 to −1.75)n = 138	−1.83 (−2.19 to −1.47)n = 129	−1.84 (−2.22 to −1.47)n = 113
A1C < 8%	−0.27 (−0.34 to −.20)n = 509	−0.24 (−0.32 to −.16)n = 325	−0.11 (−0.22 to −.01)n = 267	−0.19 (−0.28 to −.10)n = 292
Weight (lbs)
Overall	−4.27 (−4.70 to −3.85)n = 1234	−6.94 (−7.64 to −6.24)n = 1193	−7.14 (−7.95 to −6.32)n = 1143	−7.26 (−8.18 to −6.34)n = 1066
A1C ≥8%	−4.34 (−5.24 to −3.43)n = 385	−6.50 (−7.82 to −5.18)n = 373	−6.72 (−8.28 to −5.16)n = 350	−6.55 (−8.29 to −4.82)n = 320
A1C <8%	−4.24 (−4.70 to −3.78)n = 849	−7.14 (−7.97 to −6.32)n = 820	−7.33 (−8.28 to −6.38)n = 793	−7.56 (−8.64 to −6.47)n = 746
Percent weight loss (%)
Overall	−1.79 (−1.96 to −1.61)n = 1234	−2.93 (−3.21 to −2.64)n = 1193	−2.97 (−3.30 to −2.64)n = 1143	−3.0 (−3.37 to −2.63)n = 1066
A1C ≥8%	−1.79 (−2.17 to −1.40)n = 385	−2.69 (−3.22 to −2.16)n = 373	−2.74 (−3.37 to −2.12)n = 350	−2.60 (−3.28 to −1.93)n = 320
A1C <8%	−1.79 (−1.97 to −1.60)n = 849	−3.03 (−3.37 to −2.70)n = 820	−3.07 (−3.46 to −2.69)n = 793	−3.17 (−3.62 to −2.72)n = 746
BMI (kg/m²)
Overall	−0.69 (−0.76 to −.62)n = 1227	−1.12 (−1.24 to −1.01)n = 1186	−1.15 (−1.29 to −1.02)n = 1137	−1.17 (−1.32 to −1.02)n = 1060
A1C ≥8%	−0.70 (−0.85 to −.55)n = 383	−1.03 (−1.25 to −.82)n = 371	−1.07 (−1.32 to −.82)n = 348	−1.04 (−1.33 to −.76)n = 318
A1C <8%	−0.69 (−0.77 to −.62)n = 844	−1.16 (−1.30 to −1.03)n = 815	−1.19 (−1.35 to −1.03)n = 789	−1.23 (−1.40 to −1.05)n = 742

Abbreviation: BMI, body mass index.

Significant difference between members with A1C ≥8% versus <8% occurred at each study time point compared to baseline for A1C (P < .001).

Repeated measures linear models demonstrated similar results to the descriptive changes over time described previously. In unadjusted models, follow-ups at 3, 6, 9, and 12 months were assessed with −0.86%, −0.77%, −0.67%, and −0.73% reduction in A1C overall compared to baseline, respectively (Table 3). Members with A1C <8% experienced smaller magnitude but significant reductions in A1C over the study period and maintained a mean A1C under the target treatment range (Table 3).

Table 3.

Coefficients Estimated by Linear Mixed Models of Study Follow-Ups as a Predictor of Longitudinal A1C (%) Overall and Stratified by A1C at Baseline^a

	Unadjusted		Adjusted Final Model
	Coefficient (95% CI)	P Value	Coefficient (95% CI)	P Value
Full cohort (N = 1322)
Follow-up
3 mo	−0.86 (−0.95 to −.77)	<.001	−0.84 (−1.24 to −.43)	<.001
6 mo	−0.77 (−0.88 to −.67)	<.001	−0.58 (−1.06 to −.09)	.02
9 mo	−0.67 (−0.79 to −.55)	<.001	−0.47 (−1.04 to 0.09)	.10
12 mo	−0.73 (−0.85 to −.60)	<.001	−0.60 (−1.17 to −.02)	.04
A1C ≥8% (n = 411)
Follow-up
3 mo	−2.17 (−2.35 to −1.98)	<.001	−2.75 (−3.62 to −1.88)	<.001
6 mo	−2.00 (−2.25 to −1.76)	<.001	−2.56 (−3.67 to −1.44)	<.001
9 mo	−1.89 (−2.16 to −1.62)	<.001	−2.12 (−3.43 to −.82)	.001
12 mo	−2.01 (−2.33 to −1.70)	<.001	−2.77 (−4.29 to −1.25)	<.001
A1C <8% (n = 911)
Follow-up
3 mo	−0.23 (−0.29 to −.17)	<.001	−0.28 (−0.57 to 0.01)	.06
6 mo	−0.21 (−0.28 to −.13)	<.001	−0.29 (−0.64 to 0.07)	.11
9 mo	−0.13 (−0.22 to −.04)	.003	−0.36 (−0.79 to 0.07)	.10
12 mo	−0.19 (−0.28 to −.11)	<.001	−0.46 (−0.87 to −.05)	.03

Baseline is the reference category for each model. Final model includes age, sex, and race as fixed effects and weight and A1C measurement type as time-dependent effects.

In final adjusted models, all estimates were slightly attenuated but with 3 time points (3, 6, and 12 months) remaining statistically significant for the full cohort (P < .05). For members with A1C ≥8%, A1C decreased by 2.75%, 2.56%, 2.12%, and 2.77% at 3, 6, 9, and 12 months compared to baseline, with all reductions remaining statistically significant (P ≤ .001). Only the reduction in A1C at 12 months was statistically significant for members with A1C <8% in the adjusted model (Table 3).

Figure 2 presents the adjusted marginal estimates of A1C overall and stratified by A1C at baseline. Average marginal estimates for those with A1C ≥8% significantly decreased from 9.48% at baseline to 7.47% at 12 months, with the steepest decline happening from baseline to 3 months, followed by maintenance of this reduction in A1C from 3 to 12 months (Figure 2).

Figure 2.

Adjusted marginal estimates and 95% confidence intervals for A1C (%) overall and stratified by A1C at baseline (≥8% vs. <8%) over the study period.

As shown in Table 2, weight significantly decreased from baseline to each follow-up visit both overall and among those with A1C ≥8% and <8%. At 12 months, members lost on average 7.26 lbs (−8.18, −6.34) and had a 3.0% weight loss, with no significant difference by A1C group (P = .33 and P = .17, respectively; Table 2). Average marginal estimates from mixed models for weight over the study period for those with A1C ≥8% and <8% at baseline decreased from 235.7 to 228.5 lbs and from 227.6 to 220.1 lbs, respectively. In addition, members saw statistically significant reductions in BMI at 12 months (mean = −1.17 kg/m², 95% CI, −1.32 to −1.02), with no significant difference between A1C groups at baseline (Table 2). Average marginal estimates for BMI changes from baseline to 12 months for these 2 groups are depicted in Figure 3.

Figure 3.

Marginal estimates and 95% confidence intervals for BMI (kg/m²) by A1C at baseline (≥8% vs <8%) over the study period.

In a sensitivity analysis, we compared members with baseline and at least 1 follow-up A1C value to those with only a baseline A1C and members with a baseline A1C value to those without a baseline value. Members with baseline and follow-up data compared to those with only baseline data were more likely to self-report having an income over $50 000 (74.2% vs 63.3%, P = .06) and have higher baseline A1C values (7.60% vs 7.27%, P = .07). All other demographic characteristics were similar between groups. Those with a baseline A1C value compared to those without a baseline A1C value were more likely to be older (53.3 vs 50.7 years, P < .001), identify as White (71.9% vs 50.7%, P < .001), report having at least a college education (61.8% vs 55.1%, P < .001), and report household incomes of over $50 000 (73.7% vs 65%, P < .001). Additionally, all findings for LOCF analysis models were consistent in both magnitude and significance with the final model analysis (data not shown).

Discussion

Results from this retrospective cohort study show that T2DM members participating in the Omada for Diabetes program statistically significantly and clinically meaningfully reduced their A1C over a 1-year period. Members with elevated A1C (≥8%) at baseline experienced a 2-point decline in A1C over the course of 12 months of program participation (mean A1C 7.47% at study end) (Figure 2). The steepest decline happened in the first 3 months of the program, and the improvements were maintained over the remaining 9 months. The magnitude of this decline remained consistent even after adjusting for age, sex, and race/ethnicity as fixed effects and A1C measurement source type and weight as time-dependent covariates. This reduction in A1C is clinically meaningful because it has the potential to reduce the risk of comorbid illness and complications in individuals with T2DM.^6,31 Research has shown that long-term maintenance of a 1% reduction in A1C is associated with a decline in both diabetes-related deaths and microvascular complications.^33,34

Members with a baseline A1C of <8% demonstrated maintenance of blood glucose stability throughout the 12-month period, averaging 6.73% at baseline and 6.51% at 12 months (Figure 2). This is a clinically meaningful finding as well, given that maintenance is also a challenging aspect of diabetes management.³⁵ Compared to those with A1C ≥8% at baseline, members in this group differed significantly in terms of age, income, obesity status, and body weight.

Program engagement remained high throughout most of the study period in tandem with clinical improvements in A1C and body weight, suggesting that the program was successfully able to capture these members’ sustained attention and active participation over a 12-month period. This is promising because research shows that lifestyle change programs longer in duration are more successful at helping participants sustain healthy behaviors and, in turn, maintain clinical improvements.^36,37

Past research has also shown that obesity interventions typically result in rapid weight loss within the first few months of implementation, whereas weight maintenance efforts usually commence after 6 months.³⁵ The maintenance period is generally considered the more challenging facet of weight management, underscoring an exciting aspect of our study results. Clinically obese members with highly elevated A1C not only experienced reductions in A1C and weight at 3 months, but they were also able to successfully maintain those results at 12 months in the program. These findings suggest that studies assessing clinical outcomes over a 3-month time frame may reliably predict longer-term trends for Omada for Diabetes program members.

Comparable real-world evidence studies from other digital health programs have also reported significant changes in clinical outcomes; however, their studies were conducted with smaller sample sizes and over shorter periods of time.^23,26,27 One study found that individuals with T2DM participating in a 12-week digital health coaching program achieved significant reductions in A1C and BMI, with high-risk participants (A1C >9%) reducing their levels by the greatest margin.²⁶ Additional research shows that participating in digital DSMES programs has benefits beyond weight loss and achievement of glycemic targets, with growing evidence suggesting that digital diabetes care imparts positive effects on overall cardiometabolic risk.^38,39 Hence, the results from this current analysis are promising given the extent of improvements in clinical outcomes among our study sample and the likely long-term positive impact these changes have on multiple biomarkers of a member’s health.

Other digital health solutions that rely on restrictive dietary regimens, such as the ketogenic diet or an elimination diet, may result in steeper short-term weight loss and declines in A1C, but these dietary restrictions can be emotionally cumbersome and difficult to sustain long-term.^40-42 The combination of these factors can impact long-term outcomes, potentially hindering weight maintenance and glycemic stability over time.^43,44 Omada’s program focuses on healthy eating and lifestyle principles rather than an emphasis on highly restrictive diets. By focusing on mindset and balanced nutrition, members in Omada’s program learn sustainable practices to manage their condition. Studies have shown that this comprehensive approach promotes weight loss among people with obesity and/or T2DM.^45-47

Strengths

The current study has many strengths, including its sample size and 12-month time frame, with 3-month increments along the way. The multiple time points enabled us to observe both short-term outcomes and long-term maintenance of clinically meaningful improvements. The power of mixed-model analyses allowed us to use all available study data, thereby increasing the precision and confidence in our findings.⁴⁸ Our sample of members was relatively diverse in terms of demographic distribution and sample characteristics, augmenting the generalizability of our findings. Our data provide real-world evidence for the effects of a comprehensive digital health program involving diabetes education, BGM/CGM self-monitoring, and lifestyle behavior change program because we did not use a strictly self-selected sample of trial participants but rather, commercial members self-enrolled in the program.

Limitations

In addition to its many strengths, this study also had a number of limitations. First, members were enrolled in the program based on a self-reported diagnosis of T2DM, and access to medical records was not available to further validate the diagnosis. Additionally, because the study was a nonrandomized, real-world, retrospective cohort design, we were limited by data already collected and therefore could not ensure complete measurement collection at specific follow-up time periods for evaluating our main outcome.

For example, A1C data collection is self-reported and voluntary, which likely contributed to the decline in members with reported A1C values at each time point over the yearlong program period. Other factors that may have contributed to reporting variability include a member’s access to A1C testing, availability to undergo testing, awareness of their most recent A1C value after a clinic visit, and/or willingness to disclose their latest A1C value to their care team. Therefore, the sample available was subject to a degree of self-selection and reporting bias, potentially misrepresenting the overall member population of the Omada for Diabetes program and the A1C data collected. In sensitivity analyses to explore the impact of missing data, members with follow-up data (baseline + ≥1 follow-up) compared to those with only baseline A1C were similar; however, demographic differences with regards to age, race, education, and income were detected between those with a baseline A1C value compared to those without a baseline A1C value. Thus, future research should investigate the potential barriers that exist among eligible members for the program who do not self-report an A1C value at baseline. Furthermore, due to the real-world nature of this study and because we did not have a control group, we were unable to examine a cause-effect relationship and thus assessed correlations between exposures and outcomes. Lastly, the lack of a control group due to the observational study design may have introduced confounding and bias to the results, and findings could be attributable in part to regression to the mean.

Conclusion and Future Implications

The findings of this study indicate that members participating in the Omada for Diabetes program significantly reduced their A1C and maintained glycemic stability over 12 months. Members also experienced clinically significant improvements in weight and remained engaged throughout the duration of the 1-year program. A randomized controlled trial is warranted to evaluate a potential causal relationship between participation in the Omada for Diabetes program and changes in A1C. Future research should also examine potential changes in diabetes-related comorbidities and health care cost savings in order to fully capture the extent of the program’s influence in serving the T2DM population.^49,50 Finally, future research should explore opportunities to increase collaboration with physician group practices and large health systems to reduce fragmentation of care and improve clinical outcomes in members with T2DM.

Footnotes

Acknowledgements

The authors thank the data analytics team at Omada Health for their assistance with data extraction.

Declaration of Conflicting Interests

AB, JN, MM, AM, and SL are employees of Omada Health, Inc and receive salary and stock options. JJ received consulting fees from Omada Health, Inc.

Funding

This work was supported by Omada Health, Inc.

ORCID iD

Jenna Napoleone

References

Centers for Disease Control and Prevention. National diabetes statistics report: estimates of diabetes and its burden in the United States. Last Reviewed November 29, 2023. Accessed December 20, 2023. https://www.cdc.gov/diabetes/data/statistics-report/index.html?ACSTrackingID=DM72996&ACSTrackingLabel=New%20Report%20Shares%20Latest%20Diabetes%20Stats%20&deliveryName=DM72996

National Diabetes Statistics Report 2020. 2020. Accessed August 23, 2022. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf

Qaseem

Wilt

Kansagara

, et al. Hemoglobin A_1c targets for glycemic control with pharmacologic therapy for nonpregnant adults with type 2 diabetes mellitus: a guidance statement update from the American College of Physicians. Ann Intern Med. 2018;168(8):569-576. doi:10.7326/M17-0939

American Diabetes Association Professional Practice Committee. Glycemic targets: Standards of Medical Care in Diabetes—2022. Diabetes Care. 2022;45(suppl 1):S83-S96. doi:10.2337/dc22-S006

American Diabetes Association. Comprehensive medical evaluation and assessment of comorbidities: Standards of Medical Care in Diabetes—2021. Diabetes Care. 2021;44(suppl 1):S40-S52. doi:10.2337/dc21-S004

Chawla

Jaggi

Microvasular and macrovascular complications in diabetes mellitus: distinct or continuum?

Indian J Endocrinol Metab. 2016;20(4):546-551. doi:10.4103/2230-8210.183480

Torkamani

Churilov

Robbins

, et al. Diabetes and higher HbA1c levels are independently associated with adverse renal outcomes in inpatients following multiple hospital admissions. J Diabetes Complications. 2020;34(1):107465. doi:10.1016/j.jdiacomp.2019.107465

Lin

Pan

, et al. Global, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025. Sci Rep. 2020;10(1):14790. doi:10.1038/s41598-020-71908-9

American Diabetes Association. Economic costs of diabetes in the U.S. in 2017. Diabetes Care. 2018;41(5):917-928. doi:10.2337/dci18-0007

10.

Lee

McCutcheon

LRM

Fazel

Cooley

Slack

MK.

Assessment of interprofessional collaborative practices and outcomes in adults with diabetes and hypertension in primary care: a systematic review and meta-analysis [published correction appears in JAMA Netw Open. 20211;4(4):e219114]. JAMA Netw Open. 2021;4(2):e2036725. doi:10.1001/jamanetworkopen.2020.36725

11.

Lewis

Williams

Fitzgerald

, et al. Improving the implementation of diabetes self-management: findings from the Alliance to Reduce Disparities in Diabetes. Health Promot Pract. 2014;15(2 suppl):83S-91S. doi:10.1177/1524839914541277

12.

Frandsen

Joynt

Rebitzer

Jha

AK.

Care fragmentation, quality, and costs among chronically ill patients. Am J Manag Care. 2015;21(5):355-362.

13.

Powers

Bardsley

Cypress

, et al. Diabetes self-management education and support in adults with type 2 diabetes: a consensus report of the American Diabetes Association, the Association of Diabetes Care & Education Specialists, the Academy of Nutrition and Dietetics, the American Academy of Family Physicians, the American Academy of PAs, the American Association of Nurse Practitioners, and the American Pharmacists Association. Diabetes Care. 2020;43(7):1636-1649. doi:10.2337/dci20-0023

14.

Beck

Greenwood

Blanton

, et al. 2017 National standards for diabetes self-management education and support. Sci Diabetes Self Manag Care. 2021;47(1):14-29. doi:10.1177/0145721720987926

15.

Piccinino

Devchand

Gallivan

Tuncer

Nicols

Siminerio

LM.

Insights from the National Diabetes Education Program National Diabetes Survey: opportunities for diabetes self-management education and support. Diabetes Spectr. 2017;30(2):95-100. doi:10.2337/ds16-0056

16.

Tchero

Kangambega

Briatte

Brunet-Houdard

Retali

G-R

Rusch

Clinical effectiveness of telemedicine in diabetes mellitus: a meta-analysis of 42 randomized controlled trials. Telemed J E Health. 2019;25(7):569-583. doi:10.1089/tmj.2018.0128

17.

Whaley

Bollyky

, et al. Reduced medical spending associated with increased use of a remote diabetes management program and lower mean blood glucose values. J Med Econ. 2019;22(9):869-877. doi:10.1080/13696998.2019.1609483

18.

Turner

Lorig

Greenberg

DeVries

AR.

Evaluation of a diabetes self-management program: claims analysis on comorbid illnesses, health care utilization, and cost. J Med Internet Res. 2018;20(6):e207. doi:10.2196/jmir.9225

19.

Phillip

Bergenstal

, et al. The digital/virtual diabetes clinic: the future is now-recommendations from an international panel on diabetes digital technologies introduction. Diabetes Technol Ther. 2021;23(2):146-154. doi:10.1089/dia.2020.0375

20.

Adu

Malabu

Malau-Aduli

AEO

Malau-Aduli

BS.

Enablers and barriers to effective diabetes self-management: a multi-national investigation. PLoS One. 2019;14(6):e0217771. doi:10.1371/journal.pone.0217771

21.

Lorig

Ritter

Turner

English

Laurent

Greenberg

A diabetes self-management program: 12-month outcome sustainability from a nonreinforced pragmatic trial. J Med Internet Res. 2016;18(12):e322. doi:10.2196/jmir.6484

22.

Park

J-I

Jeon

Choi

HJ.

Clinical efficacy and plausibility of a smartphone-based integrated online real-time diabetes care system via glucose and diet data management: a pilot study. Intern Med J. 2020;50(12):1524-1532. doi:10.1111/imj.14738

23.

Zimmermann

Venkatesan

Rawlings

Scahill

MD.

Improved glycemic control with a digital health intervention in adults with type 2 diabetes: retrospective study. JMIR Diabetes. 2021;6(2):e28033. doi:10.2196/28033

24.

Hallberg

McKenzie

Williams

, et al. Effectiveness and safety of a novel care model for the management of type 2 diabetes at 1 year: an open-label, non-randomized, controlled study. Diabetes Ther. 2018;9(2):583-612. doi:10.1007/s13300-018-0373-9

25.

Athinarayanan

Adams

Hallberg

, et al. Long-term effects of a novel continuous remote care intervention including nutritional ketosis for the management of type 2 diabetes: a 2-year non-randomized clinical trial. Front Endocrinol (Lausanne). 2019;10:348. doi:10.3389/fendo.2019.00348

26.

Martin

Patterson

Allison

O’Connor

Patel

The influence of baseline hemoglobin A_1c on digital health coaching outcomes in adults with type 2 diabetes: real-world retrospective cohort study. JMIR Diabetes. 2021;6(2):e24981. doi:10.2196/24981

27.

Y-Z

Chang

Y-T

Chiou

H-Y

Lai

The effects of continuous usage of a diabetes management app on glycemic control in real-world clinical practice: retrospective analysis. J Med Internet Res. 2021;23(7):e23227. doi:10.2196/23227

28.

Wilson-Anumudu

Quan

Castro Sweet

, et al. Early insights from a digitally enhanced diabetes self-management education and support program: single-arm nonrandomized trial. JMIR Diabetes. 2021;6(1):e25295. doi:10.2196/25295

29.

Chen

Jasik

Dall

Siego

CV.

Impact of a digitally enhanced diabetes self-management program on glycemia and medical costs. Sci Diabetes Self Manag Care. 2022;48(4):258-269. doi:10.1177/26350106221100779

30.

Business Wire. Omada health becomes first fully-virtual healthcare provider to earn NCQA population health program accreditation. May 25, 2021. Accessed September 27, 2022. https://www.businesswire.com/news/home/20210525005094/en/Omada-Health-Becomes-First-Fully-Virtual-Healthcare-Provider-to-Earn-NCQA-Population-Health-Program-Accreditation

31.

Stratton

Adler

Neil

, et al. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ. 2000;321(7258):405-412. doi:10.1136/bmj.321.7258.405

32.

Kaiafa

Veneti

Polychronopoulos

, et al. Is HbA1c an ideal biomarker of well-controlled diabetes? Postgrad Med J. 2021;97(1148):380-383. doi:10.1136/postgradmedj-2020-138756

33.

American Diabetes Association. Standards of medical care in diabetes—2008. Diabetes Care. 2008;31(suppl 1):S12-S54. doi:10.2337/dc08-S012

34.

Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet. 1998;352(9131):837-853. doi:10.1016/S0140-6736(98)07019-6

35.

Montesi

El Ghoch

Brodosi

Calugi

Marchesini

Dalle Grave

Long-term weight loss maintenance for obesity: a multidisciplinary approach. Diabetes Metab Syndr Obes. 2016;9:37-46. doi:10.2147/DMSO.S89836

36.

Gotthelf

Chen

Y-T

Rajagopalan

EC-T

Doshi

Addy

High intensity lifestyle intervention and long-term impact on weight and clinical outcomes. PLoS One. 2018;13(4):e0195794. doi:10.1371/journal.pone.0195794

37.

Yamaoka

Nemoto

Tango

Comparison of the effectiveness of lifestyle modification with other treatments on the incidence of type 2 diabetes in people at high risk: a network meta-analysis. Nutrients. 2019;11(6):1373. doi:10.3390/nu11061373

38.

Lesser

Behal

Change in glycemic control for patients enrolled in a membership-based primary care program: longitudinal observational study. JMIR Diabetes. 2021;6(2):e27453. doi:10.2196/27453

39.

Ernawati

Wihastuti

Utami

YW.

Effectiveness of diabetes self-management education (DSME) in type 2 diabetes mellitus (T2DM) patients: systematic literature review. J Public Health Res. 2021;10(2):2240. doi:10.4081/jphr.2021.2240

40.

Memon

Gowda

Rallabhandi

, et al. Have our attempts to curb obesity done more harm than good? Cureus. 2020;12(9):e10275. doi:10.7759/cureus.10275

41.

Crosby

Davis

Joshi

, et al. Ketogenic diets and chronic disease: weighing the benefits against the risks. Front Nutr. 2021;8:702802. doi:10.3389/fnut.2021.702802

42.

Polivy

Psychological consequences of food restriction. J Am Diet Assoc. 1996;96(6):589-593. doi:10.1016/S0002-8223(96)00161-7

43.

Maston

Franklin

Hocking

, et al. Dietary adherence and program attrition during a severely energy-restricted diet among people with complex class III obesity: a qualitative exploration. PLoS One. 2021;16(6):e0253127. doi:10.1371/journal.pone.0253127

44.

Choi

Kang

J-H

Chon

Comprehensive understanding for application in Korean patients with type 2 diabetes mellitus of the consensus statement on carbohydrate-restricted diets by Korean Diabetes Association, Korean Society for the Study of Obesity, and Korean Society of Hypertension. Diabetes Metab J. 2022;46(3):377-390. doi:10.4093/dmj.2022.0051

45.

Jia

Liu

Allman-Farinelli

, et al. The use of portion control plates to promote healthy eating and diet-related outcomes: a scoping review. Nutrients. 2022;14(4):892. doi:10.3390/nu14040892

46.

Hamdy

Barakatun-Nisak

M-Y.

Nutrition in diabetes. Endocrinol Metab Clin North Am. 2016;45(4):799-817. doi:10.1016/j.ecl.2016.06.010

47.

Koliaki

Spinos

Spinou

Brinia

M-E

Mitsopoulou

Katsilambros

Defining the optimal dietary approach for safe, effective and sustainable weight loss in overweight and obese adults. Healthcare (Basel). 2018;6(3):73. doi:10.3390/healthcare6030073

48.

Yang

Zaitlen

Goddard

Visscher

Price

AL.

Advantages and pitfalls in the application of mixed-model association methods. Nat Genet. 2014;46(2):100-106. doi:10.1038/ng.2876

49.

Thorpe

Toles

Shah

Schneider

Bravata

DM.

Weight loss-associated decreases in medical care expenditures for commercially insured patients with chronic conditions. J Occup Environ Med. 2021;63(10):847-851. doi:10.1097/JOM.0000000000002296

50.

Lage

Boye

KS.

The relationship between HbA1c reduction and healthcare costs among patients with type 2 diabetes: evidence from a U.S. claims database. Curr Med Res Opin. 2020;36(9):1441-1447. doi:10.1080/03007995.2020.1787971