Healthcare Cost Impacts of Telehealth-Delivered Nutrition Therapy among US Adults with Type 2 Diabetes or Obesity: A Matched Difference-in-Differences Analysis

Abstract

Introduction

Type 2 diabetes (T2D) and obesity are associated with substantial morbidity and healthcare spending. Although telehealth-delivered lifestyle interventions have demonstrated clinical benefits, evidence on their impact on healthcare costs and utilization remains limited. This study evaluated the association between participation in a comprehensive nutrition-first digital care model integrating individualized carbohydrate-reduced nutrition support, clinician-guided medication management, health coaching, and remote biometric monitoring on total healthcare costs and utilization over 1 year and 2 years.

Methods

We conducted a claims-based, retrospective propensity score matched difference-in-differences analysis of cost and utilization outcomes among US adults with T2D or obesity who (1) enrolled in Virta Health’s Individualized Nutrition Therapy (INT) program or (2) had a primary care visit for either condition between 2017 and 2025. Per-member, per-month (PMPM) outcomes included inpatient, outpatient, and prescription medication costs, as well as inpatient, emergency department, primary care, cardiology, and endocrinology visits. For the T2D cohort only, PMPM costs for each diabetes medication were also evaluated.

Results

Among 3819 adults with T2D and 2761 adults with obesity, program participation was associated with $240 and $256 PMPM reductions in total cost of care at 12 months, respectively (−$230 and −$189 PMPM over 24 months; all p < 0.001). Among adults with T2D, savings were driven by reduced inpatient visits and deprescription of sodium-glucose cotransporter 2 (SGLT2) inhibitors (66.8% reduction from baseline cost), sulfonylureas (51.7%), insulin (43.9%), and glucagon-like peptide-1s (GLP-1s) (32.2%). Among adults without T2D and with obesity, reductions accrued across all cost categories.

Conclusions

In this large, real-world analysis, enrollment in a comprehensive nutrition-first digital care model was associated with reductions in healthcare costs and utilization over 12 and 24 months, with prescription medication cost reductions emerging within the first quarter of treatment among adults with T2D. Together with prior evidence demonstrating clinical effectiveness, these findings suggest that comprehensive digital nutritional care models can generate meaningful economic value.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s13300-026-01909-w.

Plain Language Summary

Type 2 diabetes (T2D) and obesity affect millions of people, increasing their risk for poor health outcomes and increased healthcare costs. Telehealth nutrition programs are effective at managing these conditions but less is known about whether they also reduce healthcare spending. This study used healthcare claims data to compare adults who enrolled in a telehealth nutrition program with similar adults who received care from their primary care physician (PCP). The program provided personalized nutrition guidance focused on reducing carbohydrate intake and managed T2D- and obesity-related medications . The researchers compared healthcare costs between the two groups over 1 year and 2 years. Adults who participated in the program had lower total healthcare costs than similar adults who received care from their PCP. Lowered costs were seen after 1 year and were maintained over 2 years. Among adults with T2D, the largest savings came from fewer hospital stays and lower use of several diabetes medications. Among adults with obesity, healthcare costs decreased across medications, hospital stays, and outpatient visits. The results of the study suggest that telehealth nutritional care may improve health while reducing healthcare costs for adults with T2D or obesity. These findings add to previous research showing that lifestyle programs delivered through telehealth can improve patients’ health while also reducing costs.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s13300-026-01909-w.

Key Summary Points

Why carry out this study?

Type 2 diabetes and obesity are common chronic conditions that drive substantial morbidity and healthcare costs.

Although telehealth-delivered lifestyle interventions have been shown to improve clinical outcomes, less is known about their impact on healthcare costs and utilization, particularly in real-world settings.

The study hypothesized that, among adults with T2D or obesity, enrollment in a comprehensive nutrition-first digital care model would be associated with lower healthcare costs and utilization over 1 year and 2 years.

What was learned from the study?

In a propensity score-matched, claims-based analysis, participation in the program was associated with significant reductions in total healthcare costs at 12 months ($240 per member per month for type 2 diabetes and $256 per member per month for obesity) that were sustained through 24 months.

A comprehensive nutrition-first digital care model was associated with sustained reductions in healthcare costs and utilization in a large real-world population of adults with type 2 diabetes or obesity. Together with previously published evidence demonstrating the clinical effectiveness of the care model, these findings suggest that such programs may represent a scalable cost-reducing strategy for chronic disease management.

Introduction

Type 2 diabetes (T2D) and obesity represent major interrelated public health challenges in the USA, impacting more than 40 million [1] and 100 million [2] Americans, respectively. These syndemic conditions increase the risk of comorbid conditions including cancer, kidney disease, and cardiovascular disease [3, 4] and create a staggering economic burden, with combined direct healthcare costs exceeding $500 billion annually [5, 6].

Despite significant advances in pharmacologic treatment options for both T2D and obesity, including glucagon-like peptide-1 (GLP-1) receptor agonists, real-world effectiveness is constrained by adverse effects, polypharmacy risks, and coverage and affordability barriers that limit initiation and persistence [7–16]. Moreover, even under guideline-concordant therapy, substantial residual cardiometabolic risk remains, underscoring the need for scalable and sustainable complementary approaches to chronic disease management [17].

Telehealth has become an increasingly important modality for chronic disease management by delivering continuous, between-visit support integrating remote biomarker monitoring, clinician-guided medication management, behavioral coaching, and specialist input [18–21]. Virta Health’s Individualized Nutrition Therapy (INT) program combines individualized carbohydrate-reduced nutrition, continuous remote care, and clinician-guided proactive medication management. The clinical effectiveness of the INT program has been established in both prospective clinical trials and real-world studies. Clinical trials of the INT have demonstrated sustained improvements in HbA1c and weight loss, enabling dose reduction and complete deprescription of glucose-lowering medications and supporting diabetes remission over 2 years [22–24]. Deprescription supports patient preferences, reducing medication burden and hypoglycemic risk [23, 25–27]. These findings have subsequently been replicated in multiple real-world cohorts, demonstrating consistent improvements in glycemic control, weight, medication deprescription, and metabolic health across diverse populations, including veterans, rural and urban communities, women, and individuals across varying levels of socioeconomic deprivation and healthcare access [28–33].

Beyond these established clinical benefits, growing evidence suggests that participation in the INT program may also reduce healthcare utilization and costs. Studies conducted within the Veterans Health Administration, employer-sponsored populations, and other real-world healthcare settings have reported reductions in inpatient admissions, emergency department visits, healthcare expenditures, and prescription medication spending in the INT program [28, 34, 35]. However, these evaluations were conducted within individual healthcare systems, employer groups, or specific payer populations, limiting the generalizability of their findings to broader, more heterogeneous populations. As policymakers and payers assess reimbursement pathways for digital health interventions, robust evidence of their real-world economic value has become increasingly important.

To address this evidence gap, we evaluated the association between participation in the INT program and healthcare costs and utilization over 12 and 24 months using administrative claims from more than 13,000 adults with T2D and/or obesity. We hypothesized that participation in the INT program would be associated with lower healthcare costs and utilization than matched controls over both follow-up periods.

Methods

Study Design

We conducted a retrospective matched cohort study using a difference-in-differences (DID) design to estimate the impact of enrollment in a comprehensive digitally delivered nutrition-first care program compared with primary care-based condition management alone.

Data Sources

The study used longitudinal claims data from the Komodo Healthcare Map, a nationally representative database of open and closed medical and prescription claims for more than 300 million unique patients across commercial, Medicare, and Medicaid plans [36]. Datavant privacy-preserving record linkage (PPRL) allows for identification of unique records across payers. The study used two extracts from the Komodo Healthcare Map for the study period of 1 January 2016 through 20 August 2025. The first contained claims for INT participants identified through PPRL. The second contained claims for 2.4 million patients who did not participate in INT and had either two T2D or two obesity diagnosis codes > 30 days apart during the study period. The study investigators did not have access to the source data used to create the extracts. Only closed claims were used for the analysis. While healthcare encounters occurring after a patient leaves their health plan are not observable in closed claims data and reflect limitations inherent to commercially insured populations with high levels of plan churn, closed claims provide a complete record of reimbursed healthcare services within a patient’s health plan and thereby enable complete measurement of healthcare utilization and costs from the payer perspective. The identification of closed claims periods is described in Supplementary Material Exhibit S1.

This study was a retrospective analysis of deidentified administrative claims and program data licensed from KomodoHealth. The data was deidentified using the Expert Determination deidentification standard as specified by Health Insurance Portability and Accountability Act (HIPAA) Title 45 CFR Part 164.514b1. Because the data were deidentified and the study involved secondary analysis of existing records without direct participant contact or intervention, ethics committee approval and informed consent were not required. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) and RECORD-PE guidelines for observational cohort studies [37, 38].

Study Population

We constructed two populations: (1) patients who enrolled in the INT between January 2017 and March 2025 (the INT arm) and (2) patients who had a primary care physician (PCP) visit with T2D or obesity as a primary diagnosis but did not enroll in INT during the study period (the UC arm). The INT began treating patients in 217 and the study cohort was constructed in March 2025. The study period therefore reflects a comprehensive assessment of the INT’s impact. INT and UC patients’ index date was defined as the first day of the month of INT registration or the PCP visit. The baseline period was defined as the year preceding the index date.

Patients with T2D with or without obesity participate in the Diabetes Reversal (DR) program. Patients without T2D and with obesity are treated in the INT’s Sustainable Weight Loss (SWL) program. For simplicity, we refer to INT patients in the DR program and UC patients with a PCP visit for T2D management as the “cohort with T2D.” We refer to INT patients in the SWL program and UC patients with a PCP visit for obesity management as the “cohort with T2D,” reflecting each group’s primary clinical indication.

Analyses followed an intent-to-treat approach, with participants assigned to the INT or UC arms regardless of subsequent engagement with the INT program or PCP-based care.

Patients with type 1 diabetes, heart failure, stage 4+ chronic kidney disease, end-stage renal disease, or acute psychosis during the baseline year were ineligible for the INT and excluded from both UC and INT arms. Patients without continuous closed claims during the baseline year or 90 days after the index date (allowing gaps up to 30 days), those aged < 18 years at registration, or those with cancer or pregnancy during the baseline year were further excluded from both arms. In the cohort with T2D, we excluded individuals without a baseline T2D diagnosis. In the cohort with obesity, we excluded individuals with a diagnosis of T2D, with a claim for medications indicated solely for T2D, or without a baseline obesity diagnosis. Figure 1 summarizes the application of study inclusion and exclusion criteria.

Fig. 1.

Study sample construction flowchart. Note: The figure describes the sequential application of study inclusion and exclusion criteria to construct the INT and UC cohorts. The continuous claims coverage period was defined as the period between the first and last day of closed-source medical and pharmacy claims data availability, allowing gaps ≤ 30 days, as detailed in Supplementary Material, Exhibit S1. The index event was the first day of the month of either INT registration (for the INT arm) or a PCP visit with either type 2 diabetes or obesity as a primary diagnosis (for the UC arm). The baseline period was defined as the 365 days preceding the index date. Exclusion conditions were defined by the presence of an ICD-10-CM code indicating type 1 diabetes, CKD stages 4 + or ESRD, heart failure, acute psychosis, pregnancy, or cancer during the baseline period. The cohort with type 2 diabetes was required to have a diagnosis of type 2 diabetes during baseline. The cohort with obesity was required to have a diagnosis of obesity and no diagnosis of type 2 diabetes or use of insulin, sodium-glucose cotransporter 2 (SGLT2) inhibitors, sulfonylureas, thiazolidinediones, or DPP-4 inhibitors during baseline. Abbreviations: INT, individualized nutrition therapy; UC, usual care; PCP, primary care physician; ICD-10-CM, International Classification of Diseases, 10th Revision, Condition Modification; CKD, chronic kidney disease; ESRD, end-stage renal disease; SGLT-2, sodium-glucose cotransporter 2; DPP-4, dipeptyl peptidase-4

Exposure

The key exposure was enrollment in the INT. Comparator patients received UC and did not participate in the INT during the study period. The specific content of primary care received by the active comparator group (e.g., nutrition or exercise counseling, diabetes self-management education, or referral to other services) is not observable in claims data; accordingly, the UC cohort represents the range of care delivered in real-world clinical practice rather than a standardized intervention.

Patients are eligible for the INT if their health plan offers the program and if they have T2D, prediabetes, or obesity. Patients receive limited advertising about the INT and pay no out-of-pocket costs. Participation costs are paid by the health plan on an enrolled member-month basis.

INT participants receive telehealth-delivered continuous remote care from licensed physicians and nurse practitioners integrating individualized very low-carbohydrate nutrition support (targeting generally < 30 g/day), medication management, health coaching, peer support, and regular biometric feedback via a mobile application. Medication management consists of guideline-informed initiation, titration, and adjustment of diabetes- and weight-related medications on the basis of monitoring of glycemic control, weight, and other clinical indicators, consistent with standard of care. Among participants with T2D, the care model includes biomarker-informed, clinician-guided deprescription of glucose-lowering medications as glycemic control improves. Supplementary Material Table S2 reports program participation measures for the INT arm.

Outcomes

We identified ten outcomes relevant to the management of T2D and obesity. Cost outcomes were measured per member per month (PMPM) using medical and pharmacy claims and included costs of inpatient care, outpatient care, prescription medications, and total healthcare costs, defined as the sum of these three components.

For the cohort with T2D, nine additional cost outcomes included costs for each T2D medication (metformin, thiazolidinediones, GLP-1 receptor agonists, sodium-glucose cotransporter 2 (SGLT-2) inhibitors, sulfonylureas, dipeptyl peptidase-4 (DPP-4) inhibitors, and insulin) separately, all T2D medications combined, and all non–T2D medications combined. All outcomes were measured through the end of each patient’s closed claims coverage period, up to 24 months post-index.

Because INT program fees are paid through separate contractual agreements by participating payers and are not recorded in medical or pharmacy claims, they were not included in the primary claims-based cost outcomes. To estimate the total economic impact of the INT from the payer perspective, we conducted a secondary net savings analysis that incorporated INT program fees using Virta Clinic administrative records of program prices and individual enrollment dates.

Covariates

The index date was the first day of the month of registration for the INT (INT arm) or PCP visit for T2D or obesity management (UC arm). Baseline covariates included age at index date, sex, race, comorbidities (high cholesterol, hypertension, cardiovascular disease, chronic kidney disease, liver disease, and smoking), and baseline medication use (all T2D–indicated medications and lipid- and blood-pressure-lowering medications). Comorbidities and prescription medication use during the baseline year were identified by the presence of at least one claim with a corresponding International Classification of Diseases, 10th Revision (ICD-10) code (Supplementary Material Table S3) or medication name (Supplementary Material Table S4).

Statistical Analysis

Matched Control Group Construction

Propensity score matching was used to construct a comparison cohort with similar baseline characteristics as the intervention cohort, thereby reducing confounding from observed differences between groups. For the cohorts with T2D and obesity separately, control groups were constructed using 1:1 nearest-neighbor matching on a propensity score estimated using all demographic, clinical, medication use, and healthcare utilization variables available in administrative claims data, including sex, race, 5-year age groups, payer type, US region, year of index date, baseline comorbidities, baseline prescription medication use, and baseline inpatient, outpatient, and prescription medication cost quartiles. For the cohort with obesity, the propensity score included binary indicators of baseline use of glucagon-like peptide-1 receptor agonists (GLP-1RAs), other antiobesity medications, and metformin. For the cohort with T2D, the propensity score included baseline obesity and use of each T2D–indicated medication. Matching was conducted using the MatchIt package in R and all analyses were performed using R version 4.3.1.(39) All eligible INT participants were successfully matched to a UC participant, indicating adequate common support in baseline characteristics between the INT and UC populations. The reduction in the size of the usual care comparator group after matching reflects the exclusion of individuals who were not appropriate comparisons for individuals in the intervention group. This is an expected feature of propensity score matching. Covariate balance between the matched INT and UC cohorts was assessed using standardized mean differences (SMDs), with an absolute SMD < 0.1 considered indicative of adequate balance.

Estimation Framework

We used the matched INT and UC cohorts to estimate a DID regression model of the INT’s impacts. We selected a matched DID design with an external comparator cohort because it provides a stronger quasi-experimental framework than a single-arm before-after analysis. By comparing changes over time between the matched cohorts, this design helps account for secular trends, regression to the mean, and other time-related changes in healthcare utilization and expenditures—including the rapid increase in GLP-1 prescribing during the study period—that would not be addressed by a pre–post analysis alone.

The regression model included indicators for INT participation, the post-index period, and their interaction. Individual and calendar-month fixed effects were included to account for time-invariant individual characteristics and secular trends. Because intervention and control participants were first balanced using propensity score matching, baseline covariates were not included as additional regression adjusters. Instead, the individual fixed effects account for all stable participant characteristics, including measured and unmeasured characteristics that do not vary over time. The interaction term estimated the differential change in outcomes associated with INT participation. Robust standard errors were clustered at the individual level.

The estimation sample was limited to the 12 months prior to, and 12 or 24 months following, each patient’s index date. We estimated the model for the cohort with T2D and the cohort with obesity separately. For the cohort with T2D, we additionally estimated the model for the secondary outcomes specified above. Because this study evaluated observed healthcare expenditures over a maximum 24-month follow-up period without projecting future costs, healthcare costs were not discounted.

In this study, individuals in the comparison arm initiate a comparator intervention (PCP-based chronic condition management). In both the INT and UC arms, the post-period is defined relative to each patient’s initiation date. Accordingly, we used a classical 2 × 2 DID specification with indicators for study arm, post-period, and their interaction. We did not use a traditional two-way fixed-effects DID model, in which members of the treated group receive treatment in different time periods but the comparison group does not receive a comparator treatment and therefore lacks an analogous index date and pre-/post-index periods [40].

The DID model’s key identifying assumption is that, in the absence of treatment, outcomes would have evolved similarly over time in the INT and UC arms after matching. We assessed the validity of this assumption with (1) tests of differential baseline linear trends and (2) visual assessment of pre-intervention trends using an event-study specification estimating monthly coefficients relative to the month before the index date. To assess residual confounding, we conducted a negative control outcome analysis comparing the INT's impact on specific T2D-indicated medication costs with its impacts on specific non-T2D-indicated medication costs. Because the INT’s care model only deprescribes T2D-indicated medications, we hypothesize that the INT should not impact non-T2D-indicated medications; evidence of such impacts would suggest residual confounding.

Net Savings Analysis

To estimate net cost impacts from the payer perspective, we compared the DID estimate of PMPM claims-based cost reductions with PMPM INT program costs obtained from Virta Clinic administrative billing records containing monthly program fees and enrollment dates. Because program fees are not observable in claims data, the net savings analysis was conducted separately from the primary claims-based DID analysis.

Results

Baseline Sample Characteristics

A total of 6580 INT and 94,233 UC patients met the study eligibility criteria. The matched study sample contained 6580 INT and 6580 UC patients: 3819 patients per arm in the cohort with T2D and 2761 patients per arm in the cohort with obesity. Table 1 presents the baseline characteristics of each cohort. The cohort with T2D averaged 55 years old at index, 52% male, and 49% white. The cohort with obesity averaged 48 years old, 29% male, and 52% white. Baseline medications, comorbidities, demographics and cost quartiles were comparable between the INT and UC arms. Supplementary Material Tables S5 and S6 describe the characteristics of the INT and UC arms before and after matching for each cohort.

Table 1.

Study sample characteristics for matched cohorts with obesity and type 2 diabetes

Characteristic	Cohort with obesity			Cohort with type 2 diabetes
	INT* N = 2761	UC† N = 2761	SMD	INT* N = 3819	UC† N = 3819	SMD
Age at index date, mean (SD), years	48 (10.2)	48 (11.8)	0.03	55 (8.7)	55 (10.5)	0.01
Race			0.04			0.02
White	1436 (52.0%)	1396 (50.6%)		1880 (49.2%)	1887 (49.4%)
Black or African American	352 (12.7%)	335 (12.1%)		423 (11.1%)	404 (10.6%)
Hispanic or Latino	235 (8.5%)	251 (9.1%)		342 (9.0%)	343 (9.0%)
Asian or Pacific Islander	46 (1.7%)	47 (1.7%)		157 (4.1%)	158 (4.1%)
Other	47 (1.7%)	49 (1.8%)		89 (2.3%)	84 (2.2%)
Missing	645 (23.4%)	683 (24.7%)		928 (24.3%)	943 (24.7%)
Sex			0.04			0.01
Female	1973 (71.5%)	1940 (70.3%)		1836 (48.1%)	1813 (47.5%)
Male	769 (27.9%)	807 (29.2%)		1961 (51.3%)	1985 (52.0%)
Missing	19 (0.7%)	14 (0.5%)		22 (0.6%)	21 (0.5%)
US region			0.04			0.01
Midwest	1076 (39.0%)	1070 (38.8%)		1284 (33.6%)	1274 (33.4%)
Northeast	211 (7.6%)	233 (8.4%)		485 (12.7%)	495 (13.0%)
South	1247 (45.2%)	1247 (45.2%)		1569 (41.1%)	1559 (40.8%)
West	226 (8.2%)	209 (7.6%)		481 (12.6%)	491 (12.9%)
Missing	1 (0.0%)	2 (0.1%)		0 (0.0%)	0 (0.0%)
Payer type			0.07			0.02
Commercial	2658 (96.3%)	2687 (97.3%)		3452 (90.4%)	3449 (90.3%)
Medicaid	63 (2.3%)	41 (1.5%)		97 (2.5%)	109 (2.9%)
Medicare	38 (1.4%)	30 (1.1%)		268 (7.0%)	259 (6.8%)
Missing	2 (0.1%)	3 (0.1%)		2 (0.1%)	2 (0.1%)
Baseline comorbidities
Obesity	2761 (100.0%)	2761 (100.0%)	0.00	1961 (51.3%)	1966 (51.5%)	0.00
Hypertension	1073 (38.9%)	1041 (37.7%)	0.02	2549 (66.7%)	2596 (68.0%)	− 0.03
High cholesterol	997 (36.1%)	951 (34.4%)	0.03	2707 (70.9%)	2722 (71.3%)	− 0.01
Cardiovascular disease	263 (9.5%)	257 (9.3%)	0.01	484 (12.7%)	453 (11.9%)	0.02
Metabolic dysfunction-associated steatotic liver disease	164 (5.9%)	189 (6.8%)	−0.04	327 (8.6%)	326 (8.5%)	0.00
Chronic kidney disease, stages 1–3	42 (1.5%)	39 (1.4%)	0.01	142 (3.7%)	142 (3.7%)	0.00
History of smoking	32 (1.2%)	31 (1.1%)	0.00	48 (1.3%)	35 (0.9%)	0.03
Baseline prescription medications
Diabetes or obesity medications
DPP-4 inhibitor	–	–	0.00	393 (10.3%)	406 (10.6%)	−0.01
GLP-1 receptor agonist	356 (12.9%)	356 (12.9%)	0.00	1464 (38.3%)	1465 (38.4%)	0.00
Insulin	–	–	0.00	639 (16.7%)	634 (16.6%)	0.00
Metformin	203 (7.4%)	203 (7.4%)	0.00	2857 (74.8%)	2834 (74.2%)	0.01
SGLT-2 inhibitor	–	–	0.00	916 (24.0%)	917 (24.0%)	0.00
Sulfonylurea	–	–	0.00	759 (19.9%)	790 (20.7%)	−0.02
Thiazolidinedione	–	–	0.00	209 (5.5%)	230 (6.0%)	−0.02
Other medications
Beta blocker	271 (9.8%)	273 (9.9%)	0.00	690 (18.1%)	717 (18.8%)	−0.02
Calcium channel blocker	311 (11.3%)	323 (11.7%)	−0.01	666 (17.4%)	671 (17.6%)	0.00
Diuretic	482 (17.5%)	462 (16.7%)	0.02	1070 (28.0%)	1110 (29.1%)	−0.02
Mineralocorticoid receptor agonist	85 (3.1%)	83 (3.0%)	0.00	117 (3.1%)	83 (2.2%)	0.06
RAAS inhibitor	709 (25.7%)	678 (24.6%)	0.03	2341 (61.3%)	2359 (61.8%)	−0.01
Statin	535 (19.4%)	514 (18.6%)	0.02	2546 (66.7%)	2546 (66.7%)	0.00
Baseline cost categories
Any inpatient utilization	100 (3.6%)	113 (4.1%)	0.02	157 (4.1%)	154 (4.0%)	0.00
Outpatient cost quartile‡			0.02			0.02
First (lowest)	492 (17.8%)	495 (17.9%)		781 (20.5%)	794 (20.8%)
Second	705 (25.5%)	722 (26.1%)		1047 (27.4%)	1053 (27.6%)
Third	779 (28.2%)	759 (27.5%)		990 (25.9%)	1007 (26.4%)
Fourth (highest)	785 (28.4%)	785 (28.4%)		1001 (26.2%)	965 (25.3%)
Prescription medication cost quartile‡			0.10			0.02
First (lowest)	795 (28.8%)	870 (31.5%)		352 (9.2%)	370 (9.7%)
Second	918 (33.2%)	959 (34.7%)		719 (18.8%)	739 (19.4%)
Third	664 (24.0%)	617 (22.3%)		853 (22.3%)	840 (22.0%)
Fourth (highest)	384 (13.9%)	315 (11.4%)		1895 (49.6%)	1870 (49.0%)
Baseline costs, $ per member per year
Total§	13,298 (25,726)	11,110 (21,506)	0.09	15,218 (24,705)	14,815 (31,499)	0.01
Inpatient	1351 (9351)	1401 (8497)	−0.01	1375 (8700)	1408 (11,149)	0.00
Outpatient	9412 (19,997)	7983 (16,861)	0.08	7612 (16,809)	7154 (22,964)	0.02
Prescription medications	2535 (10,138)	1726 (6566)	0.09	6230 (11,785)	6254 (15,848)	0.00
Year of index date			0.40			0.40
2017	0 (0.0%)	1 (0.0%)		3 (0.1%)	114 (3.0%)
2018	0 (0.0%)	4 (0.1%)		13 (0.3%)	129 (3.4%)
2019	0 (0.0%)	15 (0.5%)		188 (4.9%)	189 (4.9%)
2020	16 (0.6%)	31 (1.1%)		437 (11.4%)	280 (7.3%)
2021	54 (2.0%)	91 (3.3%)		590 (15.4%)	433 (11.3%)
2022	326 (11.8%)	257 (9.3%)		918 (24.0%)	667 (17.5%)
2023	478 (17.3%)	648 (23.5%)		663 (17.4%)	730 (19.1%)
2024	1689 (61.2%)	1276 (46.2%)		920 (24.1%)	970 (25.4%)
2025	198 (7.2%)	438 (15.9%)		87 (2.3%)	284 (7.4%)
Closed claims follow-up
> 6 months	2436 (88.2%)	2223 (80.5%)	0.21	3569 (93.5%)	3365 (88.1%)	0.19
> 12 months	1396 (50.6%)	1215 (44.0%)	0.13	2702 (70.8%)	2363 (61.9%)	0.19
> 24 months	427 (15.5%)	393 (14.2%)	0.03	1577 (41.3%)	1243 (32.5%)	0.18
Mean # days	449 (273.9)	426 (314.9)	0.08	728 (477.1)	660 (532.3)	0.13

^* Individualized Nutrition Therapy

^† Usual care

^‡ Quartiles for baseline outpatient and prescription medication costs were defined as the percentile relative to the pooled sample: < 25th (quartile 1), 26th–49th (quartile 2), 50th–74th (quartile 3), or > 75th (quartile 4)

^§ Total cost is defined as the sum of inpatient, outpatient, and prescription medication costs

Impacts

Table 2 reports DID estimates by cohort. The intervention was associated with significantly lower healthcare utilization and costs in each cohort. Figure 2 reports event-study DID estimates for the 12 pre- and post-intervention months, confirming the baseline parallel trends assumption.

Table 2.

Impacts on cost and utilization for cohorts with type 2 diabetes and obesity

Cohort with T2D	DID estimate (95% CI)*		Baseline mean		p, baseline trend difference‖
	1 year (n = 170,202)	2 years (n = 217,789)	UC†	INT‡
Costs, $ per member per month
Total§	−240.09*** (−357.17, −123.01)	−230.25*** (−340.36, −120.14)	1076.62	1139.61	0.74
Inpatient	−97.97** (−161.74, −34.20)	−92.77** (−150.16, −35.38)	115.54	111.45	0.87
Outpatient	18.28 (−53.23, 89.79)	4.91 (−61.34, 71.16)	418.57	489.11	0.69
All prescription medications	−160.39*** (−210.43, −110.35)	−142.40*** (−194.13, −90.67)	542.51	539.06	0.28
T2D-related	−145.16*** (−161.70, −128.62)	−135.71*** (−152.15, −119.27)	352.02	359.37	0.26
Non-T2D-related	−15.23 (−62.04, 31.58)	−6.69 (−55.33, 41.95)	190.49	179.69	0.19
p, joint baseline trend difference					0.53
Utilization, visits per 1000 members per month
Inpatient	−3.2*** (−5.0, −1.4)	−3.4*** (−5.1, −1.8)	4.7	4.3	0.94
ED	−0.02 (−3.3, 3.3)	−0.1 (−3.2, 3.0)	26.0	19.7	0.47
PCP	−60.6*** (−72.4, −48.8)	−50.2*** (−61.6, −38.7)	228.6	258.9	0.97
Cardiologist	−0.6 (−2.7, 1.4)	−0.2 (−2.1, 1.7)	8.6	6.5	0.86
Endocrinologist	−4.1*** (−6.0, −2.1)	−2.9** (−4.8, −1.0)	11.4	10.1	0.76
p, joint baseline trend difference					0.99

Cohort with obesity	DID estimate (95% CI)		Baseline mean		p, baseline trend difference‖
	1 year (n = 117,029)	2 years (n = 136,471)	UC	INT
Costs, $ per member per month (detrended)
Total§	−256.04*** (−374.55, −137.53)	−189.35** (−301.32, −77.38)	780.02	952.64	0.81
Inpatient	−116.59*** (−174.83, −58.34)	−92.04** (−146.38, −37.69)	112.59	109.29	0.96
Outpatient	−99.28* (−191.82, −6.75)	−72.09 (−158.35, 14.18)	521.40	647.98	0.79
All prescription medications	−40.17** (−66.37, −13.97)	−25.23 (−52.67, 2.21)	146.02	195.36	0.98
All antiobesity medications	−15.48* (−30.23, −0.73)	4.44 (−9.37, 18.24)	43.38	7.34	0.84
GLP-1 receptor agonist	−15.42* (−30.11, −0.73)	4.61 (−9.15, 18.36)	43.30	7.31	0.84
Non-antiobesity medications	−24.69* (−46.54, −2.85)	−29.67* (−53.49, −5.84)	102.64	188.02	0.94
p, joint baseline trend difference					0.99
Utilization, visits per 1000 members per month
Inpatient	−4.8*** (−6.8, −2.8)	−3.8*** (−5.6, −1.9)	4.5	4.5	0.13
ED	−5.3** (−9.2, −1.4)	−4.9* (−8.7, −1.1)	22.7	24.9	0.07
PCP	−74.7*** (−90.1, −59.3)	−65.6*** (−80.1, −51.0)	248.4	277.1	0.00**
Cardiologist	−2.5* (−4.5, −0.4)	−1.9 (−3.7, 0.02)	7.1	5.0	0.30
Endocrinologist	−2.6** (−4.2, −1.0)	−2.1** (−3.6, −0.6)	4.9	4.0	0.80
p, joint baseline trend difference					0.01

^***p < 0.001, ^**p < 0.01, ^*p < 0.05

^* Differences-in-differences estimates and 95% CIs are reported for the interaction term between the post-intervention term and the INT participation term. Estimates were derived from a regression model of person-month outcomes adjusted for person and calendar-month fixed effects. Standard errors are clustered at the person level

^† Usual care

^‡ Individualized Nutrition Therapy

^§ Total cost is defined as the sum of inpatient, outpatient, and prescription medication costs

^‖ The baseline trend difference is the interaction term between the indicator for INT participation and a linear month variable indicating months 1–12 during the baseline year, estimated using baseline year data only

^¶ A χ² test of the joint significance of the baseline trend difference was estimated for (1) all cost outcomes and (2) all utilization outcomes

Fig. 2.

Event-study DID estimates for key outcomes. Note: The figure displays event-study difference-in-differences estimates and 95% confidence intervals for the 12 months before and after the index month. Estimates correspond to coefficients on the interaction terms between INT participation and relative-month indicators, with month −1 as the reference category. The y-axis represents allowed costs per member per month. Month 0 is defined as the first month of INT registration or a PCP visit for type 2 diabetes/obesity management. Blue points denote post-index months; red points denote pretreatment months. Estimates are obtained from individual-level monthly regressions including individual and calendar-month fixed effects. Standard errors are clustered at the member level. Abbreviations: INT, individualized nutrition therapy; PCP, primary care physician

Cohort with T2D

Cost

Tests of pre-intervention trend differences were not statistically significant for any individual cost outcome or for the joint test across all cost outcomes, supporting the parallel trends assumption. In year 1, total healthcare costs were reduced by $240 PMPM (95% CI $123–357). The reduction was driven by inpatient costs ($98 PMPM [95% CI $34–162]) and T2D-indicated medications ($145 PMPM [95% CI $128–161]). Impacts on outpatient or non-T2D-indicated medication costs were not statistically significant.

Table 3 reports impacts by T2D medication. Cost reductions were observed across all T2D medication classes, including SGLT-2 inhibitors ($54 PMPM), GLP-1 receptor agonists ($59 PMPM), and insulin ($22 PMPM) with statistically significant, smaller impacts for metformin, DPP-4 inhibitors, and sulfonylureas, and an insignificant impact for thiazolidinediones. Figure 3 reports estimates from the negative control outcomes analysis. Estimates of the INT’s impact on non-T2D-indicated medications were consistently small in magnitude and not statistically significant.

Table 3.

Impacts on cost and utilization of specific medications for cohort with type 2 diabetes

T2D medication	DID estimate (95% CI)*		Baseline mean		Relative change§
	1 year	2 years	UC†	INT‡
Cost, $ per member per month
Metformin	−5.30*** (−8.17, 2.43)	−5.75*** (−8.63, −2.88)	18.35	18.04	−29.4%
DPP-4 inhibitor	−4.77** (−7.88, −1.66)	−4.72** (−8.01, −1.43)	26.20	26.37	−18.1%
GLP-1 receptor agonist	−58.90*** (−71.67, −46.13)	−50.83*** (−63.42, −38.24)	181.25	182.93	−32.2%
Thiazolidinedione	0.06 (−0.06, 0.18)	0.03 (−0.09, 0.15)	0.34	0.26	23.3%
Sulfonylurea	−0.38*** (−0.46, −0.30)	−0.38*** (−0.46, −0.30)	0.69	0.73	−51.7%
SGLT2 inhibitor	−53.50*** (−60.09, −46.92)	−51.32*** (−58.00, −44.64)	80.10	80.04	−66.8%
Insulin	−22.38*** (−27.71, −17.04)	−22.73*** (−28.19, −17.27)	45.09	51.00	−43.9%

^***p < 0.001, ^**p < 0.01, ^*p < 0.05

^*Differences-in-differences estimates and 95% CIs are reported for the interaction term between the post-intervention term and the INT participation term. Estimates were derived from a regression model of person-month outcomes adjusted for person and calendar-month fixed effects. Standard errors are clustered at the person level

^† Usual care

^‡ Individualized Nutrition Therapy

^§ The relative change is the DID estimate of the cost reduction for the specified medication divided by the INT group’s baseline mean cost for that medication

Fig. 3.

Negative control outcomes analysis; ^***p < 0.001, ^**p < 0.01, ^*p < 0.05. Note: The figure displays difference-in-differences estimates of the INT’s impact on specific categories of prescription medication costs over 1 year and 2 years. Estimates correspond to the coefficient on the interaction terms between INT participation and post-period indicators over 1 year and 2 years. Error bars indicate 95% confidence intervals. The y-axis represents allowed costs per member per month for the specified prescription medication category. Estimates are obtained from individual-level monthly regressions including individual and calendar-month fixed effects. Standard errors are clustered at the member level. Abbreviations: INT, individualized nutrition therapy; GLP-1 s, glucagon-like peptide-1 receptor agonists; SGLT-2i, sodium-glucose cotransporter 2 inhibitors; DPP-4, dipeptyl peptidase-4; TZDs, thiazolidinediones; AOMs, antiobesity medications; RAAS, renin–angiotensin–aldosterone system

Follow-up averaged 10.8 months in year 1 and 17.0 months over 2 years. Over 2 years, total costs were reduced by $230 PMPM, indicating sustained impacts. Table 4 compares total INT program fees against estimated cost reductions over 1 year and 2 years. Per-member net savings were $929 in year 1 and $1374 over 2 years.

Table 4.

Net savings analysis

Cohort	Year	Net price (PMPM)*	Months active†	Total program cost	DID estimate, cost reduction (PMPM)	Months of follow-up‡	Cost reduction before fees§	Net savings ‖	ROI¶
Type 2 diabetes	1	$183	9.5	$1663	$240	10.8	$2592	$929	1.56
	2	$183	14.6	$2536	$230	17.0	$3910	$1374	1.54
Obesity	1	$137	6.8	$897	$256	10.0	$2560	$1663	1.85
	2	$137	8.6	$1118	$189	13.3	$2513	$1395	1.24

^*The net price is the average per-member, per-month price over the 10-year study period, weighted by the program start year in the matched INT sample, after performance guarantee refunds and discounts

^† Months active indicates the number of months INT participants were registered in the INT within the 1- or 2-year measurement period, regardless of engagement level

^‡ Months of follow-up indicates the number of months INT participants had continuous closed claims data after registration within the 1- or 2-year measurement period

^§ The cost reduction before fees is the DID estimate for the total cost of care multiplied by the months of follow-up

^‖ Net savings are the difference between the cost reduction before fees and the total program cost

^¶ Return on investment is defined as cost reduction before fees divided by total program cost

Utilization

Tests of pre-intervention trend differences were not statistically significant for any individual utilization outcome or for the joint test across all utilization outcomes, supporting the parallel trends assumption. Compared with matched controls, INT participation was associated with lower healthcare utilization, primarily driven by reductions in inpatient, endocrinology, and primary care visits, with no significant differences in emergency department or cardiology visits. In year 1, participants had 3.2 (95% CI 1.4–5.0) fewer inpatient visits, 4.1 (95% CI 2.1–6.0) fewer endocrinologist visits, and 60.6 (95% CI 48.8–72.4) fewer PCP visits per 1000 members per month (PKMPM). Over 2 years, participants had reductions of 3.4 (95% CI 1.8–5.1) inpatient, 2.8 (95% CI 1.0–4.8) endocrinologist, and 50.2 (95% CI 38.7–61.6) fewer PCP visits PKMPM. Impacts on ED and cardiologist visits were not statistically significant.

Cohort with obesity

Cost

In our standard DID specification, there was a significant difference in the baseline antiobesity medication cost trend, driven entirely by a larger increase in GLP-1 receptor agonist costs in the UC arm. To mitigate the impact of the differential trend, for each outcome, we estimated the baseline trend difference, detrended the outcome through year 2, and estimated the model on the detrended outcome. After detrending, tests of pre-intervention trend differences were not statistically significant for any individual cost outcome or for the joint test across all cost outcomes, supporting the parallel trends assumption. Results for unadjusted outcomes are presented in Supplementary Material (Table S7). Additionally, substantial attrition was observed between years 1 and 2 in the cohort with obesity: mean follow-up was 10.0 months in the first post-index year but 13.3 months across 2 years.

In year 1, total healthcare costs were reduced by $256 PMPM (95% CI $138–375). The reduction was driven by inpatient ($117 PMPM [95% CI $58–175]), outpatient ($99 PMPM [95% CI $7–192]), and prescription medication costs ($40 PMPM [95% CI $14–66]).

Over 2 years, total costs were $189 PMPM (95% CI $77–301) lower than baseline, indicating sustained cost reductions. Table 4 compares total INT program costs against estimated claims-based cost reductions, indicating $1395 in per-member net savings over 2 years.

Utilization

Baseline trend estimates for low-frequency utilization outcomes are imprecise, and detrending these outcomes may amplify noise or lead to poorly extrapolated results. We therefore present utilization outcomes unadjusted. Compared with matched controls, INT participation was associated with lower utilization across all measured healthcare settings during the first year, with sustained reductions in inpatient and primary care visits over 2 years. Participants had 4.8 (95% CI 2.8–6.8) fewer inpatient, 5.3 (95% CI 1.4–9.2) fewer ED, and 74.7 (95% CI 59.3–90.1) fewer PCP visits PKMPM. Additionally, participants had 2.5 (95% CI 0.4–4.5) and 2.6 (95% CI 0.9–4.2) fewer cardiologist and endocrinologist visits, respectively. Over 2 years, participants had 3.8 (95% CI 1.9–5.6) fewer inpatient and 65.6 (95% CI 51.0–80.1) fewer PCP visits PKMPM.

Discussion

In this large claims-based study of 6580 adults with T2D or obesity and 6580 matched controls, participation in a clinically effective telehealth nutrition program was associated with lower total cost of care and utilization, generating approximately $1400 per member in net savings over 2 years. These findings suggest that intensive digital nutrition care can achieve both clinical effectiveness and meaningful cost reductions. As telehealth-delivered nutrition interventions become increasingly incorporated into routine diabetes and obesity care, evidence on their effects on healthcare utilization and costs can inform clinician referral decisions, health system implementation strategies, and reimbursement policies alongside evidence of clinical effectiveness.

Impacts of the INT differed by clinical indication. Among participants with T2D, savings were driven by reductions in T2D medication spending. This finding is consistent with the INT’s nutrition-first care model, which emphasizes improving metabolic health through carbohydrate reduction to reduce medication dependence through structured deprescription, and in some individuals, achieve diabetes remission [22, 24, 30]. In the negative control outcomes analysis, the absence of an association with non-T2D-indicated medication costs further supports the interpretation that reductions in medication spending reflect the care model’s clinician-guided deprescription rather than nonadherence or residual confounding.

The observed medication reductions also align with a 2-year evaluation of a pilot of the INT at the Veterans Health Administration, which demonstrated sustained reductions in T2D-related medication use and costs following the intervention. More broadly, telehealth-based T2D self-management and lifestyle interventions have demonstrated improvements in glycemic control, and in some cases, reductions in healthcare utilization and costs [41–47]. However, the magnitude and durability of these economic benefits have generally been modest. In contrast, the present study demonstrated reductions in total healthcare costs over both 12 and 24 months. These findings suggest that integrating a nutrition-first approach into chronic disease management may provide a clinically effective and cost-saving complement to traditional primary care.

In the cohort with obesity, savings accrued across a broad set of cost domains, consistent with prior evaluations of digital weight management and diabetes prevention programs demonstrating reductions in inpatient and outpatient utilization, prescription medication costs, and overall healthcare costs [44, 45, 48–51]. The present study extends these findings by demonstrating sustained reductions across inpatient, outpatient, and prescription spending over 2 years. While differences in study design preclude direct comparisons, the magnitude and durability of the observed cost reductions were greater than those reported in prior digital lifestyle interventions, which generally demonstrated smaller overall savings over shorter follow-up periods [52, 53].

Although use of GLP-1 receptor agonists for weight loss increased substantially during the study period, INT and UC participants had comparable index year distributions, minimizing potential bias from secular trends in GLP-1 use. Relative to the UC arm, INT participation was associated with a modest reduction in GLP-1 spending of approximately $15 PMPM, consistent with the INT’s nutrition-first model. Multiple large claims-based evaluations have found that GLP-1s, despite annual medication costs of $12,000–16,000 per patient [54] have not demonstrated reductions in non-GLP-1 healthcare costs within 1–5 years [55–58]. Together, these findings suggest that the observed per-member net savings of approximately $1600 is more likely attributable to the INT’s comprehensive care model than to modest differences in GLP-1 use alone.

Although the present study demonstrates significant reductions in healthcare costs associated with participation in the INT program over 1 year and 2 years, formal cost-effectiveness analyses require integration of healthcare costs with clinical outcomes or quality-of-life measures, which are not available in administrative claims data. Although these outcomes could not be evaluated in the present study, multiple prospective clinical trials and real-world evaluations of the INT have demonstrated improvements in glycemic control, weight, lipid profiles, liver enzymes, type 2 diabetes remission, diabetes medication deprescription, and liver and kidney disease risk [30, 31, 59, 60]. These findings provide important clinical context for interpreting the observed reductions in healthcare costs reported here. Future studies linking claims data with clinical or patient-reported outcomes will enable formal cost-effectiveness modeling to further assess the economic value of the INT. Formal budget impact analyses can estimate the financial implications of broader adoption of the INT at the health system or plan level. Lastly, because diabetes and obesity are chronic conditions, sustained improvements in glycemic control, weight, medication use, and hospitalization could plausibly translate into continued reductions in healthcare expenditures. The long-term economic impact of the intervention is an important area for future research.

The study has limitations. First, as with all claims-based analyses, relevant clinical variables such as diabetes duration, glycemic control, body weight, and receipt of other nutrition or lifestyle supports were unavailable and therefore could not be incorporated into the propensity score or regression models. Although propensity score matching and individual fixed-effects difference-in-differences analyses reduce confounding from observed baseline characteristics and stable participant characteristics, residual confounding from unmeasured time-varying clinical factors remains possible. We conduct three assessments of the propensity score matched DID model’s assumptions: (1) outcome-specific baseline parallel trends tests, (2) joint baseline parallel trends tests, and (3) event-study specifications. We further augment the study design with (4) detrended alternate specifications to mitigate the impacts of unobserved confounding in the cohort with obesity and (5) negative control outcome analyses for the prescription medication outcomes in the cohort with T2D. Future studies incorporating clinically relevant measures such as diabetes duration, glycemic control, body weight, and other lifestyle changes or supports would further strengthen the evaluation of the program’s effectiveness and economic impact.

Second, although mean follow-up time and retention at 6, 12, and 24 months were comparable between the INT and matched comparison groups, attrition beyond 1 year was substantial, particularly in the obesity cohort. Conditioning analyses on longer follow-up may limit generalizability of the results, given the churn inherent to commercially insured populations and could introduce immortal time bias if INT participation influenced employment duration. In addition, estimates of healthcare cost savings were derived from claims data and were therefore affected by loss to follow-up, whereas estimates of INT participation costs were obtained from Virta Clinic administrative records and thus unaffected by loss of follow-up in the claims data. This asymmetry likely biases net savings estimates toward the null. Accordingly, we report results over both 1- and 2-year follow-up periods, recognizing that the 2-year estimates reflect the data limitations inherent to commercially insured populations and the evidence available to inform commercial reimbursement decisions.

Third, digitally delivered lifestyle programs are widespread: in 2023, 85% of large employers offered disease management programs [61]. Because these programs are typically paid through direct employer–vendor contracts and not captured in claims data, we cannot rule out potential contamination of the UC arm, which would bias impact estimates toward zero.

Fourth, the study design limits generalizability of the results to several populations. The study excluded 15.3% of INT participants with T2D and 57.6% with obesity who lacked the corresponding qualifying claims-based diagnoses during the baseline year, consistent with known under-recording of obesity in claims data [62]. Linked claims and EHR data are needed to assess the generalizability of these findings to INT participants with solely BMI-defined obesity, as detailed in Supplementary Material (Table S8). The UC group was defined as adults with a primary care visit for T2D or obesity, reflecting the most common setting in which these conditions are managed in routine clinical practice. This approach excludes patients managed exclusively by specialists, limiting the generalizability of the study’s findings to other care pathways. Because most INT participants were commercially insured and propensity score matching selects UC patients with similar observable characteristics, including age and insurance type, the matched cohort was also largely commercially insured. Consequently, these findings may not generalize to populations with substantially different proportions of Medicare or Medicaid beneficiaries. Broader evidence is warranted, particularly given recent expansions in Medicare coverage for technology-supported care models. Lastly, although the lower proportion of male participants in the INT is consistent with prior studies showing lower rates of male participation in preventative health programs such as the Diabetes Prevention Program [63], less interest in structured lifestyle programs [64], and less representation in weight loss clinical trials [65], this may limit the generalizability of the findings to male populations.

Despite these limitations, this study has several strengths, including evaluation of observed healthcare expenditures using adjudicated claims rather than modeled cost projections, one of the largest cohorts evaluated to date for a telehealth-delivered nutrition intervention in adults with T2D or obesity, and a rigorous quasi-experimental design incorporating propensity score matching, assessment of both individual and joint baseline parallel trends assumptions, and negative control outcome analyses to evaluate the potential for residual confounding.

Conclusions

In this large, propensity score-matched DID analysis of commercially insured adults with T2D or obesity, enrollment in a comprehensive nutrition-first digital care model was associated with lower healthcare costs and reduced healthcare utilization over 12 and 24 months relative to routine primary care management.

These findings extend prior evidence from clinical trials and health system-specific evaluations by demonstrating the economic impact of the INT program in a large, geographically diverse real-world population. As healthcare systems and payers seek scalable approaches to address the growing clinical and economic burden of diabetes and obesity, comprehensive telehealth-delivered nutritional care may represent an effective strategy for reducing healthcare costs as a complement to existing chronic disease management strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Study conception, design, and statistical analysis were performed by Priya Shanmugam. The first draft of the manuscript was written by Priya Shanmugam, and all authors commented on previous versions of the manuscript. Priya Shanmugam, Shaminie Athinarayanan, Adam Wolfberg, Rebecca Adams, and Jeromie Ballreich critically revised the manuscript. Rebecca Adams and Jeromie Ballreich provided supervision. All authors read and approved the final manuscript.

Funding

The study was funded by Virta Health. The journal’s Rapid Service Fee was funded by Virta Health.

Data Availability

The datasets generated during and/or analyzed during the current study are not publicly available because they were licensed from KomodoHealth.

Declarations

Conflict of Interest

Priya Shanmugam, Shaminie Athinarayanan, and Adam Wolfberg are employees of Virta Health and hold stock or stock options in the company. Rebecca Adams is a former employee of Virta Health and holds stock or stock options in the company. Jeromie Ballreich received consulting fees from Virta Health for his contribution to the work.

Ethical Approval

This study was a retrospective analysis of deidentified administrative claims and program data licensed from KomodoHealth. The data was deidentified using the Expert Determination deidentification standard as specified by Health Insurance Portability and Accountability Act (HIPAA) Title 45 CFR Part 164.514b1. Because the data were deidentified and the study involved secondary analysis of existing records without direct participant contact or intervention, ethics committee approval and informed consent were not required.