Association Between Real-Time Continuous Glucose Monitor Use and Diabetes-Related Medical Costs for Patients with Type 2 Diabetes

Introduction

The impact of diabetes is a significant and growing health concern. In the United States, the prevalence of diabetes is estimated to be >34.2 million, the majority (90%–95%) of whom have type 2 diabetes (T2D). ¹ As recently reported by Lin et al., the prevalence of diabetes could rise as high as 44.6 million by 2030. ² Moreover, a significant proportion of individuals with diabetes are not achieving their glycemic targets, ^3,4 resulting in the development and progression of the debilitating chronic complications of diabetes. ^{5 –8}

In addition to the human impact of this disease, the increasing costs of providing diabetes care continues to stress the U.S. health care system. According to the latest estimates, ⁹ the cost of diabetes in the United States in 2017 was $327 billion, which includes $237 billion spent on direct medical care and $90 billion in reduced productivity. ⁹ A large percentage of these costs result from mostly avoidable hospitalizations and emergency department utilizations subsequent to diabetes-related adverse events. ¹⁰

During the past decade, an increasing number of individuals with type 1 diabetes (T1D) and T2D have adopted real-time continuous glucose monitoring (rtCGM) for their daily diabetes self-management. As demonstrated in numerous randomized controlled trials (RCTs) and real-world retrospective and prospective observational studies, use of rtCGM facilitates reductions in glycated hemoglobin (HbA1c), ^{11 –17} increased time in target glucose range ^{11 –17} and reductions in hypoglycemia risk, ¹⁸ with corresponding reductions in diabetes-related hospitalizations. ^16,18

We report findings from analyses of administrative claims data on the association between rtCGM use in individuals with T2D and diabetes-related health care resource utilization costs.

Methods

Results

Data from 571 T2D patients who met study inclusion criteria were assessed. The mean age was 51.2 ± 11.9 years, and 265 (46%) were female. Most patients were treated with bolus insulin, with or without basal insulin 454 (80%), 58 (10%) were treated with basal but not bolus insulin, and 59 (10%) were not insulin treated. Nearly all patients (99%) were covered by commercial health insurance, 58% were treated by an endocrinologist and 92% had associated diabetes complications. In addition, the most common nondiabetic-related conditions were hypertension (70%), disorders of lipid metabolism (78%), diseases of the urinary system (44%), and diseases of the heart (44%).

Average baseline diabetes-related health care resource utilization costs were $1,680 ± 4,519 PPPM and were $1,256 ± 3,679 PPPM for the follow-up period. As a result, average diabetes-related health care resource utilization costs decreased by −$424 PPPM (95% CI −$816 to −$31, P = 0.035) after initiating rtCGM treatment.

Reductions in diabetes-related medical costs were driven primary by costs associated with inpatient hospitalization admissions (−$358 PPPM, 95% CI −$706 to −$10, P = 0.044) (Fig. 1). Inpatient hospital admissions were reduced on average by −0.006 PPPM (∼7 inpatient hospital admissions/year per 100 patients), P = 0.057, and total hospital days were reduced an average of −0.042 PPPM (∼50 hospital days/year per 100 patients), P = 0.139.

OPEN IN VIEWER

FIG. 1.

Diabetes-related medical costs by medical service category. PPPM, per-patient-per-month. Color images are available online.

Discussion

Achieving optimal glucose levels remains a primary goal of diabetes management. ^21,22 However, despite the ongoing introduction of new diabetes medications, a significant proportion of individuals with T2D are not meeting their individual treatment goals. ^3,4 Results from this study demonstrate that use of rtCGM was associated with a significant 25.2% PPPM reduction in total diabetes-related health care resource utilization costs within a large group of T2D patients. The most notable reductions were seen in the cost of diabetes-related inpatient hospital admissions (−34.9%) and outpatient hospital visits (−14.3%) during the 6-month postindex observation period.

Interestingly, our cohort selection process identified 303,130 beneficiaries with a diagnosis of T2D, with 19.0% (n = 57,318) intensively treated with basal/bolus insulin, but only 7.2% (n = 4109) of this population were using some type of CGM device. These numbers suggest that a large percentage of T2D patients on IIT are not using rtCGM, a population that could benefit from this technology.

Although 80% of the cohort were treated with bolus insulin therapy, it is important to consider CGM use in those treated with less-intensive regimens. For example, recent findings from large claims database analyses and chart review studies have shown significant reductions in acute diabetes-related events, all-cause hospitalizations and HbA1c with use of CGM in individuals treated with basal insulin only and noninsulin therapies. ^{23 –27}

This suggests that cost-effectiveness studies of CGM in the broader T2D population are needed to further examine the economic value of this technology, similar to the study conducted by Roze et al. ²⁸ It is important to note that our analysis focused on only health care resource utilization costs and did not include patient medication and medical device costs, which would be factored into a cost-effective analysis of rtCGM over the patient lifespan.

A key strength of the analysis was the ability to accurately identify both our study population and medical costs. The Klompas algorithm has been shown to have a 100% sensitivity and 92% positive predictive value for identifying patients with T2D. ¹⁹ Use of NDC codes allowed us to verify use of rtCGM.

One limitation of our study was the inability to capture more granularized beneficiary data in terms of race/ethnicity, socioeconomic status, and participation in a diabetes education program. Similarly, it is not possible to determine to what extent patients consistently wore the rtCGM device after initiating use, nor do we know if they used their glucose data for their daily self-management decisions. The generalizability of the findings was limited to those with commercial health insurance and thus may not reflect patients who are uninsured or insured through Medicare or Medicaid.

The lack of a comparison group is also a notable limitation, but a comparable control group is difficult to determine given that patients likely initiated rtCGM because poor glycemic management created a self-selection for rtCGM. It is also possible that for some patients in the cohort, a diabetes-related hospitalization during the baseline period may have resulted in a referral to an endocrinologist and a subsequent initiation of rtCGM to help with glucose management. We can speculate that subsequent improved glycemic management from use of rtCGM may have contributed to a subsequent reduction in hospital care for hyperglycemic- and hypoglycemic-related events.

Although randomized controlled clinical trial designs are considered the gold standard for demonstrating the effect of therapeutic interventions such as rtCGM, many payers and regulatory agencies now rely on findings from real-world observational analyses in conjunction with RCT results when they evaluate the safety, effectiveness, and potential benefits of new medications and medical devices. ^{1 –4} Thus, although this study is limited by the one-group pre–post retrospective study design relative to an RCT, it adds to the overall body of evidence for use of rtCGM in relation to diabetes-related medical costs.

Conclusions

Despite a growing body of evidence demonstrating the clinical value of rtCGM in individuals with T2D, ^{12,17,29 –34} a large percentage of the T2D population are not using this technology. This is particularly true in those treated with nonintensive therapies who comprise >75% of the T2D population. ³⁵ Because of the increased price of rtCGM compared with traditional blood glucose testing, most individuals who choose to use this technology must rely on adequate health care coverage.

Unfortunately, the eligibility criteria established by many commercial health plans exclude individuals not treated with IIT. ^36,37 Our findings showed that use of rtCGM in real-world settings was associated with reductions in diabetes-related medical costs for people with T2D (both on IIT and less intensive therapies), and that increased access to rtCGM may help reduce diabetes-related cost of care.