Continuous glucose monitors (CGMs), which provide real-time monitoring of glucose levels, are a significant advancement in diabetes management that can lead to improved health outcomes. Although CGMs are considered the standard of care for people with insulin-treated diabetes, disparities in access to these devices persist, particularly among Medicaid populations.

Oklahoma’s Medicaid agency is using data analysis to address disparities in CGM access, including expanding eligibility to new populations who might benefit from CGMs. The Center for Health Care Strategies (CHCS) spoke with Justin Davis, PhD, clinical outcomes analyst, and Toney Welborn, MD, MPH, MS, former medical director, at the Oklahoma Health Care Authority about how the agency is identifying priority groups for CGM access and the insights gained from using data to inform Medicaid policy improvements in CGM access. Oklahoma is a participant in the Continuous Glucose Monitor Access Accelerator, a national initiative led by CHCS with support from The Leona M. and Harry B. Helmsley Charitable Trust that is helping states expand access to CGMs within Medicaid.

Q. What drove your team to analyze CGM use patterns in the state’s Medicaid population?

T. Welborn: Our Medicaid policy currently requires members to be on insulin to qualify for a CGM, but we wanted to explore expanding access to anyone with diabetes who could benefit. There were initial concerns about cost, so we needed clear evidence of a return on investment (ROI). While literature shows that better diabetes control lowers health care costs, there wasn’t much specifically on ROI for CGMs. We decided to generate that analysis ourselves.

J. Davis: To define ROI, we started by examining claims data for people already using CGMs to identify relevant outcomes. Along with spending, we looked at inpatient visits, insulin use, and other indicators tied to cost and care quality. Rather than focusing narrowly on dollars spent, we took a broader approach that considered multiple outcomes simultaneously. That complexity helped us tell a more complete story.

Q. What data sources have been most valuable in your analysis?

J. Davis: We primarily rely on Medicaid claims data, which shows who is receiving CGM prescriptions, other provider-submitted claims, demographic details, and some diagnosis codes. For example, we can see that someone had an A1c test — but not the test result.

That points to a key limitation: we don’t have visibility into how consistently someone uses their CGM. Claims data show when a device is picked up, but not how often it’s worn. To address that, we created proxy measures, including the regularity of sensor refills as an indicator of continued use. We also triangulated with emergency department visits and hospitalizations to look for correlations between presumed CGM use and adverse events.

Q. How did you define your study population and comparison groups?

J. Davis: We included members with a sustained pattern of diabetes diagnoses over time — both CGM users and non-users. To avoid including people who may have received only a one-time diagnosis, we set thresholds around the number and timing of diabetes-related claims to ensure we were capturing individuals actively managing the condition. This gave us a robust comparison group of members with diabetes who haven’t received a CGM, including many with consistent insulin use, which allows us to meaningfully compare outcomes across groups.

Q. What key outcome measures were used to evaluate your research question?

J. Davis: We focused on four main metrics: inpatient visits, insulin use, total claims costs per quarter, and total claims costs excluding insulin. Inpatient visits are a clear marker of diabetes-related health crises. Insulin use helped us understand adherence to treatment and engagement with care.

Tracking total spending per quarter gave us a baseline for assessing ROI. And separating out insulin costs helped us interpret those spending trends more accurately. Because many members start insulin use after receiving a CGM, higher costs can actually reflect more appropriate treatment, which we know will be beneficial in the long term. Removing insulin expenses gave us a more helpful picture of both cost and impact over time.

We used regression models to assess the impact of CGM use, controlling for factors such as age, sex, race, geography, and pre/post-CGM periods. While the underlying data work requires significant infrastructure within the agency, the analysis itself is straightforward to replicate using standard tools. 

Q. What are some early findings?

J. Davis: Our results consistently show a “typical” curve with CGM use: in the months before CGM use, members experience worsening health — more inpatient visits, higher spending, and more diagnoses, often with complications like retinopathy, kidney failure, or even amputations (see below figure). After the CGM is introduced, we see stabilization and then improvement. These patterns hold even after controlling for factors like age, sex, race, and geography. In most cases, outcomes return to the level of someone who never needed a CGM — sometimes even better.

One exception is insulin use. That continues to increase post-CGM and becomes more consistent, which is encouraging. It suggests that members are staying engaged in their care and using insulin more consistently over time.

T. Welborn: From a clinical perspective, the trends seen in the analysis are really powerful as they indicate better diabetes control. The CGM doesn’t directly lower someone’s blood sugar — it’s the members and their providers using that information to make changes. That’s the true value of CGMs: it supports behavioral change and better self-management.

Q. What role did descriptive analysis play in interpreting your results and informing decision making?

J. Davis: Descriptive analysis — that initial step of examining data patterns before running complex statistical models — turned out to be critical. It was tempting to jump straight into regression models, but we made sure to first review claims and diagnosis codes for hundreds of CGM users. That manual review helped us see the real-world context behind the data — and it changed how we interpreted our results.

If we only looked at the regression outputs, we might have wrongly concluded that CGM use was associated with worsening outcomes. But the descriptive analysis showed that CGMs are typically introduced in response to a health crisis, not before it. That insight helped us avoid misinterpreting cause and effect — and it’s a reminder that spending time with the data, patient by patient, really matters.

T. Welborn: These findings reinforced policy decisions we’d made prior to the CGM Access Accelerator project — like streamlining prior authorization so members using insulin in the past 90 days could access CGMs more easily. That change was meant to reduce provider burden and enable earlier use of CGMs before a crisis. But the data suggested some providers may not be aware of the change, highlighting a continued need for education and outreach. We’re also exploring how to expand CGM access to high-risk groups — like those with heart failure or vascular disease — who could benefit from earlier intervention, even if they’re not yet on insulin.

Q. Beyond informing policy change, can the CGM data analysis support other efforts to improve access to CGMs in Oklahoma?

J. Davis: To support members in accessing our chronic disease management program, the data analytics team develops monthly reports that flag members who are newly diagnosed with diabetes, newly prescribed insulin, or receiving a CGM for the first time. We pass that list to the care management team. They’ve had great success helping members initiate CGM use.

T. Welborn: We’ve also shared data trends with a Project ECHO program in our state. That’s been a great way to build awareness among providers, particularly around how CGMs can reduce hospitalizations. As part of our participation in CHCS’ CGM Access Accelerator project, we’re planning to use our findings to support provider training on the value of CGMs and how to confidently prescribe them. Our goal is to make CGMs feel more approachable for clinicians who may not yet be familiar — and over time, we hope to evaluate the impact of these training efforts on member outcomes.

We’ve also shared our findings with managed care plan quality teams. One plan responded by expanding CGM coverage to all members 18 and older with diabetes, beyond what’s included in our state plan. That kind of partnership is really promising.

Q. Looking ahead, how do you see the CGM data analysis informing future policy decisions?

T. Welborn: Right now, Medicaid coverage of CGMs is available only if you’re on insulin. But I’m concerned about people who do better with a CGM and no longer need insulin; I feel strongly they should be able to keep their CGM. We need to show leadership that the intervention worked and should continue.

We’re also using findings to build the case for strategic expansion beyond insulin-dependent members. We’re identifying high-risk subgroups where we can demonstrate both clinical benefits and cost-effectiveness. That includes people with gestational diabetes, congestive heart failure, or peripheral vascular disease. Our goal is to show where CGM access could have the greatest impact.

Q. What advice would you give other states looking to use data to support policymaking around CGMs and/or diabetes more generally?

J. Davis: Start with a deep dive into your data. Before building models or drawing conclusions, review individual claims and member histories. That descriptive work helped us understand how CGMs were actually being used — and shaped how we approached everything that followed. You’ll learn things you might miss in aggregate analysis, and it grounds your work in real member experiences.

T. Welborn: Get your providers on board. They’re key to any successful CGM policy. It’s powerful to be able to tell a story about what happens when someone receives an intervention and turns their health around. Providers want to do what’s best for their patients, and when they see the impact CGMs can have, they’re more likely to act.