These are the tools for providing top-notch diabetes care to everyone

For 30 years, doctors have known that maintaining near-normal blood sugar has huge benefits for people with Type 1 diabetes.

A 1993 clinical trial found that participants who were taught methods for tightly managing their disease -- checking their blood sugar many times each day, making adjustments to insulin doses and receiving frequent help from their medical caregivers -- reduced their risk for long-term complications, including blindness, kidney failure and peripheral nerve damage, by 50% to 70%.

Yet since that trial, physicians have struggled to roll out intensive diabetes management programs to all patients. On average, patients across the U.S. still don't achieve the level of diabetes control that would minimize their long-term risks.

"It's not because people haven't been trying," said David Maahs, MD, PhD, a pediatric endocrinologist at Stanford Medicine Children's Health. "It's a complicated condition to take care of, and for the individual with diabetes and their family, it's constant work."

Stanford Medicine experts are making headway on the problem. A new scientific study published recently in Nature Medicine describes how the research team, led by Maahs and Priya Prahalad, MD, PhD, have implemented major advances in intensive diabetes management.

First, they tackled equity issues to ensure the latest diabetes technology got into the hands of every patient as soon as they were diagnosed. They also built artificial intelligence tools that gave diabetes caregivers the ability to quickly identify which patients most needed their help. Ultimately these steps enabled adolescent Type 1 diabetes patients to maintain better control of their blood sugar levels.

It's not because people haven't been trying. It's a complicated condition to take care of, and for the individual with diabetes and their family, it's constant work.
David Maahs

Maahs talked about the methods used and the long-term ramifications. This interview was edited for length and clarity.

Your study built on recent technological advances that automate many tasks involved in living with diabetes. What are the advantages of the newer devices?

Patients can now wear continuous glucose monitors, which have a sensor inserted under the skin that reads a glucose value every 5 to 15 minutes. This is really helpful because you don't have to poke your finger six to 10 times a day to measure glucose levels, and the monitor can warn you if you're going low or high. If you're the parent of a child with Type 1 diabetes, you can get their glucose data from the cloud and onto your phone.

Another recent improvement in diabetes technology is that continuous glucose monitors can now communicate with an insulin pump. An algorithm helps control dosing to reduce or stop insulin if it predicts your blood sugar is going to go low, and it adds a bit more insulin if you're going high. You still have to give an insulin dose before you eat, but it really takes a lot of the burden out of managing Type 1 diabetes.

It's been a big challenge to get these improved diabetes devices into the hands of every U.S. patient; your earlier work shows that disadvantaged groups tend to be left behind. How did your new study tackle equity concerns?

We were testing the benefits of starting pediatric Type 1 diabetes patients on continuous glucose monitors as soon as possible after they were diagnosed, which we were usually able to do in the first week after diagnosis. Although insulin pumps were not a focus of this study, about half of our patients began using an insulin pump within a year of their diagnosis.

We all agreed that we wanted every new patient we saw to be included. If you look at earlier studies of diabetes technology, it tended to be tested in college-educated, white, privately insured people and not in other populations. We had to figure out how to meet challenges faced by less-advantaged patients. We learned that this went beyond the most obvious barriers we addressed, such as providing care in multiple languages.

For instance, at first, it seemed like some groups of patients were wearing their continuous glucose monitors less than we asked them to. But in fact, the transmission of their data to our system was incomplete because they had poor Wi-Fi access at home. That's an equity issue. We've been giving out devices to those who need them as part of the research so that everyone has enough internet connectivity to upload their data.

Also, at diagnosis, we sometimes can't tell whether someone has Type 1 or Type 2 diabetes. This happens more in minoritized populations, in youth who have an elevated body mass index, and these children are more likely to be publicly insured or non-English speakers. We made a conscious decision to include all these patients while we waited to learn the details of their diagnosis, so as not to miss anyone who might be eligible for our study.

We made a conscious decision to include (a diverse mix of) patients while we waited to learn the details of their diagnosis, so as not to miss anyone who might be eligible for our study.
David Maahs

Close to 90% of our new Type 1 diabetes patients participated in this study, and a lot of those who chose not to participate enrolled instead in a different study of artificial pancreas technology, so we did quite a good job of including everyone. It was a very diverse population: About 35% of patients were publicly insured, and only about 40% were non-Hispanic white.

Getting the new technology to every patient was important. What else was needed to make sure all patients could succeed in managing their blood sugar levels?

The 1993 trial showed that it was really useful for patients to have frequent communication with their diabetes team. That can be hard to do with the resources of a typical diabetes clinic, where each diabetes educator has many patients to track.

To address this problem, we built an AI-powered data dashboard, which filters our patients' continuous glucose monitor data and puts it into a format that helps our team identify who is struggling. Instead of spending a lot of time manually evaluating their data, we can automatically rank which patients most need our help.

We look at the percentage of time the continuous glucose monitor has been worn, and if it's below a certain threshold, that's our first starting point. Sometimes people have lost their prescription for their CGM, continuous glucose monitor, and need a new one. We're able to reach out and help them.

If a patient is having too many low blood sugar readings, which are dangerous, that's another reason for them to go to the top of the list. Our diabetes educators can contact them to help adjust their insulin dosing. Likewise, if their average glucose is out of the target range less than 70% of the time, we can flag that they need some extra attention between their visits to the clinic.

On the other hand, if someone is wearing their monitor, they're not having low blood sugar readings and they are in the right blood sugar range most of the time, they're doing well. We'll check in with them at their quarterly clinic visit, but they don't need outreach in between. That knowledge helps our team shift its attention to the patients who most need it.

How did you build the algorithm that powers this dashboard?

The platform was developed at Stanford with systems design expert David Scheinker, PhD, and his SURF team; they use tools such as machine learning and statistics to improve how health care is delivered. He teaches a class in which engineering undergraduates and grad students solve technical problems in health care.

We presented our concept to his class. Our problem was that each brand of diabetes equipment has a different system to share data. A diabetes educator with 10 patients might have had five or six different places to go look at their data, log in and so on. This made it extremely time-consuming to figure out which patients needed help.

More on diabetes

Instead, our system has all the data in one place. It was built through an iterative process between our diabetes educators and the engineering team, so that ultimately the data is presented in the way that is most useful to the diabetes educators.

We published a study showing that the dashboard has shifted the diabetes educators' workload in a helpful way. They spend less time sifting through troves of data - something an algorithm can do perfectly - and more time talking to patients who need extra help between clinic visits.

How do you know that your approach worked?

Compared with our past patients who were diagnosed with Type 1 diabetes before this research began, our newer patients were more likely to reach their glucose targets after a year of living with the disease.

One treatment target for our patients, after a year with diabetes, was a glycosylated hemoglobin A1c measurement below 7%. This laboratory test assesses patients' blood sugar control over the prior three months, and a reading below 7% is the target for optimum health.

In our earlier data, 28% of patients met this target 12 months after diagnosis; now we have 64% of our patients meeting this goal. We also looked at how many people had very high A1c measurements, with values above 9%, and that measure has reduced dramatically. Similarly, by one year after diagnosis, our patients' blood sugar was in the target range 68% of the time.

We also have data showing that compared with our historic cohort, everyone received a similar benefit from our intensive approach to treatment, meaning that if you looked at people who had public versus private insurance, or were or were not English speakers, every group had similar improvements when we implemented our study. There are still some gaps between more- and less-privileged patients, so we still have work to do, but everyone benefited a similar amount. Often, when new medical technology becomes available, more privileged people get more benefit; it is very encouraging that we could buck that trend.

Image: habrovich (stock.adobe.com)