Post-2020, it’s not a stretch to imagine a prominent role for digital technologies in health care. But Duke University researcher Gary Bennett, PhD, was leading that charge well before “COVID” became a household word. “I’ve long believed, strongly, that digital strategies aren’t simply an option to connect with medically vulnerable populations—they’re the best option. On the back end of the pandemic, that notion is much less controversial than it once was,” he said. “The explosion in the last 18 months in the digital health space has been an amazing sea change. The pandemic has had an unexpected impact on this field, and I don’t think there’s any turning back.”
As director of the Duke Digital Health Science Center, Bennett works with his colleagues to design and test digital health interventions to address the needs of medically vulnerable populations. The multidisciplinary center develops behavior-change interventions to treat obesity and other chronic diseases, with the goal of using digital tools to eliminate health disparities. The center’s core team includes 13 researchers, staff, and graduate students, and their collaborators include psychologists, physicians, nutrition scientists, registered dietitians, public health experts, software engineers, and statisticians.
More than 80% of adults living in low-income households in the United States own a cell phone, and almost half own a smartphone. For Bennett and his team, those devices offer an unparalleled opportunity to help people make choices that improve their health. “The exciting thing about working in the digital sphere is the possibility of reaching so many people. You can break down a lot of barriers to receiving health services,” said data manager and analyst Sandy Askew, MPH. “It’s exciting to be on the cutting edge and trying to find innovative solutions to help as many people as we can.”
The center’s current research focuses on people who have been historically excluded, including people from non-White populations, patients in low-income households, and those who live in rural settings. Often, Bennett’s study participants fall into two or three of those categories.
“When you’re thinking about how to deliver evidence-based treatment to populations at scale, the classic challenge has been how to access medically vulnerable populations in ways that allow us to deliver interventions well,” he said. “We have long been of the opinion that we can do that best through digital.”
Personalized goals
Bennett earned his PhD in clinical health psychology from Duke in 2002. He headed north for a postdoctoral fellowship at the Harvard School of Public Health, then joined the Harvard faculty for several years before returning to Duke in 2009. Officially, he launched the Duke Digital Health Science Center in 2017. But his work on digital interventions for obesity began much earlier—because obesity is such a significant contributor to health problems in at-risk groups, and because he saw a path for behavior change and digital technology to make a big impact. Patients in medically vulnerable groups are more likely to have obesity and to suffer from the chronic conditions that often go along with extra weight. What’s more, existing obesity treatments have been shown to be less effective in those groups.
Technology, he thought, could shrink the gap. While still at Harvard, Bennett developed the interactive obesity treatment approach (iOTA), a model designed to provide participants with personalized behavior-change goals. iOTA uses personalized algorithms to tailor goals for participants in areas such as diet, physical activity, sleep, and stress management. Participants monitor their progress toward the goals, and the system provides customized feedback through an app, text messages, or interactive voice response. “Every time a person logs self-monitoring data about how they’re doing with their goals, they get personalized feedback,” Bennett said.
The iOTA technology forms the backbone of multiple studies. In the randomized controlled “Track” trial, Bennett and colleagues, including Dori Steinberg, PhD, RD, a consulting associate professor and faculty member of the Duke Digital Health Science Center, studied an obesity intervention for adults living in low-income households who had obesity, high blood pressure, high cholesterol, and diabetes. Participants received customized goals (such as avoiding fast food, practicing portion control, and walking more than 7,000 steps per day) and tracked their progress via text prompts.
Counselors reached out to participants periodically to provide additional feedback and troubleshoot barriers to success. Participants also weighed in regularly from home using a cellular-connected scale, and the data were shared with their primary care providers at community health centers. “These digital treatments are built into the fabric of the health system,” Bennett said. “We think the best way to do these trials isn’t in a highly controlled academic care setting but where patients actually receive their care.”
More than 40% of the intervention participants lost at least 5% of their baseline weight—a threshold that has been associated with subsequent improvements in health. Those who completed more than 80% of their self-monitoring lost significantly more than those who were less engaged. In other words, the intervention was effective for those motivated to use it (American Journal of Preventive Medicineopens in new window, Vol. 55, No. 6, 2018).
A similar study explored the iOTA approach not for weight loss but for weight gain prevention. “Half of Americans with obesity are not ready to lose weight at any given moment, and that’s particularly common in medically vulnerable groups, where people have a lot of other things going on,” Bennett said. But preventing additional weight gain can reduce the risk of comorbidities such as diabetes and high blood pressure. “When patients don’t have high motivation or readiness for weight loss, weight gain prevention is an excellent secondary option,” he said.
In the “Shape” trial, Bennett’s team tested their intervention in Black women with overweight or low levels of obesity. It followed the same format as the Track trial, with tailored behavior-change goals and skills training materials, weekly self-monitoring via interactive voice response, monthly counseling calls, and a gym membership. “We showed we could prevent weight gain in that group for almost 2 years,” Bennett said. At 12 months, 62% of participants were at or below their baseline weight, compared with 45% of participants receiving primary care as usual. And those weights were maintained at 18 months (JAMA Internal Medicineopens in new window, Vol. 173, No. 19, 2013).
Following up on that project, Bennett and his team are expanding the weight gain prevention intervention beyond Black women to an entire clinical population of historically and intentionally excluded patients. The 5-year “Balance” protocol, funded by the National Institutes of Health (NIH), tests the digital intervention among patients receiving primary care through a network of federally qualified community health centers in central North Carolina (Berger, M. B., et al., BMC Public Healthopens in new window, Vol. 19, No. 596, 2019).
Precision public health
The ongoing Balance study is designed using a so-called pragmatic format. Unlike traditional explanatory trials, which are done in highly controlled environments, pragmatic trials are designed to test interventions under messy real-life conditions. “In the realm of weight loss research, participants often lose weight under controlled conditions. But what happens when they go back to the real world?” said Miriam Berger, MPH, the center’s research program lead, who manages several digital health trials. Pragmatic trials like Balance also remove a lot of the exclusions—such as strict parameters around age, comorbidities, or body mass index—that are common in obesity intervention trials. “Balance includes adults of all ages, with no upper age limit. It only requires a text-enabled phone, and it’s available in both English and Spanish,” said Berger. Funders such as NIH are increasingly backing pragmatic trials to identify interventions that work as well in someone’s kitchen as they do in an academic lab.
In another project, the team is taking an approach that is both practical and progressive. “We’re thinking about how to take these technologies and deliver them at scale in a way that preserves their evidence base,” Bennett said. To optimize the texting intervention used in many of his obesity trials, Bennett is using a framework known as multiphase optimization strategy (most) to determine which components of an intervention are most effective. “After a clinical trial, we just know that the whole package worked, but we don’t always know which parts of the intervention are responsible for the most positive effects,” he said. “most allows us to test the different elements to develop interventions that only contain active treatment components.”
It’s clear, for instance, that feedback for the user is an important component of the iOTA approach. But is it best to provide feedback each time the participant logs a healthy behavior, or to send a summary once a day, or once a week? Is it more motivating to tell a participant that she’s eating 50% more vegetables than she did last week or to tell her she’s eating more veggies than 50% of the other participants using the intervention? In the ongoing “Charge” trial, Bennett and his colleagues are testing 32 different experimental combinations to determine the impact of each stand-alone component—on their own and in combination with other elements (JMIR Research Protocolsopens in new window, Vol. 10, No. 1, 2021). “Sometimes the words you use, even the shape or color of a button, can have powerful effects on how likely someone is to stay engaged in the intervention,” Bennett said. “We’re finding that some components matter a lot, some matter less, but many of them make a big difference in interaction with one another.”
Optimizing the intervention is a key step toward making digital health tools scalable, Bennett said. And it’s an important step toward making apps that can truly stand on their own. “One of the biggest contradictions in the digital health literature is that most of what we know about digital health science is actually taken from studies that have included human counselors, including our own,” he said.
Bennett and his colleagues aim to create stand-alone obesity treatments that can be delivered exclusively through texts or apps, without the need for health coaches or counselors. “Digital health allows us to deliver interventions to large numbers of people. The big guiding question is how to leverage the technological algorithms to make them flexible enough to meet the needs of individuals,” he said. “Can we make an app that’s flexible enough to deliver hundreds of potential iterations of a treatment to meet the needs of an entire population?”
A flexible behavior-change tool
The team is also branching out beyond obesity, drawing from their expertise in behavior change to create other digital health interventions. The ongoing “Nourish” trial is testing an intervention to promote the Dietary Approaches to Stop Hypertension (dash) eating plan among adults with hypertension. The goal of this intervention is not for participants to lose weight but for them to adopt a dietary pattern that has been shown to lower high blood pressure (Miller, H. N., et al., Contemporary Clinical Trialsopens in new window, Vol. 109, 2021).
Meanwhile, Melissa Kay, PhD, an assistant professor in Duke’s Department of Pediatrics and an affiliate with the Duke Digital Health Science Center and the Duke Center for Childhood Obesity Research, is using similar digital technology to improve maternal diet quality. She’s wrapping up a pilot study that used a text-based intervention to help mothers living in low-income households make better food choices for themselves and their babies, including support with breastfeeding decisions and responsive feeding. “What’s so great about iOTA is that it’s a behavior-change technique. It happened to start with weight, but it can be applied to other behavioral outcomes,” Kay said. “How can we change behavior to improve health? Dr. Bennett has figured out a way to do that using digital technologies.”
As a technology evangelist, Bennett is often an early adopter of new tech, and he is frequently spotted on Twitter discussing digital health and behavior change (@drgarybennettopens in new window). But he takes care to meet his study participants where they are, adapting interventions to be delivered in texts or voice messages when smartphones aren’t an option. “In some ways, we retain old-school technology when it’s the most effective way to reach people,” Askew said. “We’re always finding a balance between reaching people now and thinking about the technology people will be using 2 years down the line.”
Whatever technology they use, Bennett and other members of the center are committed to making a measurable difference in the lives of medically vulnerable populations. And the pandemic has lit a fire under those efforts. “Both comfort with digital health approaches and reimbursement options for digital health have increased dramatically,” Bennett said. In the community health centers he works with in North Carolina, which serve rural and disinvested communities, 70% of clinic visits have been virtual during the pandemic. “This is a real option, and one that I think will challenge us in psychology to think differently about the ways we deliver care.”
And psychology, he adds, has much to offer as the digital health field continues to grow. “There’s a significant channel for early career scholars to find very fulfilling careers in this space,” he said. “The vast majority of our current public health challenges are predominately behavioral in nature, and psychologists have a critical role to play.”
