A panel of experts shared novel approaches to designing studies in diabetes health care during the Ask the Expert session, Old Dog, New Tricks? Making Sense of the New Research in Care Delivery, on Monday, June 23, the final day of the 85^th Scientific Sessions.

Though panelists highlighted different tools, they all stressed the importance of a collaborative approach in which patients, support networks, care teams, and even institutions participate from the onset.

Pragmatic Trials

Deborah J. Wexler, MD, MSc, introduced pragmatic clinical trials, a variation of pragmatic trials designed to study outcomes in real-world settings that are often conducive to diabetes studies.

“In type 2 diabetes, we are very rarely telling our patients to choose between an active treatment and a placebo. We are choosing between treatment options,” explained Dr. Wexler, Associate Professor of Medicine at Harvard Medical School and Chief of the Massachusetts General Hospital Diabetes Unit.

She is applying pragmatic trial approaches in the ongoing PRECIDENTD trial that compares the effectiveness of sodium-glucose cotransporter-2 (SGLT2) inhibitors versus glucagon-like peptide-1 receptor agonists (GLP-1 RAs) in the treatment of type 2 diabetes.

Implementation Science

Rachel G. Tabak, PhD, RD, examined ways public health and health care services can improve service through implementation science, which she defined as strategies to bridge gaps between research, practice, and policy.

Dr. Tabak, Associate Professor at Washington University, noted some essential components for this: beginning the process with the end goal in mind, ensuring the actions solve a problem and apply to the context, and engaging with partners throughout the process.

The definition of partners should include “not just the patients affected, but the settings, the payors, and everybody involved with helping intervention be successful,” Dr. Tabak said.

Community-Based Participatory Research

Jennifer Raymond, MD, MCR, presented case studies of two community groups as examples of patient- and family-driven collaborative care in diabetes. The groups served different populations with type 1 diabetes, but both tapped members to define and shape their programs.

Dr. Raymond, Associate Professor at the University of Southern California, and Division Chief in Pediatric Endocrinology at Children’s Hospital Los Angeles, said the communities added aspects such as spiritual components and events intentionally not focused on diabetes.

“What we learned is that what they say they needed was very different than what was represented in the research,” Dr. Raymond said.

Natural Experiments

Natural experiments are approaches often looking at something that has happened in the past and where the placement of an individual into a test group is determined, not by randomization nor by the individual, but rather through external phenomenon, such as place of residence, explained Ronald T. Ackerman, MD, MPH, the James Roscoe Miller Professor of Medicine, Chicago Center for Diabetes Translation Research, Northwestern University Feinberg School of Medicine.

As an example of this, Dr. Ackerman pointed to a case study of whether the National Diabetes Prevention Program lowers the cumulative incidence of type 2 diabetes. He noted that because this approach often relies on secondary data, researchers often compensate data holders for access.

However, these institutions can be more receptive to sharing information “if you try to understand the decisions they are struggling with and try to design the experiment so that it answers the questions relevant to those stakeholders,” Dr. Ackerman said.

AI and Big Data

Michelle Shardell, PhD, encouraged diabetes health researchers to tap artificial intelligence (AI) and machine learning to do much of the “behind the scenes” work in statistical research analysis.

Drawing on case studies, such as the effects of glucose-lowering medications on major adverse cardiovascular events, she explored how epidemiology, biostatistics, AI, and machine learning can be combined to create more impactful research. Other examples of AI use include constructing weights for use in the Cox-proportional hazards model and addressing extreme or outlier results.

Dr. Shardell, Professor at the University of Maryland School of Medicine, urged colleagues not to be afraid to reach across the gaps of epidemiology and biostatistics on one side and AI and machine learning on the other.

“The good news is you know more than you think,” Dr. Shardell said. “There is increasing crosstalk between these two sets of disciplines, and what we have been learning from each other is that we are often concerned about the same concepts, but with slightly different terminology.”

On-demand access to recorded presentations from this session will be available to registered participants of the 85^th Scientific Sessions through August 25.