Over the past two decades, research has helped identify ways to reduce complications among people with diabetes and laid the foundation for primary prevention. However, prevention and treatments are still unequally applied, and social, economic, and age-related disparities persist.
The June 2020 supplement issue of Medical Care argues for and presents the results of natural experiments as a way to fill current gaps in knowledge and implementation when randomized trials are not practical or feasible. The issue explores 8 natural experiment studies across different settings. All were conducted under an umbrella project funded by the CDC, National Institute for Diabetes and Digestive and Kidney Diseases (NIDDK), and Patient-Centered Outcomes Research Institute (PCORI).
Natural Experiments
A natural experiment exploits real-world events that occur to change policy, intervention, or health to measure impacts on a broad scale, comparing patients or communities affected by the events with others who are not. Of course, these analyses are still subject to bias from selection into treatment as well as measurement bias. The judicious use of statistical methods can lessen these problems. Natural experiments allow researchers to evaluate how policy and programs work in the world outside of controlled research environments, moving from individual-level analysis in the clinical setting to population-level analysis.
In one paper, Duru and colleagues evaluated a care coordination program. The program integrated medical, behavioral, and social support for high-cost, high-need Medicaid patients. The authors compared patients who enrolled in the program with others who were eligible but never contacted. The results show that the care coordination program was linked to lower rates of emergency department or hospital use for some Medicaid categories.
Another article by Rodriguez et al investigates the impact of the CMS State Innovation Models Initiative on 30-day hospital readmissions. Using a quasi-experimental design, they did not find any effect on 30-day readmission rates among adults with diabetes.
Infographics depict the key takeaways from two additional NEXT-D2 articles. One shown here summarizes findings related to exempting preventive medications from deductibles. Another captures the role of the ACA in reducing ethnic disparities in diabetes treatment and prevention.
Methods
The issue also covers some of the methods used in natural experiment research. More extensive systems for collecting patient data, such as EHRs, hold promise for research given the large numbers of included patients. However, confidentiality and intercompatibility make research in this area more complicated than it might seem at first glance. A paper in the supplement by Huguet and colleagues attacks a different problem: attrition over time. Using data from almost a million patients across 20 states, the authors found average attrition of 34% over 3 years. Interestingly, the rates were less than 25% for patients with diagnoses of diabetes or hypertension.
Stakeholder Engagement
Stakeholders can play an essential role in research and help to set priorities both for research and policy. The NEXT-D2 network incorporates stakeholder engagement at all research sites and at various stages of the research process. An article by Poger and colleagues describes the various stages of stakeholder involvement across different locations:
- Study meetings
- Proposal development
- Training and educational opportunities
- Data analysis
- Results dissemination
The NEXT-D2 network also conveys best practices and suggestions within the network.
Where do we go from here?
Natural experiment studies hold promise to bridge the implementation gap in diabetes research. They allow us to understand which policies can help reduce the burden of diabetes across the nation. Applying rigorous methods to evaluate which programs are making a difference can help guide the decision-making of policymakers and other health stakeholders.
