Speaker
Christopher J. Lindsell, PhD
Professor and co-Chief, Biostatistics, Duke University
Director, Data Science and Biostatistics, DCRI
Director, Biostatistics and Bioinformatics, CTSI
Editor in Chief, Journal of Clinical and Translational Science
Keywords
Study Design; Diversity; Health Disparities; Evidence Generation
Key Points
- Health disparities are factors that contribute to preventable differences in health status and outcomes. They can be environmental, sociocultural, behavioral, and biological. These are preventable differences with adverse effects for populations.
- When research teams don’t consider factors that change outcomes for certain populations but not others, research can contribute to a difference in health status and outcomes. A flawed evidence generation system compounds the problem.
- One popular solution is to measure and adjust for diversity variables; however, research teams often get the variable wrong or use it incorrectly.
- By designing for diversity, research teams can begin to address the generalizability of evidence; develop an understanding of factors that contribute to success or failure of interventions among diverse populations; and remove the evidence generation system as a contributor to health disparities.
- Designing for diversity is an optimization problem. Historically, study designs have been optimized for the researcher; Dr. Lindsell proposed that researchers optimize for the participant.
- Research that is optimized for the participant is rigorous and flexible; safe and practical; and complete and simple. Participants should be embedded in every part of the research process. This can be difficult – there are tradeoffs involved – but it is effective.
- The interface between data generation and data use is crucial. Making systems that work to bring in the right information and systems that use that information appropriately are two pieces of the same puzzle.
- Dr. Lindsell included a call to ditch the ordinary subgroup analysis, noting that groups are not binary and not all groups have meaning. He suggested a focus on interaction terms.
Discussion Themes
Institutions should be supporting research teams in their ability to achieve diversity; this should be a matter of course, rather than something to be achieved without support and then celebrated as a remarkable accomplishment.
Looking forward, as data infrastructure becomes increasingly robust, research teams and communities may be able to collaborate to build a more complete understanding of individuals’ health states and who may be in need of an intervention.