Does Geography Affect Health? Insights Derived From Integration of Geography-Based Social Determinants of Health With Patient-Level Real-World Data


Discussion Leader: Won Lee, PhD, Decision Science, HEOR/RWE, Axtria, Inc., San Francisco, CA, USA
Discussants: Ron Preblick, PharmD, MPH, Sanofi, Bridgewater, NJ, USA; Keran Moll, PhD, Regeneron Pharmaceuticals, Inc., Tarrytown, NY, USA; Jennifer Ken-Opurum, PhD, Decision Science, HEOR/RWE, Axtria, Inc., Berkeley Heights, NJ, USA


Social determinants are an increasingly important lens for understanding patient health, through which patient- and geography-based disparities in access to care and clinical outcomes are pronounced despite advances in medicine and technology. Patient-level social determinants of health (SDoH) are widely unavailable from many real-world databases. In the absence of individual-level SDoH data, an opportunity in drawing meaningful insights arises by integrating patient-level health data with geography-level SDoH. Furthermore, FDA’s draft guidance for the industry provided recommendations to sponsors on the development of a Diversity Plan to enroll representative participants from underrepresented racial and ethnic populations in clinical trials.

This workshop aims to provide appropriate methods for statistical analysis, and industry examples of using SDoH data to understand disparity and improve diversity, equality, and inclusion (DE&I).


This workshop has wide application across healthcare research stakeholders, addressing whether location, i.e., community resource availability and demographic composition, influences patient health outcomes.

Four analytical methods for generating rich insights will be highlighted across research objectives.

They are (1) running separate models for patient-level data in different geographies, where results are interpreted qualitatively by geography-level SDoH, (2) aggregating patient-level data to the geography-level so the unit of analysis is common, where results are at the geography-level, (3) ascribing geography-level SDoH to each patient living in that geography, such that interpretation of results is, e.g., “patients in communities with higher rates of unemployment have greater odds for heart disease,” and (4) mixed-effects modeling with geography as the random effect, where results are interpretated the same as the third method, but modeling accounts for the dependent nature of health and geography.

Subsequently, similarities and differences between methods will be discussed, along with examples using RWD with SDoH data to improve DE&I in clinical studies. Finally, the workshop will conclude with challenges / opportunities for integrating these methods.




Methodological & Statistical Research