Internal medicine

Pathology

Sociology

Behavioral Risk Factor Surveillance System

Population

Geocoding

Obesity

Demography

Concordance

American Community Survey

Geography

Overweight

Cartography

Gerontology

Census

Medicine

Environmental health

Objective Health system data were assessed for how well they can estimate obesity prevalence in census tracts. Methods Clinical visit data were available from two large health systems (Kaiser Permanente Washington and University of Washington Medicine) in King County, Washington, as were census tract–level obesity prevalence estimates from the Behavioral Risk Factor Surveillance System (BRFSS). The health system data were geocoded to identify each patient’s tract of residence, and the cross‐sectional concordance between census tract–level obesity prevalence estimates computed from the two health systems in 2005 to 2006 and the concordance between University of Washington Medicine and BRFSS from 2012 to 2016 were assessed. Results The spatial distribution of obesity was similar between the health systems (Spearman r = 0.63). The University of Washington Medicine estimates of rank order correlated well with BRFSS estimates (Spearman r = 0.85), though prevalence estimates from BRFSS were lower (mean obesity prevalence = 26% for University of Washington Medicine versus 20% for BRFSS, Wilcoxon rank sum test p &lt; 0.001). Across all data sources, obesity was more prevalent in tracts with less educational attainment. Conclusions Health system clinical weight data can reliably replicate census tract–level spatial patterns in the ranking of obesity prevalence. Health system data may be an efficient resource for geographic obesity surveillance.

From the clinic to the community: Can health system data accurately estimate population obesity prevalence?