1. Daley, Matthew F. MD
  2. Barrow, Jennifer C. MSPH
  3. Tabano, David C. MA
  4. Reifler, Liza M. MPH
  5. Kraus, Emily M. PhD, MPH
  6. Davies, Sara Deakyne MPH
  7. Williford, Devon L. MPH
  8. White, Benjamin MPH
  9. Shupe, Alyson PhD
  10. Davidson, Arthur J. MD, MSPH


Context: Although local childhood obesity prevalence estimates would be valuable for planning and evaluating obesity prevention efforts in communities, these data are often unavailable.


Objective: The primary objective was to create a multi-institutional system for sharing electronic health record (EHR) data to produce childhood obesity prevalence estimates at the census tract level. A secondary objective was to adjust obesity prevalence estimates to population demographic characteristics.


Design/Setting/Participants: The study was set in Denver County, Colorado. Six regional health care organizations shared EHR-derived data from 2014 to 2016 with the state health department for children and adolescents 2 to 17 years of age. The most recent height and weight measured during routine care were used to calculate body mass index (BMI); obesity was defined as BMI of 95th percentile or more for age and sex. Census tract location was determined using residence address. Race/ethnicity was imputed when missing, and obesity prevalence estimates were adjusted by sex, age group, and race/ethnicity.


Main Outcome Measure(s): Adjusted obesity prevalence estimates, overall, by demographic characteristics and by census tract.


Results: BMI measurements were available for 89 264 children and adolescents in Denver County, representing 73.9% of the population estimate from census data. Race/ethnicity was missing for 4.6%. The county-level adjusted childhood obesity prevalence estimate was 13.9% (95% confidence interval, 13.6-14.1). Adjusted obesity prevalence was higher among males, those 12 to 17 years of age, and those of Hispanic race/ethnicity. Adjusted obesity prevalence varied by census tract (range, 0.4%-24.7%). Twelve census tracts had an adjusted obesity prevalence of 20% or more, with several contiguous census tracts with higher childhood obesity occurring in western areas of the city.


Conclusions: It was feasible to use a system of multi-institutional sharing of EHR data to produce local childhood obesity prevalence estimates. Such a system may provide useful information for communities when implementing obesity prevention programs.