The Spatial Distribution of Adult Obesity Prevalence in Denver County, Colorado: An Empirical Bayes Approach to Adjust EHR-Derived Small Area Estimates. Issue 1 (6th December 2017)

• Record Type: Journal Article
• Authors: Tabano, David C.; Bol, Kirk; Newcomer, Sophia R.; Barrow, Jennifer C.; Daley, Matthew F
• Is Part Of: EGEMS. Volume 5:Issue 1(2017)
• Publication Date: 2017-12-06
• Subjects: Medical records -- Data processing -- Periodicals; Medical care -- Data processing -- Periodicals; Medical Records; Automatic Data Processing; Medical care -- Data processing; Medical records -- Data processing; Periodicals; Periodicals; 651.504261
• DOI: 10.5334/egems.245
• Languages: English
• ISSNs: 2327-9214

Abstract

Objectives: Measuring obesity prevalence across geographic areas should account for environmental and socioeconomic factors that contribute to spatial autocorrelation, the dependency of values in estimates across neighboring areas, to mitigate the bias in measures and risk of type I errors in hypothesis testing. Dependency among observations across geographic areas violates statistical independence assumptions and may result in biased estimates. Empirical Bayes (EB) estimators reduce the variability of estimates with spatial autocorrelation, which limits the overall mean square-error and controls for sample bias. Methods: Using the Colorado Body Mass Index (BMI) Monitoring System, we modeled the spatial autocorrelation of adult (≥ 18 years old) obesity (BMI ≥ 30 kg m 2 ) measurements using patient-level electronic health record data from encounters between January 1, 2009, and December 31, 2011. Obesity prevalence was estimated among census tracts with >=10 observations in Denver County census tracts during the study period. We calculated the Moran's I statistic to test for spatial autocorrelation across census tracts, and mapped crude and EB obesity prevalence across geographic areas. Results: In Denver County, there were 143 census tracts with 10 or more observations, representing a total of 97, 710 adults with a valid BMI. The crude obesity prevalence for adults in Denver County was 29.8 percent (95% CI 28.4–31.1%) and ranged from 12.8 to 45.2 percent across individual census tracts. EB obesity prevalence was 30.2 percent (95% CI 28.9–31.5%) and ranged from 15.3 to 44.3 percent across census tracts. Statistical tests using the Moran's I statistic suggest adult obesity prevalence in Denver County was distributed in a non-random pattern. Clusters of EB obesity estimates were highly significant (alpha=0.05) in neighboring census tracts. Concentrations of obesity estimates were primarily in the west and north in Denver County. Conclusions: Statistical tests reveal adult obesity prevalence exhibit spatial autocorrelation in Denver County at the census tract level. EB estimates for obesity prevalence can be used to control for spatial autocorrelation between neighboring census tracts and may produce less biased estimates of obesity prevalence.