Analysis of Asthma Prevalence Using Clinical Electronic Health Records and Air Pollution Data

BACKGROUND: Asthma is a chronic disease affecting over 500,000 children and adults in Wisconsin. Air pollution components including PM_2.5 and ozone are known risk factors for asthma exacerbations. Electronic health records (EHR) are a source of practical data for clinical and public health research. The goal of this research was to examine asthma prevalence and risk factors associated with asthma in the EHR and air pollution data.

METHODS: The UW eHealth-PHINEX collaboration between the University of Wisconsin Departments of Family Medicine, Pediatrics and Internal Medicine clinics, Applied Population Laboratory (APL) and Wisconsin Division of Public Health Information Network (PHIN) have allowed us to identify a patient population with asthma at a census block group level using electronic health records (EHR). The database contains extracted clinical care fields, geocoding to the census block group neighborhood level, and detailed socio-demographic data. In addition, we have included EPA criteria air pollutant data modeled to the census block group. Average annual (2008) values for criteria pollutants CO, NO₂, PM_2.5, PM₁₀, and SO₂ were modeled from stack emissions and aggregated to the census block group. Frequency tables and multivariate logistic regression models were developed using EHR and modeled pollution data. GIS and spatial analyses were used to map areas of high air pollution and asthma disparity.

RESULTS: Between 2007 and 2011, approximately 462,000 patients, including 45,000 asthmatics, seen in the University of Wisconsin health system were identified using EHR. Asthma prevalence was 9.7 percent in the clinic population, ranging from 6.4 percent (among adults aged 65+ years) to 13.5 percent (among children aged 5-11 years). By race-ethnicity, asthma prevalence was highest among Black non-Hispanics (17.9 percent). Covariates significantly associated with asthma prevalence (p<0.0001) in multivariate analysis included younger age (childhood), female gender, Black non-Hispanic race-ethnicity, former smoking status, elevated body mass index, and reporting Medicaid or Medicare insurance. The only air pollutant that was significantly associated with asthma prevalence in the adjusted model was PM_2.5 (p=0.04). GIS analyses identified asthma patients and areas of high air pollution at the neighborhood level in the Madison, WI area.

CONCLUSIONS: EHRs provide exciting opportunities in asthma surveillance, including small area estimation of asthma prevalence by linking to databases on community-level demographic and socioeconomic factors, behaviors, geographic, and environmental characteristics. This could further highlight areas of asthma disparity, allow discovery of novel risk factors, and improve targeting of education and healthcare interventions.