JMIR Publications

ABSTRACT

Background:

Electronic Health Records (EHRs) have potential to enhance chronic disease surveillance. Population health surveillance for hypertension can be complemented using EHRs to characterize disease burden at the local level.

Objective:

We aimed to derive and validate computable phenotypes to estimate hypertension prevalence for population-based surveillance.

Methods:

This retrospective study developed six candidate computable phenotypes for essential (primary) hypertension in an adult population from a medium-sized Midwestern metropolitan area. Two independent clinician reviewers validated the phenotypes through manual chart review of 150 randomly selected patient records. We assessed precision by calculating positive predictive value (PPV) and validity using prevalence-adjusted-bias-adjusted kappa (PABAK).

Results:

Among a cohort of 548,232 adults, six computable phenotypes produced PPVs ranging from 71.0% (95% confidence interval [CI]: 64.3% - 76.9%) to 95.7% (95% CI: 84.9% - 98.9%). The PABAK revealed a high percentage agreement of 0.88 for hypertension. Similarly, inter-rater agreement for individual phenotype determination demonstrated substantial agreement (Range: 0.70 – 0.88) for all six phenotypes examined. The most sensitive phenotype included diagnosis, blood pressure measurements, and medications, and identified 210,764 individuals (38.4%) with hypertension during the study period (2014-2015).

Conclusions:

We identified several high performing phenotypes to identify hypertension prevalence for local hypertension surveillance using EHR data. Given increasing availability of EHR systems in the United States, leveraging EHR data has potential for enhancing surveillance of chronic disease in health systems and communities.