Determining the prevalence of hypertension in a large cohort using self-report, electronic medical records, and records of prescription antihypertensive medications

Background: Hypertension affects approximately 33% of adults in the United States and is associated with substantial clinical and economic burden. When blood pressure measurements are unavailable, accurate identification of hypertensive patients through administrative databases could be valuable.

Objective: We sought to assess agreement among self-reported hypertension, electronically recorded International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for hypertension, and electronically recorded prescription dispensing records for antihypertensive medications in US military personnel within a large, prospective cohort study.

Methods: Self-reported hypertension was identified using baseline Millennium Cohort data collected from July 2001 through June 2003. Specific hypertension ICD-9-CM codes were extracted from outpatient and inpatient electronic medical records. The Pharmacy Data Transaction Service and the Department of Defense Pharmacoeconomics Center databases were queried for antihypertensive prescription medications dispensed. Kappa statistics were calculated to evaluate overall agreement among self-reported hypertension, ICD-9-CM codes for hypertension, and antihypertensive medications dispensed.

Results: There was moderate agreement between self-reported hypertension and at least one electronically recorded ICD-9-CM code (kappa=0.49; 95% CI: 0.47, 0.50). Defining hypertension by two or more outpatient and/or one or more inpatient codes showed similar agreement (kappa=0.50; 95% CI: 0.48, 0.51). Moderate agreement was observed between self-reported hypertension and at least one antihypertensive medication dispensed (kappa=0.48; 95% CI: 0.47, 0.50).

Recommendations: In identifying hypertensive patients in the absence of blood pressure measurement, ICD-9-CM codes and dispensing records can be an important complement to self-report. Understanding the strengths and limitations of administrative databases are critical when classifying patients in large population studies.