1044. Identifying Hypertension-Related Comorbidity using Administrative Data
Ann M Borzecki, MD, MPH, CHQOER, Bedford VAMC and Boston University School of Public Health, AT Wong, CHQOER, Bedford VAMC and Boston University School of Public Health, EC Hickey,
CHQOER, Bedford VAMC, AS Ash,
Boston Medical Center and Boston University School of Medicine, DR Berlowitz,
CHQOER, Bedford VAMC and Boston University School of Public Health
Objectives: To determine the best strategy for identifying outpatients with hypertension-related diagnoses using the Outpatient Clinic File, an administrative VA database.
Methods: We retrospectively reviewed the medical charts of 1176 regular ambulatory care users at ten VA sites in 1999, taking the presence/absence of a diagnosis for each of eleven diagnoses relevant to hypertension management as the gold standard for identifying the comorbidity. We calculated observed and kappa measures of agreement, and sensitivity and specificity, for the chart versus several administrative data-based algorithms, e.g. requiring one, or two OPC records with the diagnosis, or using one or two years of data.
Results: Using the same year for both sources and requiring one administrative diagnosis, observed agreement ranged from 0.98 for atrial fibrillation to 0.85 for hyperlipidemia. Kappas were high, indicating at worst moderate agreement beyond chance for tobacco use (0.47) to almost perfect agreement for diabetes (0.92). Sensitivity varied from 38% for tobacco use to 97% for diabetes. Specificity exceeded 91% for 10/11 diagnoses. Using two years of administrative data and requiring two diagnoses, increased specificity, observed agreement and kappas, with minimal sensitivity decrease.
Conclusions: We found good agreement between the outpatient encounter file and chart diagnoses. The administrative data varied in its ability to identify all patients with a given diagnosis but could accurately identify those without these diagnoses. The best strategy for case-finding for quality assessment required two diagnoses in a two-year period.
Impact: This information may be used to profile provider treatment patterns in patients with hypertension and monitor guideline adherence.