Coding of heart failure diagnoses in Saskatchewan: a validation study of hospital discharge abstracts.

BACKGROUND Validity of Heart Failure (HF) diagnoses from administrative records has not been extensively evaluated, especially with respect to small / unselected hospitals. OBJECTIVES To determine the positive predictive value of a primary / most responsible diagnosis of HF among a general population of subjects discharged from Saskatchewan hospitals. METHODS Using administrative health records from the Province of Saskatchewan, Canada, we identified subjects experiencing their first HF hospitalization between 1994 and 2003. From this cohort, we randomly selected 500 subjects for individual validation using Framingham and Carlson criteria. RESULTS The 466 charts available for analysis, 74% (345/466) and 63.9% (298/466) of subjects met criteria for a clinical diagnosis of HF based on Framingham or Carlson criteria, respectively; 57.5% (268/466) met both criterion. Provincial hospitals (located in the largest urban centres) were associated with the highest proportion of confirmed HF diagnoses (87.8% by Framingham criteria) compared to progressively smaller hospitals (regional 77.9%; district 64.2%; and community 60.0%). Accuracy also differed when stratified by physician category. Cardiologists and internists were associated with the highest rates of confirmed diagnoses [(97.5% (39 / 40) and 85.0% (34 / 40)]) compared to general practitioners [(73.1% (95 / 130)]) and other physicians [(69.1% (177 / 256)]), by Framingham criteria. CONCLUSIONS Hospital discharge abstracts indicating HF are frequently inaccurate. These findings have important implications for the epidemiologic study of HF as well as the clinical management of patients.