OBJECTIVES Administrative data are often used to identify patients for inclusion in quality performance measures. We examined the specificity of ICD-9 codes from administrative data for identifying patients hospitalized for heart failure (HF). METHODS Charts of 90 adults (≥25 years) admitted to two university hospitals between November 2005 and October 2006 were selected on the basis of ICD-9 discharge diagnosis [primary diagnosis of HF (n=45), or without HF (n=45)]. The ICD-9 codes used to identify HF were those used in the Joint Commission National Hospital Quality Measures. Medical charts were abstracted to identify the physician discharge diagnoses. Specificity was calculated by comparing the number of patients identified with HF via ICD-9 codes from administrative data to the number identified by medical record review [(true negatives/(true negatives + false positives)]. The main analysis was based on the reference standard being the physician primary discharge diagnosis of HF. In sensitivity analyses we expanded the definition of the reference standard to include; A) physician secondary discharge diagnosis of HF with shortness of breath (SOB) or respiratory failure (RF) as the primary diagnosis; and B) physician secondary discharge diagnosis of HF with SOB, RF, arrhythmia, or valvular disease as the primary diagnosis. RESULTS The specificity of the administrative data was 75.0% (95% CI 62.1-85.3%) for the main analysis. Using the expanded definitions of the reference standard A and B above, the specificity was 77.6 % (95% CI 64.7-87.5 %) and 81.8% (95% CI 69.1-90.9%), respectively. CONCLUSIONS Administrative data have relatively high specificity (>75%) for identifying hospitalized patients with heart failure. Additional analyses are needed to determine the extent to which quality among those captured by administrative data reflects overall quality, and to determine the sensitivity of administrative data for identifying HF patients.