Abstract 1454: Classification of Heart Failure in the Atherosclerosis Risk in Communities (ARIC) Study: A Comparison With Other Diagnostic Criteria
Introduction: Population-based research on heart failure (HF) is hindered by lack of consensus on diagnostic criteria. In addition, established criteria such as Framingham (FRM), National Health and Nutrition Examination Survey (NHANES), Modified Boston (MBS), or Gothenburg (GTH) are not designed to differentiate acute decompensated HF (ADHF) events from chronic stable HF (csHF). We developed a classification protocol to distinguish ADHF from csHF in the Atherosclerosis Risk in Communities (ARIC) Study and conducted a validation study with a panel of physician reviewers as gold standard to determine the degree to which established criteria accurately identify ADHF.
Methods: A probability sample of 1598 hospitalizations among residents 55 years of age or older in four geographically defined communities was abstracted for data contributing to the above criteria. Reports of echocardiography, chest X-ray, ECG, catheterization, other cardiac imaging and the discharge summary were used by two independent ARIC physician reviewers to classify each event as ADHF or csHF. Discrepancies were adjudicated by a third reviewer.
Results: Thirty-nine percent of men and 33% of women met ARIC criteria for ADHF and 9.6% and 6.5% were classified as csHF, respectively. Among events classified as HF by FRM criteria, 69% were identified as ADHF, 9.1% as csHF and 22% as no HF. Similar ADHF classification rates were seen for the MBS, NHANES, and GTH schema. Agreement among criteria in classifying the same event was poor. The highest percent agreement (and kappa) between ARIC and other criteria was found with FRM 70% (0.32). Overall, the highest agreement (kappa) in event classification was found between FRM and MBS, 88% (0.62). The sensitivity of the 4 criteria to classify ADHF ranged from 80 to 90%, positive predictive values ranged from 63 to 69%, and false positive rates ranged from 60 to 80%.
Conclusions: Although commonly used criteria for classifying hospitalized HF are moderately sensitive in identifying ADHF, their high false-positive rate suggest their specificity is poor when applied to a broad set of random community hospitalizations. Classification criteria such as those used in the ARIC study may be applied in community surveillance of ADHF in hospital discharges.