Probability of Accurate Heart Failure Diagnosis and the Implications for Hospital Readmissions

Abstract Heart failure (HF) is a complex syndrome with inherent diagnostic challenges. We studied the scope of possibly inaccurately documented HF in a large healthcare system among patients assigned a primary diagnosis of HF at discharge. Through a retrospective record review and a classification schema developed from published guidelines, we assessed the probability of the documented HF diagnosis being accurate, and determined factors associated with HF-related and non-HF-related hospital readmissions. An arbitration committee of 3 experts reviewed a subset of records to corroborate the results. We assigned a low probability of accurate diagnosis to 133 (19%) of the 712 patients. A subset of patients were also reviewed by an expert panel, which concluded that 13%-35% of patients probably did not have HF (inter-rater agreement, kappa=0.35). Low-probability HF was predictive of being readmitted more frequently for non-HF causes (P=0.018), as well as documented arrhythmias (P=0.023), and age >60 years (P=0.006). Documented sleep apnea (P=0.035), percutaneous coronary intervention (P=0.006), non-white race (P=0.047), and B-type natriuretic peptide >400 pg/ml (P=0.007) were determined to be predictive of HF readmissions in this cohort. In conclusion, approximately 1 in 5 patients documented to have HF were found to have a low probability of actually having it. Moreover, the determination of low-probability HF was twice as likely to result in readmission for non-HF causes, and thus should be considered a determinant for all-cause readmissions in this population.