Background: Self-reports of heart failure (HF) are relied upon to quantify the burden of HF in the community. While self-report of HF is often considered subject to recall bias, accuracy of self-reported HF following a HF hospitalization has not been estimated.
Methods: Hospital admission of ARIC cohort members is monitored on an ongoing basis. Starting in 2005, hospital records with HF were abstracted and events validated with a standardized protocol based on review of history and physical examination, echocardiography, nuclear imaging, chest X-ray, catheterization reports, cardiac markers, and discharge summary. HF was classified as acute decompensated HF (ADHF) or chronic stable HF (CSHF). Also starting in 2005, ARIC participants were asked annually to report whether a physician had told them they have HF or a weak heart. Those self-reporting HF were asked to authorize release of medical information and their providers of medical care were sent a survey to confirm their patient’s HF status, HF characteristics, and treatment status.
After excluding participants with a HF hospitalization prior to 2005 (defined by an ICD-9 discharge code of 428 in any position), self-reported HF following an incident HF hospitalization was quantified as the proportion of hospitalized participants who self-reported HF (sensitivity) and the proportion of non-hospitalized participants who did not self-report HF (specificity). The proportion of self-reported HF confirmed by physicians was calculated among those with a returned survey.
Results: Of the 10,509 ARIC participants alive from 2005-2013, 512 were hospitalized with ADHF and 154 were hospitalized with CSHF prior to self-reporting their HF status and 328 self-reported having HF. Comparing self-reported to hospitalized HF, sensitivity was poor (0.28) and specificity excellent (0.99), although some participants who self-reported HF but were not hospitalized may have been diagnosed in the outpatient setting. A physician survey was available for 92 participants who self-reported HF. In 64 (70%; 95% confidence interval (CI): 61, 80) instances, self-reported HF was confirmed by their physician. Of the latter, 23 (36%; 95% CI: 24, 48) were reported as diastolic dysfunction while 19 (30%; 95% CI: 18, 41) and 8 (13%; 95% CI: 4, 21) were characterized as systolic dysfunction or mixed, respectively. HF type was not specified by the physician in 14 cases. Seventy-eight percent (95% CI: 68, 89) of patients with confirmed HF, the physician also reported HF treatment.
Conclusion: Self-report of HF was highly specific and quite insensitive calling into question the validity of self-reports of HF. Our results underscore the need for methods more effective than self-report to identify persons with clinically manifest HF.