Abstract Body: Introduction: Observational studies often use participants’ self-reported health data to identify clinical endpoints. Information on the accuracy of self-reported cardiovascular and pulmonary events among Hispanic/Latino populations is lacking.
Research Question: Are self-reported hospitalizations for cardiovascular and pulmonary events in an observational study of Hispanic/Latino participants consistent with physician adjudication of medical records?
Methods: We assessed 526 participants (mean age: 54 years; 64% female) who self-reported hospitalizations during 2008-2016 annual follow-up interviews in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). Participants reported all hospitalizations within the past year and the reason for each hospitalization. Medical records for reported hospitalizations were abstracted and reviewed by two physicians to determine event classification. Disagreements between reviewers were adjudicated by a third physician. Reviewers classified each event as myocardial infarction (MI), heart failure (HF), stroke, or chronic lower respiratory disease (CLRD; defined as asthma, chronic obstructive pulmonary disease, chronic bronchitis, or emphysema). We calculated the positive predictive value (PPV) of these self-reported events as the proportion verified by physician review. We examined the PPV by gender and acculturation (measured by language preference: Spanish or English).
Results: There were 90 MI, 96 HF, 112 stroke, and 538 CLRD events self-reported during follow-up. Using physician adjudication as the standard criterion, the PPV of self-reported events was 42% for MI, 25% for HF, 54% for stroke, and 18% for CLRD. The PPV of stroke among females was 43% (95% CI: 32%, 56%); the PPV among males was 65% (95% CI: 53%, 76%). The PPV of CLRD among females was 20% (95% CI: 17%, 24%); the PPV among males was 11% (95% CI: 7%, 18%). For CLRD, the PPV among those whose preferred language was Spanish was 16% (95% CI: 12%, 20%); the PPV among those whose preferred language was English was 23% (95% CI: 18%, 29%). Among self-reported MI events not confirmed by physician adjudication, the most common diagnosis was unspecified chest pain.
Conclusions: In an observational study of Hispanic/Latino participants, the PPV of self-reported cardiovascular and pulmonary events is low and varies by demographics. Relying solely on self-report to ascertain cause-specific clinical events may contribute substantial bias in observational studies.