Validation of self-reported family history of myocardial infarction using nationwide health care data

Abstract Background Self-reported family history of myocardial infarction (MI) is used clinically and in research to estimate cardiovascular risk. However, previous validations of self-reported family history are limited to small study samples, and to death-certificates or self-reported history as reference. We sought to assess the accuracy of self-reported family history of MI in a population-based cohort, using national register data as reference. Methods We included 25,302 subjects from the Swedish CArdioPulmonary bioImage Study (SCAPIS), of which both parents were identified in the Swedish Multi-Generation Register. Self-reported family history of MI in parents and siblings, with and without regards to age of diagnosis, were derived from the SCAPIS questionnaire. Reference occurrences for MI in parents and siblings were retrieved from the Swedish National Patient Register and Cause of Death Register. Data on education were collected from the SCAPIS questionnaire, whereas data on income were retrieved from national tax registers. Cohen’s κ, sensitivity, specificity and predictive values for each definition of family history were calculated with register-derived diagnoses as reference. Analyses were also performed in subgroups by sex, educational attainment and quintiles of register-derived income. Results The median age was 57.4 years. Generally, the agreement and sensitivity of self-reported and register-verified family history were moderate (table). A reported family history of MI in any parent or sibling rendered a κ of 0.491 (95% confidence interval CI 0.480 – 0.503), a sensitivity of 57.6% (95% CI 56.6% – 58.6%) and specificity of 89.0% (95% CI 88.5% – 89.5%) (Panel A). Reported maternal and paternal MI showed similar measures of agreement, however, reported early-onset disease had a lower κ (0.396, 95% CI 0.367 – 0.426 for any parent) although the sensitivity was 58.1% (95% CI 54.1% - 62.0%)(Panel B). Male subjects generally reported family history of MI less accurately than female counterparts. Subjects with a completed university degree and a high disposable income reported parental history more accurately, however, only for history without age restriction. Conclusions Self-reported family history of MI notably underestimates the prevalence of disease assessed by national registers, especially for early-onset disease which also is the clinically most relevant.Measures of agreement and Venn-diagrams