Accuracy of offspring reports of parental cardiovascular disease history: The Framingham offspring study

Background: Family history is used to infer the risk for heart disease; however, little is known about the accuracy of family history reports. Objective: To examine the accuracy of offspring reports of parental cardiovascular disease. Design: Validation study. Setting: Framingham Heart Study. Participants: Offspring participants of the multigenerational Framingham Heart Study with both parents in the original cohort. Measurements: 791 men and 837 women (mean age, 57 years) completed a family history questionnaire from 1995 to 1998. Offspring reports were compared with confirmed medical evidence of parental status, and positive and negative predictive values and likelihood ratios were calculated. Results: Positive reports of high blood pressure, diabetes, and high cholesterol levels in fathers were accurate: Positive predictive values were 83% (95% CI, 80% to 86%), 76% (CI, 70% to 82%), and 78% (CI, 73% to 83%), respectively. Corresponding positive predictive values for reports in mothers were 91% (Cl, 89% to 93%), 79% (CI, 73% to 85%), and 88% (CI, 84% to 92%), respectively. Positive predictive values for reports of paternal heart attack occurring before 55 years of age and for stroke occurring before 65 years of age were 28% (CI, 22% to 34%) and 43% (CI, 33% to 53%), respectively, whereas the positive likelihood ratios were 8.6 (CI, 6.8 to 10.9) and 11.2 (CI, 9.2 to 13.6), respectively. Negative predictive values for parental history reports were greater than 90%, except for high blood pressure and high cholesterol level (negative predictive values, 33% to 55%, and negative likelihood ratios, 0.47 to 0.88). Limitations: This study does not determine whether more accurate measures of family history would meaningfully improve estimation of cardiovascular risk. Conclusions: Negative parental history reports were reliable, except for hypertension and high cholesterol levels. Although reports of parental premature heart attack and stroke had high likelihood ratios, their predictive values were low because the prevalence of these conditions was low in parents. If patients were more aware of their parents' medical illnesses, they might be able to estimate their risk for disease more accurately and perhaps motivate themselves to follow a healthy lifestye.