Can an Internet-Based Health Risk Assessment Highlight Problems of Heart Disease Risk Factor Awareness? A Cross-Sectional Analysis

BACKGROUND: Health risk assessments are becoming more popular as a tool to conveniently and effectively reach community-dwelling adults who may be at risk for serious chronic conditions such as coronary heart disease (CHD). The use of such instruments to improve adults’ risk factor awareness and concordance with clinically measured risk factor values could be an opportunity to advance public health knowledge and build effective interventions. OBJECTIVE: The objective of this study was to determine if an Internet-based health risk assessment can highlight important aspects of agreement between respondents’ self-reported and clinically measured CHD risk factors for community-dwelling adults who may be at risk for CHD. METHODS: Data from an Internet-based cardiovascular health risk assessment (Heart Aware) administered to community-dwelling adults at 127 clinical sites were analyzed. Respondents were recruited through individual hospital marketing campaigns, such as media advertising and print media, found throughout inpatient and outpatient facilities. CHD risk factors from the Framingham Heart Study were examined. Weighted kappa statistics were calculated to measure interrater agreement between respondents’ self-reported and clinically measured CHD risk factors. Weighted kappa statistics were then calculated for each sample by strata of overall 10-year CHD risk. Three samples were drawn based on strategies for treating missing data: a listwise deleted sample, a pairwise deleted sample, and a multiple imputation (MI) sample. RESULTS: The MI sample (n=16,879) was most appropriate for addressing missing data. No CHD risk factor had better than marginal interrater agreement (κ>.60). High-density lipoprotein cholesterol (HDL-C) exhibited suboptimal interrater agreement that deteriorated (eg, κ<.30) as overall CHD risk increased. Conversely, low-density lipoprotein cholesterol (LDL-C) interrater agreement improved (eg, up to κ=.25) as overall CHD risk increased. Overall CHD risk of the sample was lower than comparative population-based CHD risk (ie, no more than 15% risk of CHD for the sample vs up to a 30% chance of CHD for the population). CONCLUSIONS: Interventions are needed to improve knowledge of CHD risk factors. Specific interventions should address perceptions of HDL-C and LCL-C. Internet-based health risk assessments such as Heart Aware may contribute to public health surveillance, but they must address selection bias of Internet-based recruitment methods.