Abstract P234: Cardiovascular Risk Prediction With Minimal Physiologic Variables
Introduction: Cardiovascular risk prediction relies on physiologic parameters, limiting the utility of tools such as the Pooled Cohort Risk Equation (PCRE) for self-administration outside clinical contexts. Tools that can be widely administered in places such as stores or pharmacies that offer blood pressure (BP) assessment could increase detection of high risk individuals and motivate care seeking. We developed a risk prediction approach that relies only on BP and otherwise self-reported variables and compared it with the PCRE.
Methods: We identified 16,833 in the REasons for Geographic AND Racial Differences in Stroke (REGARDS) study without coronary heart disease (CHD) and <80 years of age at baseline. CHD and stroke were expert adjudicated. Data included variables required to calculate the PCRE, as well as variables for the BP-self-report models (sociodemographics, medical history, and items assessed by validated questionnaires [e.g., Short Form-12]). We developed two predictive models for the composite endpoint of incident CHD or stroke using 1) survival trees (Tree) and 2) generalized additive models (GAM). For 5 iterations the cohort was randomly partitioned into equal training and testing subsets. Tree models and GAM were fit using the training subset and predicted risks calculated for the testing subset using the PCRE, Tree and GAM methods. C-statistics were calculated overall and stratified by age dichotomized at 65 for each model.
Results: For the overall cohort, the mean C-statistics over the 5 iterations were PCRE: 0.69 (Range 0.65-0.74); GAM BP-self-report: 0.70 (0.67-0.72); Tree BP-self-report: 0.59 (0.55-0.66). For participants aged over 65 years the GAM had the highest mean of 0.64 vs. 0.58 (PCRE) and 0.54 (Tree). See Figure.
Conclusion: A predictive statistical model using only self-reported data and systolic BP can have similar predictive performance to the PCRE, with potential marginal improvement noted for subjects over age 65 years.
Author Disclosures: M.M. Safford: B. Research Grant; Modest; Amgen. B. Research Grant; Significant; significant. G. Consultant/Advisory Board; Modest; diaDexus. G. Consultant/Advisory Board; Significant; modest. L. Colantonio: None. J. Richman: None.
- © 2015 by American Heart Association, Inc.