Looking to the Past to Predict Future Outcomes
Improving Cardiovascular Disease Prediction Models Using Periodic Health Screening Summary Measures
This article requires a subscription to view the full text. If you have a subscription you may use the login form below to view the article. Access to this article can also be purchased.
Cardiovascular disease (CVD) remains a leading cause of morbidity and mortality worldwide.1 As such, extensive efforts have been made particularly during the past 2 decades to develop multivariate statistical models capable of estimating the absolute risk of incident CVD outcomes within asymptomatic populations.2 The majority of clinically utilized risk prediction tools (eg, the Framingham score3) require input of single measurements of multiple risk factors to derive an individual’s absolute risk of CVD, the outcome of which can be used to guide clinical care decision making.4 Yet, model performance remains far from perfect.5 Attempts to improve performance has often included incorporating novel laboratory or imaging biomarkers as potential predictors; however, the magnitude of improvement in accuracy compared with existing models, and whether there is sufficient difference in an individual’s predicted risk to alter current clinical management,6 needs to be weighed up against the economic costs and patient burden associated with the new biomarker.7 To date, such attempts at model improvement have resulted in little change to tangible clinical outcomes.8
See Article by Cho et al
In this month’s issue of Circulation: Cardiovascular Quality and Outcomes, Cho et al9 posit that poor disease prediction using existing models may be due, in part, to the time-dependent nature of some CVD risk factors which is not adequately captured in single-measurement models and propose that more cost-effective improvement in performance could be achieved by incorporating long-term averages and variability of serially measured (standard) risk factors into prediction models. The …