Generalizability of Cardiovascular Disease Clinical Prediction Models

A new study published in the Journal of the American Medical Association (JAMA) finds that cardiovascular disease clinical prediction models show promise for improving risk prediction, but more work is needed to increase their generalizability.

A Balanced Heart Is A Healthy Heart

Are you living from the heart? Find out. Take our brief quiz to get your own #LiveFromTheHeart tip sheet.

Take the quiz!

The study, “Generalizability of Cardiovascular Disease Clinical Prediction Models: 158 Independent External Validations of 104 Unique Models,” included 158 external validations of 104 unique models. The models were developed using data from 11 populations and applied to data from 27 additional populations.

The study found that the models demonstrated good discrimination, with an area under the curve (AUC) ranging from 0.64 to 0.76. However, there was significant heterogeneity in model performance across populations, with only moderate agreement between observed and expected AUCs. In addition, the study found that most of the models overestimated 10-year risk in low-risk populations and underestimated risk in high-risk populations.

The study’s authors note that these findings suggest that “cardiovascular disease clinical prediction models developed in 1 population may not perform well when applied to another population.” They say more research is needed to understand the sources of this heterogeneity and to develop strategies for improving the generalizability of these models.

This new study provides valuable insights into the generalizability of cardiovascular disease clinical prediction models. While the models show promise for improving risk prediction, more work is needed to increase their generalizability. The study’s findings suggest that more research is needed to understand the sources of heterogeneity in model performance and to develop strategies for improving the generalizability of these models.