Related Experiment Videos

How generalizable are coronary risk prediction models? Comparison of Framingham and two national cohorts

Y Liao1, D L McGee, R S Cooper

  • 1Department of Preventive Medicine and Epidemiology and the Division of General Internal Medicine, Loyola University Stritch School of Medicine, Maywood, IL 60153, USA.

American Heart Journal
|April 30, 1999
PubMed

Insights

The Framingham Heart Study model reasonably ranks coronary heart disease (CHD) risk but overestimates mortality in men. Newer national data show improved CHD survival, especially in women.

Area of Science:

  • Cardiovascular epidemiology
  • Public health research
  • Biostatistics

Background:

  • Established coronary heart disease (CHD) risk prediction models, like the Framingham Heart Study, are based on older data.
  • Significant declines in CHD mortality rates have occurred due to improved risk factor management and medical interventions.
  • The generalizability of older models to current populations is uncertain.

Purpose of the Study:

  • To compare the predictive accuracy of the Framingham CHD risk model with functions derived from more recent national health surveys.
  • To assess the performance of risk prediction models across different US population cohorts.

Main Methods:

  • Utilized data from the Framingham Heart Study (1954-1958), NHANES I (1971-1975), and NHANES II (1976-1980).
  • Included participants aged 35-69, free of CHD at baseline, and followed for up to 24 years.
  • Compared Cox regression coefficients for major risk factors (age, blood pressure, cholesterol, smoking) and evaluated model performance using receiver operating characteristic (ROC) curves.

Main Results:

  • Heterogeneity was observed in individual risk factor coefficients among the cohorts, particularly in men.
  • Despite heterogeneity, risk functions collectively ranked individual risk similarly across cohorts (ROC areas: men 0.71-0.76, women 0.76-0.81).
  • The Framingham model accurately predicted CHD survival in women but overestimated CHD mortality rates in men from NHANES I and NHANES II.

Conclusions:

  • The Framingham risk model offers a reasonable rank ordering of CHD mortality risk for the US white population (1975-1990).
  • The model's prediction of absolute CHD mortality risk is less accurate for contemporary populations.
  • Updated models incorporating recent cohort data may be necessary for precise absolute risk prediction.
Abstract

Related Concept Videos