Failure of Traditional Risk Factors to Adequately Predict Cardiovascular Events in Older Populations

Jarrod E. Dalton, PhD; Michael B. Rothberg, MD; Neal V. Dawson, MD; Nikolas I. Krieger, MS; David A. Zidar, MD, PhD; Adam T. Perzynski, PhD


J Am Geriatr Soc. 2020;68(4):754-761. 

In This Article

Abstract and Introduction


Background: Accurate assessment of atherosclerotic cardiovascular disease (ASCVD) risk across heterogeneous populations is needed for effective primary prevention. Little is known about the performance of standard cardiovascular risk factors in older adults.

Objective: To evaluate the performance of the American College of Cardiology/American Heart Association Pooled Cohort Equations (PCE) risk model, as well as the underlying cardiovascular risk factors, among adults older than 65 years.

Design and Setting: Retrospective cohort derived from a regional referral system's electronic medical records.

Participants: A total of 25 349 patients who were 65 years or older at study baseline (date of the first outpatient lipid panel taken between 2007 and 2010).

Measurements: Exposures of interest were traditional cardiovascular risk factors, as defined by inclusion in the PCE model. The primary outcome was major ASCVD events, defined as a composite of myocardial infarctions, stroke, and cardiovascular death.

Results: The PCE and internally estimated models produced similar risk distributions for white men aged 65 to 74 years. For all other groups, PCE predictions were generally lower than those of the internal models, particularly for African Americans. Discrimination of the PCE was poor for all age groups, with concordance index (95% confidence interval) estimates of 0.62 (0.60–0.64), 0.56 (0.54–0.57), and 0.52 (0.49–0.54) among patients aged 65 to 74, 75 to 84, and 85 years and older, respectively. Reestimating relationships within these age groups resulted in better calibration but negligible improvements in discrimination. Blood pressure, total cholesterol, and diabetes either were not associated at all or had inverse associations in the older age groups.

Conclusion: Traditional clinical risk factors for cardiovascular disease failed to accurately characterize risk in a contemporary population of Medicare-aged patients. Among those aged 85 years and older, some traditional risk factors were not associated with ASCVD events. Better risk models are needed to appropriately inform treatment decision making for the growing population of older adults.


Effective primary prevention of atherosclerotic cardiovascular disease (ASCVD) requires the accurate assessment of risk across a broad, heterogeneous population to identify those most likely to benefit from medical therapy. The American College of Cardiology and the American Heart Association (ACC/AHA) have established the Pooled Cohort Equations (PCE) risk model to inform initial ASCVD risk assessment and advocated for its use in decisions to prescribe cholesterol-lowering and antihypertensive therapies.[1,2]

The PCE includes standard cardiovascular risk factors (age, sex, race, blood pressure, total and high-density lipoprotein [HDL] cholesterol, smoking, diabetes, and antihypertensive medication use) and was derived from cohort data on 24 626 individuals aged 40 to 79 years. Several studies have investigated the discrimination (ability to differentiate between individuals who will and who will not experience ASCVD events, defined throughout this article as nonfatal myocardial infarction, nonfatal stroke, or cardiovascular death, unless otherwise noted) and calibration (agreement between estimated ASCVD event-free survival probabilities and observed frequencies) of the PCE. Validation studies have suggested both overestimation and underestimation of risk, depending on the contexts.[3–6] A recent study found that incrementally improved overall calibration is achievable with the use of more contemporary data.[7]

However, these models assume a fixed set of relationships within four distinct populations, defined according to combinations of race and sex. While age is included as a risk factor, the models do not allow for changes in risk relationships as patients age. While the models are not intended for individuals older than 79 years, some online risk calculators nonetheless provide estimates for adults older than 79 years by truncating age to 79 years*.[1] Clinicians who use ASCVD risk estimation tools that disallow age values of 80 years or higher, such as the official ACC tool,[2] may nonetheless truncate age to 79 years. Validation studies of these models to date have included relatively small numbers of individuals over age 80.[6]

Increases in the size and diversity of the older population have intensified the need for research examining the determinants of cardiovascular risk within that population. In this study, we (1) evaluated the performance of the PCE among older populations and (2) evaluated the extent to which accuracy of cardiovascular risk models that are based on established clinical risk factors might be improved by estimating model coefficients separately for older patients.

*Pooled Cohort Risk Assessment Equations. Accessed November 8, 2019.
ASCVD Risk Estimator Plus.!/calculate/estimate/. Accessed November 8, 2019.