Relative risk forests for exercise heart rate recovery as a predictor of mortality

Hemant Ishwaran, Eugene H. Blackstone, Claire E. Pothier, Michael S. Lauer

Research output: Contribution to journalReview articlepeer-review

33 Scopus citations


Recent studies have confirmed heart rate fall after treadmill exercise testing, or heart rate recovery, as a powerful predictor of mortality from heart disease. Heart rate recovery depends on central reactivation of vagal tone and decreased vagal activity is a risk factor for death. If heart rate recovery is defined as the fall in heart rate after 1 minute following peak exercise, then a heart rate recovery value of 12 beats per minute (bpm) or lower has been shown to be a good prognostic threshold for identifying patients at high risk. Although this finding establishes a simple, useful relationship between heart recovery and mortality, a working understanding of how heart rate recovery interacts with other characteristics of a patient in determining risk of death is still largely unexplored. Such knowledge, addressed in this article, could improve the prognostic value of the exercise test. Our analysis is based on over 23,000 patients who underwent exercise testing. A rich assortment of data was collected on these patients, including clinical and physiological information, heart rate recovery, and other exercise test performance measures. Our approach was to grow relative risk forests, a novel method that combines random forest methodology with survival trees grown using Poisson likelihoods. Our analysis reveals a complex relationship between peak heart rate, age, level of fitness, heart rate recovery, and risk of death.

Original languageEnglish (US)
Pages (from-to)591-600
Number of pages10
JournalJournal of the American Statistical Association
Issue number467
StatePublished - Sep 2004
Externally publishedYes


  • Cox regression
  • MARS
  • Proportional hazards
  • Random forests
  • Relative risk trees
  • Stochastic variable selection

ASJC Scopus subject areas

  • Mathematics(all)
  • Statistics and Probability


Dive into the research topics of 'Relative risk forests for exercise heart rate recovery as a predictor of mortality'. Together they form a unique fingerprint.

Cite this