Validity of a simple ST-elevation acute myocardial infarction risk index: are randomized trial prognostic estimates generalizable to elderly patients?

Journal Article (Journal Article)

BACKGROUND: Risk-stratification scores derived from randomized clinical trial (RCT) data should be evaluated in community-based populations. A simple risk-stratification index for patients with ST-segment elevation myocardial infarction derived from an RCT population was recently proposed, but it has not been validated in a community-based cohort. METHODS AND RESULTS: We evaluated the simple risk index using data from 49 711 patients > or =65 years of age hospitalized with ST-elevation myocardial infarction. We evaluated the distribution of patients in the 5 simple risk index groups, compared observed and published 30-day mortality rates, and assessed the score's discrimination and calibration. The simple risk index provided poor discrimination (c=0.62) and calibration (goodness of fit P<0.001) for survival at 30 days. Risk score distribution was skewed, because two thirds (66.1%) of all patients were classified in the highest-risk group, whereas fewer than 11.0% were classified in the 3 lowest-risk groups. Thirty-day mortality estimates were lower than those observed in the cohort (risk group 2 to 5: 1.9% to 17.4% versus 5.3% to 27.9%). Risk index discrimination, calibration, score distribution, and mortality estimates were worse among patients who did not receive acute reperfusion therapy than among those who did. CONCLUSIONS: The limited performance of the simple risk index highlights the limitations of applying prognostic models derived in RCT populations to the general population of patients 65 years and older. Prognostic scores must be validated in community-based cohorts before integration into clinical practice.

Full Text

Duke Authors

Cited Authors

  • Rathore, SS; Weinfurt, KP; Gross, CP; Krumholz, HM

Published Date

  • February 18, 2003

Published In

Volume / Issue

  • 107 / 6

Start / End Page

  • 811 - 816

PubMed ID

  • 12591749

Electronic International Standard Serial Number (EISSN)

  • 1524-4539

Digital Object Identifier (DOI)

  • 10.1161/01.cir.0000049743.45748.02


  • eng

Conference Location

  • United States