Bayesian reclassification statistics for assessing improvements in diagnostic accuracy.

Published

Journal Article

We propose a Bayesian approach to the estimation of the net reclassification improvement (NRI) and three versions of the integrated discrimination improvement (IDI) under the logistic regression model. Both NRI and IDI were proposed as numerical characterizations of accuracy improvement for diagnostic tests and were shown to retain certain practical advantage over analysis based on ROC curves and offer complementary information to the changes in area under the curve. Our development is a new contribution towards Bayesian solution for the estimation of NRI and IDI, which eases computational burden and increases flexibility. Our simulation results indicate that Bayesian estimation enjoys satisfactory performance comparable with frequentist estimation and achieves point estimation and credible interval construction simultaneously. We adopt the methodology to analyze a real data from the Singapore Malay Eye Study. Copyright © 2016 John Wiley & Sons, Ltd.

Full Text

Duke Authors

Cited Authors

  • Huang, Z; Li, J; Cheng, C-Y; Cheung, C; Wong, T-Y

Published Date

  • July 10, 2016

Published In

Volume / Issue

  • 35 / 15

Start / End Page

  • 2574 - 2592

PubMed ID

  • 26875442

Pubmed Central ID

  • 26875442

Electronic International Standard Serial Number (EISSN)

  • 1097-0258

Digital Object Identifier (DOI)

  • 10.1002/sim.6899

Language

  • eng

Conference Location

  • England