Affinity-based measures of medical diagnostic test accuracy

TL;DR

This paper introduces new affinity-based summary measures for diagnostic test accuracy, grounded in geometrical principles, with covariate-specific versions, Bayesian estimators, and demonstrated through simulations and prostate cancer data.

Contribution

It proposes novel affinity-based accuracy measures, develops covariate-specific versions, and introduces Bayesian estimators with theoretical analysis and practical application.

Findings

New affinity-based measures align with geometrical principles.

Covariate-specific measures assess test performance conditioned on predictors.

Bayesian estimators perform well in simulations and real data.

Abstract

We propose new summary measures of diagnostic test accuracy which can be used as companions to existing diagnostic accuracy measures. Conceptually, our summary measures are tantamount to the so-called Hellinger affinity and we show that they can be regarded as measures of agreement constructed from similar geometrical principles as Pearson correlation. A covariate-specific version of our summary index is developed, which can be used to assess the discrimination performance of a diagnostic test, conditionally on the value of a predictor. Nonparametric Bayes estimators for the proposed indexes are devised, theoretical properties of the corresponding priors are derived, and the performance of our methods is assessed through a simulation study. Data from a prostate cancer diagnosis study are used to illustrate our methods.