Confidence intervals for the area under the receiver operating characteristic curve in the presence of ignorable missing data

TL;DR

This paper compares various methods for constructing confidence intervals for the AUC of ROC curves when data has ignorable missingness, recommending multiple imputation with specific techniques based on missingness severity.

Contribution

It evaluates and compares different Wald-type confidence interval methods for AUC in the presence of missing data, providing practical recommendations.

Findings

Multiple imputation with logistic regression offers robust coverage.

Newcombe's Wald method performs well with less severe missingness.

Choice of confidence interval method is less critical when using multiple imputation.

Abstract

Receiver operating characteristic (ROC) curves are widely used as a measure of accuracy of diagnostic tests and can be summarized using the area under the ROC curve (AUC). Often, it is useful to construct a confidence intervals for the AUC, however, since there are a number of different proposed methods to measure variance of the AUC, there are thus many different resulting methods for constructing these intervals. In this manuscript, we compare different methods of constructing Wald-type confidence interval in the presence of missing data where the missingness mechanism is ignorable. We find that constructing confidence intervals using multiple imputation (MI) based on logistic regression (LR) gives the most robust coverage probability and the choice of CI method is less important. However, when missingness rate is less severe (e.g. less than 70%), we recommend using Newcombe's Wald method for constructing confidence intervals along with multiple imputation using predictive mean matching (PMM).