Net reclassification indices have recently become popular statistics for measuring the prediction increment of new biomarkers. We review the various types of net reclassification indices and their correct interpretations. We evaluate the advantages and disadvantages of quantifying the prediction increment with these indices. For pre-defined risk categories, we relate net reclassification indices to existing measures of the prediction increment. We also consider statistical methodology for constructing confidence intervals for net reclassification indices and evaluate the merits of hypothesis testing based on such indices. We recommend that investigators using net reclassification indices should report them separately for events (cases) and nonevents (controls). When there are two risk categories, the components of net reclassification indices are the same as the changes in the true-positive and false-positive rates. We advocate use of true- and false-positive rates and suggest it is more useful for investigators to retain the existing, descriptive terms. When there are three or more risk categories, we recommend against net reclassification indices because they do not adequately account for clinically important differences in shifts among risk categories. The category-free net reclassification index is a new descriptive device designed to avoid pre-defined risk categories. However, it suffers from many of the same problems as other measures such as the area under the receiver operating characteristic curve. In addition, the category-free index can mislead investigators by overstating the incremental value of a biomarker, even in independent validation data. When investigators want to test a null hypothesis of no prediction increment, the well-established tests for coefficients in the regression model are superior to the net reclassification index. If investigators want to use net reclassification indices, confidence intervals should be calculated using bootstrap methods rather than published variance formulas. The preferred single-number summary of the prediction increment is the improvement in net benefit.

No abstract

New methodology has been proposed in recent years for evaluating the improvement in prediction performance gained by adding a new predictor, Y, to a risk model containing a set of baseline predictors, X, for a binary outcome D. We prove theoretically that null hypotheses concerning no improvement in performance are equivalent to the simple null hypothesis that Y is not a risk factor when controlling for X, H0: P (D = 1|X, Y) = P (D = 1|X). Therefore, testing for improvement in prediction performance is redundant if Y has already been shown to be a risk factor. We also investigate properties of tests through simulation studies, focusing on the change in the area under the ROC curve (AUC). An unexpected finding is that standard testing procedures that do not adjust for variability in estimated regression coefficients are extremely conservative. This may explain why the AUC is widely considered insensitive to improvements in prediction performance and suggests that the problem of insensitivity has to do with use of invalid procedures for inference rather than with the measure itself. To avoid redundant testing and use of potentially problematic methods for inference, we recommend that hypothesis testing for no improvement be limited to evaluation of Y as a risk factor, for which methods are well developed and widely available. Analyses of measures of prediction performance should focus on estimation rather than on testing for no improvement in performance.

Roughly 9% of cerebrovascular events are missed at initial ED presentation. Risk of misdiagnosis is much greater when presenting neurologic complaints are mild, nonspecific, or transient (range 24%-60%). This difference suggests that many misdiagnoses relate to symptom-specific factors. Future research should emphasize studying causes and designing error-reduction strategies in symptom-specific subgroups at greatest risk of misdiagnosis.

No abstract