From blue November to broader diagnosis: The Youden index to evaluate the performance of any diagnostic tests

This is a methodological study. The goal is to describe and implement statistical tests based on the Youden index to evaluate the performance of diagnostic tests, using Prostate-Specific Antigen (PSA) as the primary example and including additional diagnoses to illustrate how these evaluations can be generalized.

Quantitative analysis using the Youden index was applied to assess diagnostic test performance across three different experimental designs: a single condition, two independent conditions (between-groups), and two dependent conditions (within-group), revisiting 2 × 2 tables from previous studies.

The Youden method combines sensitivity and specificity into a single index and requires only a 2 × 2 contingency table summary, incorporating both point estimates and confidence intervals. This allows for the evaluation of many studies where raw data are unavailable.

PSA seems insufficient for effective prostate cancer screening, despite numerous efforts over decades claiming improvements in sensitivity, specificity, or diagnostic capability. However, the statistical method presented here can be applied to any symptom, sign, or laboratory test, current or future. By providing open-source code, the authors aim to bridge the gap between statistical methods and their practical application, improving diagnostic processes. The R package and other supplemental materials to replicate this study are available on Harvard Dataverse at https://doi.org/10.7910/DVN/5QTMBW.