A confidence ellipse for the Net Reclassification Improvement.

Area of Science:

• Biostatistics
• Medical Informatics
• Epidemiology

Background:

• Net Reclassification Improvement (NRI) is widely used to assess disease prediction model enhancements.
• Current interpretations and applications of NRI vary significantly across studies.
• Reliance on statistical significance of NRI estimates is common but increasingly discouraged.

Purpose of the Study:

• To address the inconsistent usage and interpretation of Net Reclassification Improvement (NRI).
• To propose an alternative method for evaluating variability in NRI estimates.
• To demonstrate the utility of confidence ellipses for NRI assessment using real-world data.

Main Methods:

• Investigated the application and interpretation of Net Reclassification Improvement (NRI).
• Proposed the use of confidence ellipses for visualizing variability in event and non-event NRI.
• Applied the proposed method to practical examples from the EPIC-Potsdam study.

Main Results:

• Highlighted the challenges in interpreting Net Reclassification Improvement (NRI) based solely on statistical significance.
• Demonstrated that confidence ellipses provide a robust measure of variability for NRI point estimates.
• Illustrated the practical application of confidence ellipses for enhanced NRI evaluation.

Conclusions:

• Confidence ellipses offer a superior approach to assessing the variability of Net Reclassification Improvement (NRI) estimates.
• Statistical testing for NRI should be reconsidered in favor of graphical methods like confidence ellipses.
• The proposed method enhances the reliability of evaluating disease prediction model improvements.