Binary regression: total gain in positive and negative predictive values.

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Prediction Modeling

Background:

• Risk prediction models are valuable for clinicians and patients.
• Model utility depends on confirming observed outcomes and accurately predicting future events.
• Existing metrics like sensitivity, specificity, and predictive values have limitations.

Purpose of the Study:

• To introduce standardized measures for evaluating the performance of binary outcome prediction models.
• To quantify the gain in positive and negative predictive values.
• To provide a tool for assessing prediction model accuracy.

Main Methods:

• Development of standardized total gain measures for positive and negative predictive values.
• Application to binary outcome prediction models.
• Focus on assessing the accuracy of predictor variables.

Main Results:

• The proposed measures offer a standardized approach to evaluating prediction model performance.
• These metrics provide insights into the accuracy of predicting binary outcomes.
• The measures facilitate comparison and selection of optimal prediction models.

Conclusions:

• Standardized total gain measures are effective tools for assessing prediction model accuracy.
• These measures enhance the evaluation of predictor variables for binary outcomes.
• The introduced concept aids in selecting optimal prediction models for future research.