Screening as an application of decision theory

Area of Science:

• Decision theory
• Medical screening
• Biostatistics

Background:

• Optimizing screening procedures is crucial for early disease detection.
• Current methods like the Youden index may not fully account for differential costs of misclassification.
• Accurate threshold setting balances sensitivity and specificity effectively.

Purpose of the Study:

• To develop a novel decision-theoretical framework for determining optimal screening test thresholds.
• To incorporate the economic and clinical consequences of false positives and false negatives into threshold selection.
• To provide a flexible alternative to traditional methods like maximizing the Youden index.

Main Methods:

• Developed a decision-theoretical approach using custom loss functions.
• Elicited loss functions from subject-matter experts to quantify decision consequences.
• Investigated various loss function classes and outcome distributions.
• Outlined extensions for mixture distributions and composite loss functions.

Main Results:

• The proposed method offers a principled way to set screening thresholds based on decision consequences.
• Demonstrated the method's applicability through simulated data and real-world datasets.
• Showcased flexibility in handling different loss functions and data distributions.

Conclusions:

• The decision-theoretical approach provides a robust and adaptable framework for optimizing screening procedures.
• Incorporating specific loss functions leads to more informed threshold setting than general optimization criteria.
• This method enhances the clinical utility and cost-effectiveness of screening programs.