Computer-aided detection pragmatic threshold calibration in TB active case finding: A mixed-methods study in Tondo,...

Area of Science:

• Public Health
• Medical Imaging
• Infectious Disease Control

Background:

• Tuberculosis (TB) remains a significant global health challenge, particularly in high-burden urban settings.
• Computer-aided detection (CAD) offers potential to enhance active case finding (ACF) efficiency.
• Optimizing CAD thresholds is crucial for balancing sensitivity and operational feasibility in TB screening programs.

Purpose of the Study:

• To develop and assess a pragmatic approach for calibrating CAD thresholds in a TB ACF project in the Philippines.
• To evaluate the feasibility and acceptability of CAD-assisted TB screening in a high-burden urban environment.
• To establish a framework for tailoring CAD deployment to local contexts and system capacities.

Main Methods:

• A mixed-methods approach combining retrospective data analysis, prospective monitoring, and semi-structured interviews with healthcare workers and stakeholders.
• Integration of CAD4TB v.7 software into ACF activities in Tondo, Manila, with referrals based on CAD scores above a dynamic threshold.
• Retrospective analysis of pre-CAD data to set initial thresholds, followed by prospective monitoring and adjustments based on referral rates, screening yield, and operational feedback.

Main Results:

• Over 14,739 individuals were screened, identifying 633 TB cases with an overall screening yield of 4.3%.
• Dynamic calibration of CAD thresholds, adjusted based on referral rate targets and screening yield, improved ACF efficiency.
• Stakeholder input and real-time monitoring facilitated adjustments, balancing diagnostic sensitivity with operational constraints in an underserved population.

Conclusions:

• Pragmatic calibration of CAD thresholds, informed by referral targets, screening yield, and operational factors, enhances TB ACF efficiency and user confidence.
• Adaptive threshold adjustments based on real-time monitoring and stakeholder feedback are vital for optimizing CAD performance in diverse settings.
• The developed framework provides a scalable model for deploying CAD technology in TB control programs, adaptable to local epidemiology and resource availability.