Mathematical modelling to guide the development of novel tuberculosis diagnostics

Tuberculosis (TB) remains the world's deadliest infectious disease, yet significant gaps persist across the diagnostic care cascade. In many contexts, access to accurate diagnostics can be limited or delayed, with this contributing to onwards transmission and poorer health outcomes. Substantial investment in the research, development, and implementation of affordable, scalable, and context-appropriate diagnostic tools is thus needed. This thesis used mathematical modelling and cost-effectiveness analyses to generate evidence to guide decisions on the development and implementation of these tools across the care cascade. This included modelling to inform the accuracy requirements in the latest TB screening and diagnostic target product profiles. Here symptom-agnostic screening tools emerged as critical for large-scale active case-finding, although balanced sensitivity and specificity were noted to be essential for programme feasibility. This thesis also explored trade-offs in accuracy and access for both TB detection and drug-susceptibility testing, and highlighted that diagnostics which improved access to testing and results could achieve greater population-level impact even at lower accuracy thresholds. Together, these results underscore that diagnostic performance cannot be evaluated in isolation from the health-system contexts in which they are deployed. Additionally, cost as an element of programme feasibility was explored in analyses for South Africa. In this, molecular diagnostics were demonstrated to have high returns on investment, showing the importance of these tools in disease control. Overall, this thesis highlighted how modelling can be used to support decision makers and, when conducted iteratively and with stakeholder engagement, can be translated into national and global policy.