Search

Tuberculosis (TB) is among the leading infectious killers in the world. Despite increases in case detection, there are still three million people with TB who are undiagnosed, untreated or not reported to national programmes. Scaling up of systematic TB screening and use of more sensitive diagnostic tools is required to identify missing people with TB. Chest X-ray (CXR) and computer aided diagnostic (CAD) software for TB are now recommended by the World Health Organization as useful tools for TB screening and triage. Policymakers in lower-and middle-income countries must, however, contextualize this guidance and consider which tools to adopt, and how they will incorporate these into their algorithms. Operational modelling using the Witness package, a visual and interactive modelling tool, has been demonstrated to aid policymakers in decision making processes. It, however, remains unclear where operational modelling would best fit in the policy process and its influence and usefulness in the policy development process has not been formally studied. The overall aim of this study was to generate new knowledge related to CXR use in TB screening, develop an operational model using this information and other sources, and assess if the modelling approach is a feasible technique to influence TB policy in Kenya. This mixed-methods, multidisciplinary study incorporated clinical research, operational modelling and policy analysis. Secondary retrospective quantitative analysis was conducted on cross-sectional study data using individual-level participant CXR data from adult community members who took part in the 2016 Kenya National TB Prevalence Survey. An operational model was built to assess the impact of scale up of CXR in different screening and diagnostic algorithms. Finally, a qualitative, retrospective, and prospective analysis of lung health policy in Kenya was conducted employing the policy triangle and heuristic frameworks. This process generated novel findings related to CXR use in TB screening with important health policy implications. Firstly, the accuracy of CAD met the optimal WHO target product profile for a community TB screening tool. The specificity was lower in adults with previous TB and those aged 41 years or older hence an adaptive approach to setting CAD thresholds will be required to optimize use. Secondly, the use of CXR for TB population-based studies identified many patients with non-TB related abnormalities that would likely be missed by use of CAD. Implementation of CXR TB screening offers an opportunity to integrate disease screening efforts and improvement on all future CAD versions would require scoring for non-TB abnormalities. The modelling demonstrated that a strategy using CXR-CAD screening for all, then GeneXpert though the most expensive, had the ability to identify more persons with TB. Though a relatively new concept, operational modelling was an acceptable and feasible tool to aid in TB policymaking in Kenya and a framework for its adoption in policymaking was developed.