Using data analysis and mathematical modelling to study tuberculosis contact tracing in London, with reference to the national strategy and guidance

Background: In January 2015, Public Health England and NHS England published a collaborative TB strategy for the years 2015-20; this strategy highlighted contact tracing as a key element. In January 2016, the National Institute of Health and Care Excellence (NICE) made changes to the UK TB guidance, including no longer recommending that contacts of non-pulmonary TB cases be screened. This thesis attempts to address several issues arising from these policy documents. Methods: I utilized a range of quantitative approaches. I undertook a cohort analysis of TB cases in London between 2012-15 (inclusive), including logistic regression, to understand contact tracing outcomes in London, and how these differed between population subgroups. To understand the impact of changes to NICE guidance I carried out an economic analysis using a simple static model. I then utilized a pairwise transmission model to understand how transmission differs between those with primary and reactivation disease. Results: In London from 2012-15, 91% of pulmonary index cases had at least one contact identified (a median of four per case), and 86% of these identified contacts were evaluated. In this period, 80% of those contacts determined to have TB had an isolate that was indistinguishable from their index case, implying probable transmission. Assuming each contact with PTB infects 1 person/month, screening contacts of ETB cases costs £78000/QALY (95% CI: 39000 to 140000). Pairwise modelling suggests that the number of infections generated by those with a reactivation disease is only slightly greater than those with disease following recent infection. Conclusions: While contact tracing outcomes in London were good relative to similar countries and previous UK studies, our results highlight several groups for whom outcomes are worse. Our results also show that the impact of contact tracing is not limited to those occasions where transmission between index cases and contact has occurred. Our results also show that screening contacts of non-pulmonary index cases is almost certainly not cost-effective at a £30000/QALY threshold. More work is required if pairwise modelling is to be used effectively to model M. Tb transmission.