In 2020, Tuberculosis (TB) was the second-leading cause of death by a single infectious agent, trailing only COVID-19. Improved strategies for case finding and prompt treatment initiation are therefore a public health priority. However, routine TB case finding through health facilities depends on an individual recognising that they have TB symptoms and taking the initiative to seek care, and is not sufficient to control TB transmission. This is because a considerable number of people with TB symptoms delay and sometimes never attend healthcare facilities, meaning that they remain infectious in the community, and are at high risk of severe illness and death. By contrast, while community-based TB case-finding interventions are complementary to facility-based services, they are substantially more resource intensive, logistically challenging, and are generally only justifiable when targeted at groups of people at high risk of TB disease. Because the prevalence of microbiologically confirmed pulmonary TB is difficult to measure and only rarely exceeds 1% even in the highest burden settings, identifying population groups likely to have a high burden of undiagnosed TB disease who could benefit from community-based case finding interventions can be challenging. Therefore, the aim of this thesis was to investigate how statistical spatial modelling of epidemiological patterns of spatial heterogeneity in urban TB epidemiology in Blantyre could be used to improve the targeting of community-based active case-finding (ACF) interventions for TB, and so increase the efficiency and effectiveness of the delivery of these interventions. First, a systematic review of the effectiveness of spatially-targeted community ACF interventions demonstrated that spatially-targeted interventions are feasible and that they have potential as alternative design strategies for community ACF, but that more rigorous approaches to design and analysis are required (Chapter 3, Khundi et al. 2021). The systematic review identified ten studies of spatially targeted interventions from six countries between 1 January 1993 and 22 March 2021: three directed against TB, three against leprosy, three against malaria, and one against HIV. Although data were limited, and understanding of effectiveness was limited by high risk of bias (particularly in classification of hotspots and ascertainment of outcomes), this demonstrated that spatially-targeted interventions have real potential to identify communities with a higher yield of identified cases, communities with a high prevalence of cases, and hence potential for accelerating reductions in the TB case notifications rates. Second, multi-level spatial regression modelling of TB case fatality rates across the city of Blantyre (Chapter 4, Khundi et. al., 2021) – in which we evaluated 4397 newly-diagnosed TB cases, 10.9% (479) of whom died – found strong evidence that, while undergoing TB treatment, age, being HIV positive, and distance to TB treatment clinic were associated with an increased odd of death. Distance to TB treatment clinic is a proxy of ease of access to care: individuals that have to travel longer distances to get health care are at an increased risk of adverse health. In our study population, distance increased the odds of death only for patients that were registered at the referral clinic but not a primary health care clinic. This is consistent with the hypothesis that high quality facility-based TB screening and care is complementary to community-based interventions in reducing TB mortality. Third, using data from a citywide TB prevalence survey and Malawi Liverpool Wellcome’s Blantyre enhanced TB surveillance system (Chapter 5, Khundi et al., 2022), we developed a novel Bayesian statistical spatial modelling approach to enable identification of TB hotspot neighbourhoods, based on ranking of 72 neighbourhood prevalence-to-notification ratios. In 2019, the prevalence of microbiologically-confirmed TB in Blantyre was 215 per 100,000 population. The model derived mean neighbourhood prevalence-to-notification ratio was 4.49 (95% credible interval [CrI]: 0.98–11.91, range: 1.70–10.40, standard deviation: 1.79). This indicates that, overall, there remains a substantial burden of undiagnosed TB. Our model should support researchers and health workers in other settings with similar characteristics to urban Blantyre in identifying potential hotspot neighbourhoods without the need for conducting a full prevalence survey. Looking forward, the approaches developed in this thesis, need to be validated, and then further developed and applied in other similar urban settings to prioritise neighbourhoods by burden of undiagnosed TB, and hence efficiently direct community based ACF interventions.