Mathematical models of the impact of host and environmental risk factors on the incidence of Tuberculosis (TB) within a national cohort

Tuberculosis (TB) is an infectious disease that can prove fatal if untreated. Despite a re-emergence of TB in Ireland, research has failed to provide insight to the causes of recent increases. The study acquires national surveillance data and systematically identifies a number of significant trends related to TB. From these findings, epidemiological models are constructed and simulated and put through various scenarios. The primary aim of the study is to develop, simulate, and forecast deterministic epidemic models for the spread of TB with application to an Irish setting. The study utilities anonymised cross-sectional surveillance data acquired from the Health Protection Surveillance Centre (HPSC). Ethical approval was granted by the Adelaide and Meath Hospital, and ethical approval recognized by Trinity College. Two SEIR (Susceptible Exposed Infection Recovered) models consisting of systems of ordinary differential equations (ODEs) were developed and simulated. Approximate Bayesian Computation and Metropolis-Hastings inference algorithms were implemented to estimate the basic reproductive number, R0, and other model parameters in preparation for simulation and forecasting. Statistically significant differences were calculated between native and foreign-born TB notifications, which is in line with previously published literature. Significant seasonality was discovered in Irish TB notifications, which has not been previously shown in published research. Migrant and seasonal SEIR models were presented for analysis. The models forecast a modest decline in notifications nationally up until 2023. Key parameters were identified in each model to help strategies that involve population management. A scenario analysis conducting numerical simulations calculated marginal increases in notifications (from one to three cases annually) when a change in vaccination procedure from universal vaccination to selective vaccination is considered. Numerical simulations of the seasonal epidemic model suggest that it would be more cost effective to implement an infection control strategy such as vaccination during the period from January to June, rather than all year round. Further research is required to investigate the causes and effects of seasonality in TB notifications and whether foreign-born and native-born populations interact with each other in an Irish setting. The epidemiological parameters estimated in this thesis form a basis for future surveillance and modelling to take place in Ireland and other settings. The top contributing countries of the foreign-born population should be surveyed to ensure these trends continue, as variance in notifications for this group is larger than that of the native-born population. Further research is required to model vulnerable populations in Ireland such as the homeless, refugee, and unemployed populations