MATHEMATICAL MODELLING FOR REAL-TIME AND PROSPECTIVE ASSESSMENT OF INFECTIOUS DISEASES

Infectious diseases are a major threat to population health and often require decisions to be made under uncertainty about current and future dynamics. Mathematical models provide a way to combine mechanistic knowledge with surveillance data and to evaluate alternative intervention strategies. Building on this perspective, this thesis develops and applies mathematical modelling frameworks for the real-time and prospective assessment of infectious diseases, with a focus on how surveillance, forecasting, and intervention analysis can support public health decision-making. After introducing the core tools of infectious disease modelling, the thesis addresses three applied problems. First, it investigates how incomplete recent data, caused by reporting delays, influence estimates of the time-varying reproduction number. A nowcasting algorithm based on recurrent updating is retrospectively validated using COVID-19 surveillance data from Italy. By comparing estimates obtained with and without nowcasting, the analysis quantifies how correcting for right truncation can improve the detection of changes in transmission, with direct implications for the timely adjustment of control measures. Second, the thesis examines short-term forecasting of norovirus outbreaks on cruise ships using data from seven epidemics. Different semi-mechanistic models are formulated and compared against multiple endpoints: incidence by symptom onset date, incidence by diagnosis date, and final outbreak size. Models that incorporate superspreading of secondary cases and reduced transmission due to case isolation provide the most accurate projections, highlighting the importance of capturing key epidemiological mechanisms when generating real-time forecasts in confined settings. Third, a transmission model for Chlamydia Trachomatis (CT) among men who have sex with men enrolled in an HIV pre-exposure prophylaxis (PrEP) programme in the province of Verona, Italy, is developed and calibrated to multiple data sources for 2018 to 2024. The model is used to explore alternative screening intervals, strategies excluding asymptomatic screening, and scenarios that introduce doxycycline post-exposure prophylaxis for high-activity individuals. The results show how the balance between intensified or reduced screening, behavioral changes among PrEP users, and potential antibiotic use shapes CT prevalence over the next decade. Overall, the thesis illustrates how flexible modelling frameworks, grounded in data and tailored to specific pathogens and settings, can be used to characterise transmission dynamics, improve outbreak forecasts, and compare intervention strategies.