Browsing by Author "Hesketh, Zsofia"

Rapid identification of infectious outbreaks is critical for the timely deployment of containment measures and for better prevention in the future. However, since surveillance mechanisms can be costly and complex to develop, lower-income countries may lack capacity to monitor prevalence data. Outbreaks therefore tend to spread extensively before authorities are notified. To overcome this, patient data can be collected and analysed more thoroughly to yield actionable epidemiological evidence. Using paediatric patient records from western Rwanda, the primary aim of this Thesis was to develop a syndromic surveillance methodology and accompanying visual dashboard to identify localities and times of year with higher prevalence of priority syndromes. The raw dataset of over 100,000 paediatric consultations was collected between December 2021 and July 2023, spanning 31 health facilities in two districts. A secondary aim was to uncover any statistically significant space-time dependencies in a sub-group of these syndromes, allowing for outbreak detection and evidence-based inference regarding seasonal, geographical, or socio-economic risk factors. The surveillance methodology consists of a pipeline of data pre-processing, binary syndromic variable coding and visual dashboard-building for six categories of syndromes: respiratory, febrile, diarrhoeal, nutritional, parasitic, and CNS. The prototype dashboard was built in PowerBI and comprises interactive graphs and maps to present prevalence results in an easily interpretable format for health policymakers. For the secondary aim, two scan statistics models were applied to detect the presence of significant high-prevalence clusters for six top interest syndromes. For each syndrome, spatio-temporal clusters were deemed significant when the p-value < 0.01. The descriptive visualisations generated from our syndromic data revealed several interesting trends. We found that respiratory and febrile syndromes exhibited clearer seasonal fluctuations, particularly increasing at the start and end of the rainy season. Diarrhoeal and malarial syndromes had strong relationships to health facility location, possibly pertaining to factors like elevation and proximity to the lake. On the other hand, nutritional syndromes appeared similarly prevalent throughout the year and across all health facilities. Our statistical dependency analyses also yielded meaningful results, finding at least one significant space-time cluster in four of the six selected syndromes. These results demonstrate the utility of our surveillance pipeline and visual dashboard for uncovering previously unknown epidemiological trends. If data is consistently collected and consulted by policymakers, outbreaks may be caught early and averted ahead of time. They also suggest that the prevalence of certain syndromes is significantly linked to space-time variables like health facility, village of origin and month of occurrence. In the future, further inferential and predictive analyses, like regression modelling, may be applied to evaluate the independent effect of more specific variables like rainfall, temperature, average income and sanitation levels.