Exploratory Geospatial Analysis and Modelling to Support Trachoma Elimination and Surveillance

Trachoma, caused by ocular infection with the bacterium Chlamydia trachomatis, is the leading infectious cause of blindness worldwide. Trachoma prevalence surveys are essential to target and prioritise interventions, measure progress toward trachoma elimination goals, and help validate a country as having eliminated trachoma as a public health problem. In an increasingly resource-constrained global health programme environment, there is a need for enhanced tools for identifying areas at higher risk of having continued high trachomatous inflammation—follicular (TF) prevalence in children aged 1–9 years (TF1–9). Geospatial analysis is one such tool, helping identify areas with higher TF1–9 within and between trachoma-endemic evaluation units (EUs). This thesis aimed to investigate geospatial patterns in trachoma surveys and develop a method to identify “zones of concern” (defined here as areas of higher TF1–9, based on pooled data from neighbouring clusters). A systematic literature review was conducted to identify potential environmental, climatic, sociodemographic, socioeconomic, and geospatial factors associated with TF and rachomatous trichiasis ((TT); the late-stage blinding sequalae of trachoma). Higher TF prevalence was associated with areas of lower precipitation, temperature, and elevation, as well as with areas that were more rural, were less accessible, and had limited infrastructure. Higher TT prevalence was associated with areas of higher aridity and lower nightlight density. In addition to identifying associated factors, the review highlighted a gap in the literature regarding systematic, multi-country analyses of the relationship between these covariates and trachoma outcomes. This gap was addressed in the subsequent analysis using geospatial modelling across 20 countries and 24 covariates. Covariate coefficient directions varied by country and EU, with greater stability observed at finer spatial scales (e.g., districts) for both TF1–9 and TT prevalence in adults aged ≥15 years (TT≥15). To explore spatial structure in trachoma survey data, analyses were conducted across 20 countries to assess spatial heterogeneity, scale, and dependency in TF1–9 and TT≥15. Some EUs showed sufficient spatial structure for geospatial analysis, but in most cases, spatial autocorrelation was weak. These findings highlight the complexity of TF1–9 and TT≥15 spatial patterns and the limitations of applying broad-scale geospatial models to routine trachoma prevalence survey data. Building on these results, a novel approach was developed to identify zones of concern. The approach was piloted in two areas in Uganda and pooled prevalence data within a given distance (buffer) of each survey location. The approach showed flexible application across EUs, with key decisions including buffer size, prevalence thresholds, and identification method (simple threshold, overlap, or Local Moran’s I). This thesis demonstrates that geospatial analysis is possible with trachoma survey data but the approach should be tailored to intended use, data availability, and local context. These findings are relevant for both country programme decision-makers and researchers supporting trachoma- endemic countries in applying geospatial models for trachoma elimination.