Analyses of complex networks for clustering of symptoms to characterise severe schistosomiasis in rural Uganda

external supervisor

Professor Renaud Lambiotte, Mathematical Institute, Univeristy of Oxford

background

Symptoms remain a critical form of preliminary diagnosis or triage in low-income countries where laboratory diagnostics may be limited. Symptoms also are used as the first indicator for the need for care, as a patient must recognise their health problem before seeking care. Individuals may have clusters of conditions with diverse causes that complicate diagnoses and increase the burden on health systems due to repeated patient visits. At the population level, there is an interest in understanding simply what symptoms tend to be grouped together to identify what regimens of available medicines or disease management strategies should be available at clinics. The structure of symptom clusters can provide important information on the severity of comorbidities, future conditions, and dominant conditions (Chami et al 2018 J. Roy. Soc. Interface). 

RESEARCH EXPERIENCE, RESEARCH METHODS AND TRAINING

This project will develop methods to identify the clustering of symptoms across and within patients in rural poor villages of rural Uganda. The focus of the project will be on characterising disease states for hepatosplenic schistosomiasis (for Schistosoma mansoni), which in its most severe form presents with abdominal distention and vomiting blood. There are no international clinical guidelines for the management of severe schistosomiasis within health facilities despite an estimated 250 million people worldwide with schistosome infections and 700 million people are at risk. This parasitic infection affects individuals over their life course.

The overarching aim is to identify symptom clusters associated with severe schistosomiasis. This project will:

1. investigate different methods for symptom clustering and cluster validation,
2. model the temporal trajectory of individuals and population trends,
3. identify symptom clusters attributable to schistosomiasis while controlling for other causes with shared aetiologies,
4. determine the variation attributable to covariates including sociodemographics, health access, and spatial dependence.

This project will use data from an ongoing prospective cohort study (SchistoTrack). Ethical approvals have been obtained. Detailed symptom data are collected yearly from study nurses in free-text format.

The student will gain skills in text processing, clinical data analysis, complex networks, clustering algorithms (unsupervised and semi-supervised), data cleaning, and research presentation.

FIELD WORK, SECONDMENTS, INDUSTRY PLACEMENTS AND TRAINING

This DPhil project requires one to two months of fieldwork in rural Uganda. Experienced SchistoTrack teams will lead the primary data collection.

PROSPECTIVE  STUDENT

The ideal candidate will have a Master’s degree in statistics, mathematics, machine learning, or a related discipline. This post is particularly suited to someone with a very strong quantitative background with good programming skills in R (preferably) or Python.