Diarrhea is the second leading cause of death in children under-5; it kills more children than HIV/AIDS, measles, and malaria combined. Despite this significant health burden, our ability to anticipate and prepare for diarrhea outbreaks remains limited. Precipitation and temperature variability have been shown to affect diarrhea dynamics and therefore contribute to outbreak predictions, but the observed environment-diarrhea relationships are complex and context-specific, depending on local pathogen distribution, host population behavior, and physical environments. To date, studies in sub-Saharan Africa, where the burden of under-5 diarrhea is particularly high, are limited due to sparse diarrheal disease surveillance data. In this dissertation, we leverage unique under-5 diarrhea incidence data to explore the effects of meteorological variability on childhood diarrhea incidence and develop a real-time forecasting system for diarrheal disease in Botswana, where diarrhea remains an important cause of childhood morbidity and mortality. The study focuses in Chobe District, which has an annual dry (April – September) and wet (October – March) season, during which the Chobe River, the primary source of drinking water in the region, floods. Weekly cases of under-5 diarrhea in Chobe District exhibit strong seasonal dynamics with biannual outbreaks occurring during the wet and the dry season. In Chapter 1, we show that wet season diarrhea incidence is strongly associated with increased rainfall and Escherichia coli concentrations in the Chobe River, while dry season incidence is associated with declines in Chobe River flood height and increased total suspended solids in the river. In Chapter 2, we confirm the existence of an El Niño-Southern Oscillation teleconnection with southern Africa by demonstrating that La Niña conditions are associated with cooler temperatures, increased rainfall, and higher flooding in Chobe District during the wet season. In turn, we show that La Niña conditions lagged 0-5 months are associated with higher than average incidence of under-5 diarrhea in the early wet season (December – February). In Chapter 4, we develop and test an epidemiological forecast model for childhood diarrheal disease in Chobe District. The prediction system uses a compartmental susceptible-infected-recovered-susceptible (SIRS) model coupled with Bayesian data assimilation to infer relevant epidemiological parameter values and generate retrospective forecasts. The model system accurately forecasts diarrhea outbreaks up to six weeks before the predicted peak of the outbreak, and prediction accuracy increases over the progression of the outbreak. Many forecasts generated by the model system are more accurate than predictions made using only historical data trends. This dissertation work is an important step forward in our understanding of the links between proximal and distal climatic variability and childhood diarrhea in arid regions of sub-Saharan Africa. Furthermore, it advances methods for generating accurate long-term and short-term forecasts of under-5 diarrhea. We demonstrates the potential use of ENSO data, which are publicly available, to prepare for and mitigate diarrheal disease outbreaks in a low-resource setting up to 5 months in advance, and develop a model-inference system that can generate accurate predictions during an outbreak. Deaths caused by diarrhea are preventable using low-cost treatments. Hence, accurate predictions of diarrhea outbreak magnitudes could help healthcare providers and public health officials prepare for and mitigate the significant morbidity and mortality resulting from diarrhea outbreaks.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/d8-78dv-d120 |
| Date | January 2019 |
| Creators | Heaney, Alexandra Krosnick |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.0021 seconds