Exploring the Association between Remotely Sensed Environmental Parameters and Surveillance Disease Data: An Application to the Spatiotemporal Modelling of Schistosomiasis in Ghana
Wrable, Madeline.
2017
-
Abstract:
Schistosomiasis control in sub-Saharan Africa is enacted primarily through mass drug
administration, where predictive modeling plays an important role in filling knowledge
gaps in the distribution of disease burden. Remote sensing (RS) satellite imagery is
used to predictively model infectious disease transmission in schistosomiasis, since
transmission requires environmental conditions ... read moreto sustain specific freshwater snail
species. Surveys are commonly used to obtain health outcome data, and while they provide
accurate estimates of disease in a specific time and place, the resources required make
performing surveys at large spatiotemporal scales impractical. Ongoing national
surveillance data in the form of reported counts from health centers is conceptually
better suited to utilizing the full spatiotemporal capabilities of publically available
RS data, as most open source satellite products can be utilized as global continuous
surfaces with historical (in some cases 40-year) timespans. In addition RS data is often
in the public domain and takes at most a few days to order. Therefore, the use of
surveillance data as an initial descriptive approach of mapping areas of high disease
prevalence (often with large focal variation present) could then be followed up with
more resource intensive methods such as health surveys paired with commercial, high
spatial resolution imagery. Utilization of datasets and technologies more cost
effectively would lead to sustainable control, a precursor to eradication (Rollinson et
al. 2013). In this study, environmental parameters were chosen for their historical use
as proxies for climate. They were used as predictors and as inputs to a novel climate
classification technique. This allowed for qualitative and quantitative analysis of
broad climatic trends, and were regressed on 8 years of Ghanaian national surveillance
health data. Mixed effect modeling was used to assess the relationship between reported
disease counts and remote sensing data over space and time. A downward trend was
observed in the reported disease rates (~1% per month). Seasonality was present, with
two peaks (March and September) in the north of the country, a single peak (July) in the
middle of the country, and lows consistently observed in December/January. Trend and
seasonal patterns of the environmental variables and their associations with reported
incidence varied across the defined climate zones. Environmental predictors explained
little of the variance and did not improve model fit significantly, unlike district
level effects which explained most of the variance. Use of climate zones showed
potential and should be explored further. Overall, surveillance of neglected tropical
diseases in low-income countries often suffers from incomplete records or missing
observations. However, with systematic improvements, these data could potentially offer
opportunities to more comprehensively analyze disease patterns by combining wide
geographic coverage and varying levels of spatial and temporal aggregation. The approach
can serve as a decision support tool and offers the potential for use with other
climate-sensitive diseases in low-income
settings.
Thesis (M.S.)--Tufts University, 2017.
Submitted to the Dept. of Civil Engineering.
Advisor: David Gute.
Committee: Elena Naumova, Karen Kosinski, and Magaly Koch.
Keyword: Environmental health.read less - ID:
- 1g05fp87x
- Component ID:
- tufts:22469
- To Cite:
- TARC Citation Guide EndNote