Return to search

Use of climate in a simple entomological framework to improve dynamic simulation and forecast of malaria transmission

Malaria is a serious and life-threatening mosquito-borne disease that every year affects over 200 million individuals and causes 400,00 deaths. An additional 0.5 billion people globally are at risk of malaria infection. The unique role of climate in influencing malaria transmission outcomes across individual communities by acting on multiple dimensions of the malaria vector and parasite ecology has been long recognized. This recognition has led to the development of explicit and implicit climate-driven models of malaria transmission designed to better understand and predict patterns of population vulnerability and uncover potential challenges to malaria control. However, existing implicitly-forced process-based models of malaria have relied on indirectly correlated predictors of malaria transmission, instead of direct relationships among climate, vector entomology and parasite ecology. The lack of biologically-motivated modulation of malaria transmission compromises meaningful interpretation of the ecological role played by climate in malaria transmission.

Similarly, the specific influence of climate on vector and parasite dynamics is obscured, limiting the utility of these simple and powerful model forms. This dissertation focuses on elaborating the direct ecological relationships between climate, the malaria vector and parasite to enhance the ecological utility of lower dimensional mathematical models of malaria transmission. In the 2nd chapter of this thesis, a climate-driven entomological modeling framework is developed, consisting of a simple dynamic model that explicitly tracks malaria transmission in human populations and implicitly represents the malaria force of infection through climate-regulation of multiple aspects of the Entomological Inoculation Rate (EIR). The EIR-model construct is found to accurately capture seasonal malaria dynamics under free-simulation, when coupled to local rainfall and temperature climatology across multiple local regions in Rwanda. Furthermore, local rainfall modulation of sub-adult survivorship is found to be a more critical driver of seasonal malaria dynamics than other environmentally-regulated components of EIR.

In chapter 3, the model framework is paired with data assimilation methods to dynamically simulate interannual malaria incidence in Rwanda, infer parameters of malaria transmission and validate the malaria model. Results indicate that the implicitly-forced transmission model is able to reproduce interannual and seasonal malaria incidence at regional and local scales. However, accuracy of model description of malaria incidence is more varied at the more resolved local level. Intensified malaria control efforts during the later years of the study are suspected to increase the discrepancy between the vector and parasite dynamics dictated by climate and the observed widespread decline in malaria activity in the region. Nonetheless, the parameters of transmission identified across populations in Rwanda were comparable to existing estimates of malaria, further validating the transmission model and data assimilation approach.

For the 4th chapter, a state-of-the-art Bayesian inference forecasting system for the EIR-model framework is developed, as well as a multi-model forecasting system consisting of weighted-average predictions from the dynamic malaria model and historical expectance predictions. Retrospective forecasts of four years of malaria data from 42 regions in Rwanda indicate that the model-inference forecasting system predicts malaria incidence more accurately than historical expectance alone, particularly for predictions with 1-6 weeks lead times. Although slightly less skillful, the multi-model system was found to substantively enhance forecast reliability of the EIR-model system, bolstering the utility of the malaria model as a robust forecaster of malaria in the region.

The concluding chapter describes areas for improving the specification of the parsimonious model construct. The need to include malaria control coverage data as exogenous forces of transmission, non-climate drivers and alternate sources of climate exposure that support transmission are highlighted. Future works on forecast calibration needed to improve model performance for real-time prediction are also detailed. In addition, areas for application within information systems for evaluating malaria risk and for advising malaria control efforts, specifically relating to local variability in malaria burden and characterization of entomological drivers of local malaria, are identified and further discussed. The model systems developed in this thesis advance the capabilities of lower dimension dynamic models to connect the ecological drivers of malaria transmission to climate variation. Such process-based formulations could provide better climate-driven descriptions of malaria, while limiting model complexity, without compromising representation of entomological relationships that are potentially valuable for improved understanding and control of malaria transmission.

Identiferoai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/d8-x4mv-8n52
Date January 2021
CreatorsUkawuba, Israel Uchenna
Source SetsColumbia University
LanguageEnglish
Detected LanguageEnglish
TypeTheses

Page generated in 0.002 seconds