Modelling approaches for disease biogeography : investigating the spatio-temporal relationships of infectius disease systems
(2022)
Files
DaRe_PhThesis_publicDefence.pdf
Open Access - Adobe PDF
- 26.65 MB
Details
- Authors
- Supervisors
- Abstract
- Understanding the causal processes determining the geographical distribution of species is a fundamental question in ecology but has relevant implications in epidemiology. Infectious diseases are a public health concern for humans, livestock, and wildlife, and their relevance has fostered the interest in tools allowing the delineation of areas at risk for pathogen transmission. In the past two decades, the prompt availability of new spatio-temporal explicit datasets and coding environments has led to extensive use of modelling tools to infer the geographical distribution of the species involved in infectious disease systems. However, the validity of these models was questioned, underlining the lack of biological realism and causal-based reasoning as the main limitations. In this doctoral dissertation, I tried to include biological realism and a causal-based perspective on correlative and mechanistic modelling approaches aiming to infer the spatio-temporal distribution of vector and host species involved in vector-borne disease systems. I first applied the modelling relation framework in a species distribution modelling exercise through the Structural Equation Modelling approach, a methodology that includes and evaluates causal pathways within a linear modelling framework. I moved towards mechanistic models and built dynamAedes, a spatially-explicit model inferring the population dynamic of four Aedes mosquito species at different spatial scales. I then explored how the choice of the model parameters and spatial scales affect the outcomes of an epidemiological model estimating the number of Chikungunya’s secondary cases. Finally, since host abundance is an epidemiological parameter as substantial as vector abundance, I presented a downscaling methodology to disaggregate livestock censuses aggregated at different administrative unit levels. The results highlighted how a causal-based approach increases the biological realism and predictive accuracy of the modelling approaches tested. However, the knowledge of essential biological parameters is scattered, fragmented and not standardized, affecting the models’ outcome quality and reliability. The choice of the spatial scale affects as well the models' outputs, as coarser training and testing datasets produce, on average, better results because of the effects of the Modifiable Areal Unit Problem. To amend such limitations and promote the effective use of spatial-explicit model outputs for public health decisions, clear communication and dialogue with policymakers are essential to enable them to understand the assumptions of the models and the uncertainty of their predictions.
- Affiliations
Citations
Da Re, D. (2022). Modelling approaches for disease biogeography : investigating the spatio-temporal relationships of infectius disease systems. https://hdl.handle.net/2078.5/103441