Hybrid modelling methods for epidemiological studies
Amadi, Miracle (2022-06-23)
Väitöskirja
Amadi, Miracle
23.06.2022
Lappeenranta-Lahti University of Technology LUT
Acta Universitatis Lappeenrantaensis
School of Engineering Science
School of Engineering Science, Laskennallinen tekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:ISBN:978-952-335-833-1
https://urn.fi/URN:ISBN:978-952-335-833-1
Tiivistelmä
Epidemiological modelling plays an important role in the study of the distribution of a disease and its impact in a given population, and helps to suggest effective control and prevention measures. The complexity of models as well as the modelling approach vary depending on a number of factors such as how much is understood about the disease epidemiology, the objective of the study and the nature of the data available. Compartmental models have shown to capture the macro level dynamics of an infectious disease outbreak and have been utilised to develop control policies and outbreak responses. However, they contain a limited account of the complex processes of dynamics of most infectious diseases. Unlike the continuous modelling framework, the Agent-based modelling (ABM) approach features the simulation of heterogeneous communities subjected to more realistic transmission scenarios and can incorporate complex and stochastic issues affecting diseases. This work provides an example of how to utilise the strength of both kinds of models through a hybrid approach that combines in situ field data with the parameters of a classical malaria model. The ABM simulations provide a computational laboratory for generating data on the impact of some complex factors on malaria prevalence. The ABM results can be extended to continuous time by inserting the values fitted by the classical response surface regression as the key coefficients of compartmental models. Another regression approach presented in this work is a cluster-integrated regression which helps to screen the incidence clusters where the available explanatory variables fail to predict, using a panel data. The cluster-integrated regression method also improves the accuracy of the model by providing more explanatory variables. In addition, the spatial autocorrelation study using global Moran’s I, the Geary’s C and Moran’s scatter plot was made to measure the timely spatial pattern of disease incidence in a country and to form the grouping. This was combined with a proposed metapopulation model that parameterizes and reassesses non-pharmaceutical interventions. The uncertainty quantification of model outputs using Markov Chain Monte Carlo (MCMC) techniques was done based on the notion of randomness in the modelling approach.
Kokoelmat
- Väitöskirjat [1037]