Exploring sociodemographic drivers of the COVID-19 pandemic in Belgium: Insights from hierarchical and fractal-based spatiotemporal models

Abstract

The COVID-19 pandemic profoundly reshaped multiple aspects of daily life, from societal norms to research priorities and public health perspectives. In its aftermath, the urgency of preparing for and responding to future large-scale health crises has become a central concern for many countries, including Belgium. This underscores the importance of critically evaluating and refining methodological approaches to ensure accurate estimation, nuanced interpretation, and actionable insights that can guide policy makers and the broader public health community. In this doctoral thesis, the use of hierarchical models (specifically Bayesian spatial models implemented through INLA and linear mixed models) as a framework for analyzing spatiotemporal COVID-19 data was explored. These models revealed important demographic and contextual drivers of the pandemic, including age, sex, income, population density, student population size, and human mobility. Additional influencing factors, such as vaccination coverage and the stringency of government interventions, were also highlighted. On top of this, fractal dimension analysis was introduced as a complementary tool to characterize epidemic dynamics. Unlike hierarchical models, fractal-based approaches focus on quantifying the complexity of time-series data without imposing strong assumptions about the underlying spatial mechanisms. By summarizing epidemic dynamics through indicators such as mean, variance, and autocorrelation of local fractal dimension curves, this method enables rapid characterization of epidemic patterns at very fine administrative scales. This is particularly valuable when data are noisy, fragmented, or rapidly evolving, conditions frequently encountered during emerging outbreaks. The proposed fractal-based clustering framework further demonstrates the potential of this approach to identify vulnerable areas and provide actionable insights for targeted interventions. Together, the hierarchical modeling framework and the fractal dimension approach illustrate the value of combining established and innovative methodologies. Hierarchical models excel at capturing structured relationships and quantifying the effects of covariates, while fractal analysis offers an efficient way to summarize and compare epidemic complexity across space and time. Integrating these approaches can enrich epidemiological analysis, improve outbreak surveillance, and ultimately strengthen preparedness for future health crises.