Comparing the spatial factors responsible for multiple waves of COVID-19 in Bangladesh: A GIS-based approach

AbstractBackground Since 2019, the world has been going through multiple waves of COVID-19 and many new epidemics have been inducing with time. The underlying reasons behind each wave are inconsistent, and the distribution of resources to tackle the pandemics is also not identical. To understand the variation of contributors and develop future pandemic management policies, determining the associated factors of each wave is necessary. Therefore, this study compares the spatial factors influencing five COVID-19 pandemic waves in Bangladesh. Methods Data on the number of confirmed cases and tests done in 64 districts of Bangladesh were used to define the COVID-19 incident rate. The data on 27 spatial attributes encompassing demographic, infrastructural, healthcare facilities, meteorological, and economic factors were accumulated from various secondary sources to explore their influences in each wave. Global and local Moran's I autocorrelation techniques (i.e., GMI and LISA) were employed to know the spread pattern. Both univariate and multivariate approaches were applied to screen the spatial factors. Later, three global regression models named Ordinary Least Square Regression (OLS), Spatial Lag Model (SLM) and Spatial Error Model (SEM), and Geographically Weighted Regression (GWR) as a local model were accommodated to develop the relationship among spatial determinants and incident rate across the waves. The relationship between incident rates and attributes was explained and justified by using the cluster analysis approach. Results The findings from autocorrelations showed that the first and third waves experienced multiple dominant clusters of incident rate, whereas most infections happened randomly in the second, fourth, and fifth waves. While exploring the contributory attributes, the SLM model performed better in the scenario of clustered spatial patterns, and all three models depicted almost similar results during the random distribution of incident rate. GWR model could not significantly improve the developed local models yet help to explain the changes in relationships among spatial attributes and pandemic spread. Despite shifting the ancillary factors and their performances that happened over the waves, urban population density, precipitation, mismanagement in healthcare facilities, concentration of local gathering places, and local economic conditions served as the most dominant influencing factors. Different pharmaceutical (i.e., vaccination) and non-pharmaceutical measures (i.e., containment strategies) somewhat controlled wave-wise variation of responsible determinants. Moreover, some countrywide applied preventive strategies and the evolution of new virus strains could also affect the wave scenario along with spatial factors as resulted by the explanation parameters of the developed models. Conclusion The findings of this paper would provide insight into how spatial policies could be designed to tackle future epidemics and pandemics. Controlled urban population growth and assurance of cautious management of healthcare facilities are crucial to mitigate the adverse impacts of any pandemic. Local economic, demographic and climatic conditions have to be addressed in framing the preventive measures and resource allocation of pandemic waves.