Human interaction significantly contributes to the spread of infectious diseases like COVID-19, making it essential to consider human activities when analyzing transmission patterns. Typically, data on infectious disease transmission is aggregated across various spatial and temporal scales due to data accessibility limitations and privacy concerns. This study investigates the relationship between COVID-19 weekly incidence, human activity index, and inter-regional mobility flows at fine spatial scale. We modeled weekly COVID-19 incidence for Middle Layer Super Output Areas in the Greater London area from March to December 2020 using Bayesian hierarchical spatiotemporal models. We constructed two types of indicators: mobility indicators by weighting travel-to-work flows with destination activity levels, and spatial indicators by weighting activities with geographic adjacency. The models accounted for area-specific spatial random effects, temporal effects, and district-level spatiotemporal effects. Both activity-based models showed modest improvements over baseline, with spatial random effects proving essential. Models using geographic adjacency performed nearly identically to those using travel-to-work estimates. Temporal stratification revealed substantial variation, with mobility-disease coefficients shifting from positive during moderate transmission to negative during explosive growth at year’s end. We presented a model to jointly study human mobility, human activity, and COVID-19 cases, providing new insights into their combined effects. Simple geographic proximity captures transmission patterns as effectively as travel-to-work data when robust spatiotemporal modeling is employed. The temporal variation in mobility-disease relationships underscores the importance of phase-stratified analyses during evolving epidemics. The models successfully reproduced spatial transmission patterns, validating their utility for surveillance applications at small areal units.
Niraula et al. (Thu,) studied this question.