The epidemic forest reveals the spatial pattern of the spread of acute respiratory infections in Jakarta, Indonesia

Abstract

Abstract Acute respiratory infection (ARI) is a communicable disease of the respiratory tract that implies impaired breathing. The infection can expand from one to the neighboring areas at a region-scale level through a human mobility network. Specific to this study, we leverage a record of ARI incidences in four periods of outbreaks for 42 regions in Jakarta to study its spatio-temporal spread using the concept of the epidemic forest. This framework generates a forest-like graph representing an explicit spread of disease that takes the onset time, spatio-temporal distance, and case prevalence into account. To support this framework, we use logistic curves to infer the onset time of the outbreak for each region. The result shows that regions with earlier onset dates tend to have a higher burden of cases, leading to the idea that the culprits of the disease spread are those with a high load of cases. To justify this, we generate the epidemic forest for the four periods of ARI outbreaks and identify the implied dominant trees (that with the most children cases). We find that the primary infected city of the dominant tree has a relatively higher burden of cases than other trees. In addition, we can investigate the timely ( $$R_t$$ R t ) and spatial reproduction number ( $$R_c$$ R c ) by directly evaluating them from the inferred graphs. We find that $$R_t$$ R t for dominant trees are significantly higher than non-dominant trees across all periods, with regions in western Jakarta tend to have higher values of $$R_c$$ R c . Lastly, we provide simulated-implied graphs by suppressing 50% load of cases of the primary infected city in the dominant tree that results in a reduced $$R_c$$ R c , suggesting a potential target of intervention to depress the overall ARI spread.

Keywords

• ARI
• Epidemic forest
• Early detection