Social network analysis and exponential random graph modelling of dengue spatiotemporal linkage networks in an urban Malaysian district

Abstract

Dengue burden is increasingly concentrated in rapidly urbanising Asian cities, yet routine surveillance rarely captures fine-scale residential spatiotemporal co-occurrence. We characterised a case-level linkage network and identified sociodemographic and residential correlates of linkage formation in Petaling District, Malaysia, using social network analysis and exponential random graph models (ERGMs). Cases were linked when their residences were within 200 m and symptom onset occurred within 14 days. Among 19,549 laboratory-confirmed dengue cases reported in 2023, 12,978 had at least one linkage, forming 21,424 linkages in a highly fragmented network. Linkages represent residential co-occurrence rather than confirmed transmission. Among connected cases, the conditional odds of linkage were elevated for pairs of high-rise residents (OR 6.66, 95% CI 6.40–6.93), low-rise residents (OR 4.34, 95% CI 4.15–4.54) and non-citizen cases (OR 3.96, 95% CI 3.70–4.22). Student status was modestly associated with linkage formation (OR 1.07, 95% CI 1.05–1.10). Degree-related structural dependence was not supported. Attribute associations were more stable than structural dependence across evaluated network definitions. Integrating linkage analysis into routine surveillance could support targeted prioritisation of vector control in high-rise buildings and priority communities when operational capacity is constrained.