Kawasaki disease is an uncommon acute systemic vasculitis, predominantly affecting children younger than 5 years [1]. Its defining features are fever lasting more than 5 days, bilateral nonpurulent conjunctivitis, changes to oropharyngeal mucous membranes, palmar and/or plantar erythema, polymorphous exanthema, and cervical lymphadenopathy [2]. It has been described with varying incidence worldwide [3]. In Australia, the annual Kawasaki disease incidence in children ages 0‐4 years is 14.31/100,000, as measured by the rate of intravenous immunoglobulin treatment [4]. The etiology is poorly understood, with many putative causes proposed, investigated, and excluded [3,5]. However, one or more preceding infective causes are widely suggested as the trigger for the disease and fit with epidemiology [1,3,6,7]. Local outbreaks in two cities in the United States in 1979 and 1980 showed that the only difference between patients and controls was a history of respiratory illness, which may imply concurrent circulation of a viral infectious trigger [8]. Since then, seasonal patterns of Kawasaki disease have been observed in Japan, Korea, and France, with annual peaks occurring in winter and an additional peak in summer in Japan [9-11]. In Victoria, Australia, Kawasaki disease historically peaks in late winter and early spring, which has also been described in studies from the northern hemisphere [4,9]. The co-occurrence of respiratory viral infections in conjunction with Kawasaki disease has been variously reported [12]. A prospective study in Taiwan found that, in addition to a greater prevalence of respiratory symptoms among children with Kawasaki disease, there was a higher rate of polymerase chain reaction (PCR) viral detection in patients with Kawasaki disease compared to controls, including enteroviruses (16.8% vs 4.4%), rhinoviruses (26.5% vs 9.7%), adenoviruses (8% vs 1.8%) and non–SARS-CoV-2 coronaviruses (7.1% vs 0.9%) [13]. Observational studies have reported an association between Kawasaki disease and adenovirus [14], human rhinovirus [15], and non–SARS-CoV-2 coronavirus [16]. An ecologic study in South Korea of over 53,000 Kawasaki disease cases recently identified temporal correlations between rhinovirus, respiratory syncytial virus (RSV), and varicella peaks 1‐3 months preceding Kawasaki disease outbreaks [17]. Reductions in Kawasaki disease presentations in the United States during the COVID-19 pandemic further support the role of respiratory viruses in this disease [18], given that its decline was coincident with reduced respiratory virus circulation throughout the pandemic. Conversely, the appearance of pediatric inflammatory multisystem syndrome temporally associated with SARS-CoV-2 (PIMS-TS) throughout the COVID-19 pandemic has compounded the complexity of a Kawasaki disease diagnosis, as PIMS-TS clinically resembles Kawasaki disease [19-21].

Given the variability of viral associations with Kawasaki disease that have been identified to date and the methodological limitations of some previous studies, we aimed to investigate spatiotemporal associations of Kawasaki disease incidence with respiratory virus circulation in Victoria, Australia over 10 years.

This is a large-scale population-based ecological spatiotemporal analysis of Kawasaki disease in conjunction with respiratory viral circulation over 10 years. We found a novel association between Kawasaki disease and HMPV and (separately) with RSV. The RSV association is similar to the findings of Kang et al [17] using a different temporal ecologic analysis, although our spatiotemporal association was contemporaneous, while their temporal association supported a 1‐ to 3-month lag between outbreaks and Kawasaki disease peaks. Despite being associated with HMPV and RSV, viruses that commonly peak in the winter of each year, there were no independent associations between Kawasaki disease incidence and seasons. This adds credence to the association between Kawasaki disease and these two viruses, suggesting a closely related temporal co-occurrence that cannot be generalized to the month of incidence [9]. The spatial distribution of the data sets is as expected and suggests the relatively equal likelihood of Kawasaki disease presentations across the state of Victoria.

In contrast with our study, Kang et al [17] also identified associations with varicella and rhinovirus. Our laboratory assays identified picornavirus circulation, which includes both rhinovirus and enteroviruses—two viruses that have contrasting seasonal trends and consequently may not have been able to be associated with Kawasaki disease by our model, potentially accounting for the discrepant findings [31]. Varicella was not included in our analysis, and the childhood varicella vaccine has been implemented in Australia since 2005, with rare outbreaks [32].

We did not observe any viral associations with Kawasaki disease incidence throughout the pandemic, which may, in part, have reflected the lockdown restrictions in Victoria from 2020 to 2021 that altered viral circulation in the community [33] and quantifiably reduced infection-related hospitalization in children by 65% [34-36]. The reduction in monthly presentations of Kawasaki disease throughout the pandemic years has been reported internationally, supporting a putative role for transmissible viruses in Kawasaki disease etiology [13,37-41] but not supporting the suggestion that SARS-CoV-2 infections may cause Kawasaki disease [42]. Nevertheless, the fact that there were ongoing presentations of Kawasaki disease despite the limited viral transmission in this period also suggests that Kawasaki disease etiology is more complex [43]. Pitzer et al [7] have suggested that Kawasaki disease onset may be triggered by two concurrent infections; an acute infectious agent (such as a respiratory virus) and an infection with a longer duration (such as a superantigen elaborating colonizing bacteria). A plausible alternative interpretation is that one or more infectious agents that are not included in our multiplex PCR assay and for which transmission was not substantially reduced by the lockdowns are also responsible for Kawasaki disease cases [44,45]. The lack of association between SARS-CoV-2 and Kawasaki disease in our study suggests that Kawasaki disease is not related to PIMS-TS as was initially suggested [19-21]. Our ecologic association findings may inform future investigations of the possible mechanistic roles of respiratory viruses in Kawasaki disease etiology.

We acknowledge some limitations in this study. An important limitation of our study was the skewed geographical distribution of the respiratory PCR test data. These two hospitals had broad coverage of respiratory PCR test data, with virus assay results demonstrated for patients residing in all SA3s in Victoria. However, virus circulation data predominated in the eastern, inner, and inner western suburbs of Melbourne, where these hospital networks are located. MH also provides pathology services for eastern and southern Victoria, including respiratory PCR assays. As such, this introduced potential bias into the analysis. The associations identified in this study may be an over- or underrepresentation of the actual viral associations with Kawasaki disease due to insufficient viral data representation in some parts of the state. An additional limitation was that Kawasaki disease presentation data were obtained from statewide administrative hospital admissions records. Admissions data for Kawasaki disease may overestimate true disease episodes, as was demonstrated in a comparison of emergency department and hospital admission diagnoses in 2015, with clinicians possibly overdiagnosing Kawasaki disease to avoid missing this important diagnosis [46,47]. However, hospitalization data and intravenous immunoglobulin use for Kawasaki disease correlate highly in Australia [4].

An important consideration is that this study used viral infections in Victoria as predictors of Kawasaki disease, irrespective of the host age. We chose this approach because all-population viral infections were used as a surrogate of potential exposure to all circulating microorganisms, in contrast to true infection. We also recognized that restricting the age group of patients with positive respiratory PCR tests would give us a representation of viral infections in children severe enough to lead to hospitalization rather than any virus exposure or mild infection. Notably, PCR data obtained from RCH would represent viral infections in children predominantly (with the occasional adult test conducted on site for parents or guardians). We recognize that this would skew test results toward respiratory viruses that are more common in children due to a sampling bias. Consequently, PCR data used in this analysis may not accurately represent potential exposure to all circulating microorganisms. However, the larger sample size in itself may improve the accuracy of the results, and this has been demonstrated in reviews of nonhealth data [48]. Additionally, data completeness is only one of several dimensions of data quality, and other features including amount of data, consistency, geographic coverage, usability, and maintainability are other important features that these PCR data possess [49].

We have identified potential viral triggers of Kawasaki disease and demonstrated the application of our novel statistical approach. While ecologic studies can only inform association rather than causation, these findings may be used to inform alternate etiological methodologies. We have therefore demonstrated the utility of the Snotwatch ecological data platform to investigate an uncommon disease and to identify novel associations with common viral pathogens. If these findings are validated, with the rapid clinical development of RSV vaccines and the recent revolution in vaccine technologies, we look forward to a time when Kawasaki disease may become (at least partially) a vaccine-preventable disease.

This spatiotemporal analysis of Kawasaki disease identified two potential associations between Kawasaki disease and HMPV and RSV before the COVID-19 pandemic in Victoria, Australia. We used a novel statistical technique to investigate an uncommon disorder at a population level. The additional granularity of the spatial component of our analysis adds accuracy to our findings by exploring associations within each region’s cohort, while the large sample size adds power. Future investigation of respiratory virus associations with Kawasaki disease may confirm that RSV and HMPV have etiological roles in Kawasaki disease and are important targets for intervention.