Regional WHO Office for the Western Pacific. Dengue situation updates 2020.
The COVID-19 pandemic has led to substantial societal disruption in 2020, including changes to human movement behaviours and the closure of specific venues and modes of transport where humans often mixed, through government-imposed public health and social measures.
Dengue transmission in the Asia-Pacific region: impact of climate change and socio-environmental factors.
Declines in human mobility—either voluntarily or through restrictions (ie, public health and social measures)—could reduce dengue virus transmission, but might also disrupt vector control and thus increase dengue virus transmission. The effects of COVID-19-related disruption might also depend on the relative importance of inside versus outside the home for dengue virus transmission.
Some countries might be experiencing the continuance of this global dengue epidemic, while others might be experiencing below average transmission due to the build-up of immunity. Finally, concerns have been raised about under-reporting of dengue case statistics in 2020 given reduced treatment-seeking rates, higher potential for clinical misdiagnosis, and reduced availability of laboratory testing for dengue.
The COVID-19 pandemic should not jeopardize dengue control.
Research in context
Evidence before this study
Previous studies have shown that human movement, heterogeneity in environmental risk, and mosquito control practices all strongly influence the transmission of dengue virus. Restrictions put in place in response to the COVID-19 pandemic led to substantial changes in how people move, where they spend time, and the continuity of disease control programmes, but the net effect on dengue remains unclear. We searched PubMed for studies published between database inception and April 4, 2021, without language restrictions, using the search terms “(COVID-19 OR coronavirus OR SARS-CoV-2) AND (lockdown OR interventions OR restriction OR human mobility) AND (dengue* OR DENV*)”. We also searched WHO Weekly Report and government websites for dengue case data reported for countries in Latin America and southeast Asia. Although 15 studies warned about the risk of COVID-19 exacerbating dengue transmission and the subsequent pressure on intensive care resources, only three studies analysed dengue and COVID-19 data from 2020. Among the three studies that have looked for associations between COVID-19 restrictions and dengue, findings have been mixed—with protective effects, enhancing effects, and no significant effects seen in different countries. Assessing the effect of the COVID-19 pandemic on dengue is challenging due to the high immunity levels against dengue caused by an unusually large global dengue outbreak in 2019 and previously incomplete dengue datasets from 2020.
Added value of this study
To our knowledge, this study is the first to analyse dengue data throughout 2020 from 23 countries spanning the main dengue endemic regions of Latin America and southeast Asia. Our findings show that there is a consistent association between various measures of COVID-19-related disruption and reduced dengue transmission that cannot be explained by seasonal or extra-seasonal dengue cycles or underreporting. Although attributing change to specific restrictions or behaviours was restricted by collinearity, we present evidence that suggests specific roles for schools and other commonly visited non-residential venues.
Implications of all the available evidence
This combined evidence base emphasises the importance of high-traffic, high-mixing venues for dengue transmission and could lead to new interventions and targeting strategies. Although we are unlikely to ever see 2020-like restrictions used to control dengue outbreaks, targeted testing and mosquito control based on patient-reported recent movements could offer new approaches for a disease that continues to evade control by existing approaches.
For dengue, the COVID-19 pandemic provides a unique opportunity to understand how different environments and human movement contribute to transmission and could lead to new interventions and strategies after the public health and social measures are relaxed.
Assessing the impact of COVID-19 border restrictions on dengue transmission in Yunnan Province, China: an observational epidemiological and phylogenetic analysis.
We aimed to conduct the first multi-continent assessment of the effects of public health and social measures on dengue incidence using data from 23 countries, with the goal of quantifying the strength and magnitude of associations between COVID-19-related disruption and dengue virus transmission dynamics.
19 of 23 countries reported lower dengue incidence in 2020 (cumulative Jan–Dec, 2020) than average (vs 2014–19, figure 1A), with exceptions seen in Brazil, Peru, Bolivia, and Singapore. Compared with 2019, incidence decreased by 44·1% across the study area of Latin America and southeast Asia (2·28 million cases in 2020 vs 4·08 million cases in 2019), with a 40·2% decrease in Latin America (569·26 to 340·33 cases per 100 000 population) and 58·4% decrease in southeast Asia (297·31 to 123·58 cases per 100 000 population, figure 1B). This decline becomes even more pronounced when comparing incidence from April 2020 onwards (figure 1C); exceptions include Singapore, which saw above average caseloads throughout 2020, and Ecuador, Brazil, and Peru, which had extra-seasonal increases later in the year. At the time of analysis, we were unable to obtain complete (Jan–Dec, 2020) reported dengue case values for several large dengue-endemic countries, including India, Sri Lanka, Nepal, Myanmar, Paraguay, and Indonesia.
These declines occurred at the beginning of the dengue season in many countries, with cases in southeast Asia, Central America, and the Caribbean typically increasing between June and September. Nine of 11 countries in Central America and the Caribbean, and the Philippines in southeast Asia, saw complete suppression of their 2020 dengue season, with most other countries experiencing a much suppressed dengue season (appendix pp 13–14). In countries where public health and social measures began at the peak of the dengue season, such as in South America, sharper than expected declines were seen despite above average incidence earlier in the year (appendix pp 13–14).
These abnormal declines coincide with the introduction of public health and social measures (late March to early April) and the subsequent shift of human movement behaviours towards time spent in residential premises in late March to April (figure 1D; appendix pp 13–18). The observed climate in 2020 was similar to the average of the previous 6 years, with the exception of mildly higher temperatures in Jamaica and lower temperatures in Venezuela (appendix p 19). We found no evidence that this decline in incidence is due to underreporting. If cases were underreported, we would expect to see higher case fatality rates than reported due to reporting of severe cases being less adversely affected than mild cases. Case fatality rates for 2020 were within the range of the previous 6 years for all countries except Venezuela, which had a mild increase in incidence over its 2017 peak (appendix p 15).
To investigate the association between public health and social measures, human movement behaviours, and dengue incidence, we also considered climatic and immunological factors that can also influence seasonal and extra-seasonal dengue cycles. In the historical model fit to data before the COVID-19 epidemic (2014–19), we retained the climate variables of convective precipitation, surface temperature, and short-term and long-term autocorrelation effects (appendix p 8). This model specification resulted in the largest improvements in both within-sample explanatory and out-of-sample predictive performance, with a decrease in deviance information criterion of 76·30 and cross-validated mean log score of 0·033 over a baseline model of spatiotemporal effects (appendix pp 8, 21–24). The model accurately replicated seasonal dynamics in all countries, explained large outbreak years (eg, 2019) globally and prolonged periods of low transmission (eg, 2017–18 in Central America; appendix pp 29, 31), and estimated approximate seasonal dynamics and comparative outbreak size between countries when making predictions up to a year ahead (appendix p 32).
We used an intervention sub-model to explain the difference between observed case counts in 2020 and predicted case counts in 2020 from the historical model (appendix pp 29, 33). In our univariable intervention model, seven (88%) of eight public health and social measures (except closing public transport) and the composite stringency index showed significant negative correlations with dengue risk. Similarly, three (50%) of six human movement behaviours (except residential, retail or recreation, and park) showed significant positive associations with dengue relative risk (appendix pp 25–26). Although these findings suggest a potential association between one or multiple public health and social measures or human movement behaviours and reduced dengue risk, collinearity prevents us from identifying associations with specific variables when using univariable analyses alone. Therefore, we used a hierarchical cluster analysis to quantify the structural collinearity between public health and social measures and human movement behaviours timeseries. This analysis clustered variables into five distinct clusters, two of which were statistically significant (approximate p>0·05), including the stay at home requirement and closing public transport cluster, and the cluster of all non-residential human movement metrics (figure 2; appendix p 27). For public health and social measures, three variables (cancel public events, school closing, and restrictions on gathering size) were highly colinear (correlation coefficient >0·8; appendix p 28) indicating they were consistently applied at the same time. As expected, the composite stringency index showed high collinearity with all eight of the specific public health and social measures indicators (correlation coefficient range 0·73–0·91; appendix p 28); therefore, we excluded stringency index from subsequent multivariable analysis. All non-residential human movement metrics were highly colinear (absolute correlation coefficient range 0·82–0·94) with change in mobility in grocery or pharmacy showing lower, but still high, collinearity (0·82–0·86). No public health and social measures, nor human movement behaviours, had strong correlation with any environmental variables.
After covariate selection, the human movement behaviours model retained all non-residential variables except change in workplace, whereas the public health and social measures model only retained school closing (figure 2). Consistent with the univariable analysis, these variables were negatively associated with dengue risk (RR range 0·01–0·17), but the magnitude of association varied over different lag periods (Figure 2, Figure 3). School closing was associated with the biggest decrease in dengue risk at short lags (0–1 month) and—to a lesser extent—long lags (3 month; figure 3). To reduce the impact of collinearity, we aggregated human movement behaviour variables (arithmetic mean) into residential and non-residential. Only the non-residential variable was retained by the human movement behaviour model, and a positive association with dengue risk was identified (figure 3). Low values of non-residential movement showed the strongest protective effects at short (0–1 month) and medium to long (2–3 months) time lags.
The selected variables belonged to highly colinear clusters. This means that we cannot accurately rule out an association between dengue risk and restrictions on gathering size or cancelling public events. This analysis does, however, suggest that there is relatively less evidence for an association between dengue risk and stay at home requirements, closure of public transport, restrictions on domestic and international movement, and workplace closures.
By comparing observed and predicted cases (via the historical model) between April and December, 2020, we estimate that 0·72 million (95% CI 0·12–1·47) fewer dengue cases occurred (table), representing a 35% (9–56) decrease that is potentially attributable to COVID-19-related disruption.
TableThe number of cases observed and predicted (April–Dec, 2020) after the implementation of public health and social measures
Data are based on the historical model projected on 2020 environmental and epidemiological conditions.
This reduction was more pronounced in countries in southeast Asia than in Latin America (table). In southeast Asia all countries except Singapore were predicted to have substantial reductions in dengue cases with the largest reductions seen in the Philippines and Cambodia. In Latin America, most (nine of 16) countries had fewer cases than expected; however, Belize, Bolivia, Brazil, Costa Rica, Honduras, Nicaragua, and Peru experienced more cases than anticipated. Brazil, in particular, remains a major outlier that negatively skews regional and global estimates of the percentage of averted cases, accounting for 51% of all observed dengue cases between April and December, 2020. This discrepancy between expected and observed cases in 2020 in Brazil might be related to the less stringent public health and social measures, variable adherence,
Dataset on SARS-CoV-2 non-pharmaceutical interventions in Brazilian municipalities.
and more modest changes in human movement behaviours that occurred in the country in 2020 (appendix pp 17–18).
We then tested what proportion of the difference between expected (historical model) and observed case counts in April to December, 2020 could be explained by the specific public health and social measures and human movement behaviour variables in our analysis (appendix p 10). School closures in the public health and social measures model explained 70·95% (95% CI 55·55–80·48) of the reduction, whereas reductions in movement in non-residential locations in the human movement behaviour model explained 30·95% (15·57–43·65; appendix p 30). Even in countries with low or negative estimates of averted cases (table), such as Brazil, variation in monthly case counts could be explained by the public health and social measures models (figure 4), suggesting these countries would have experienced lower dengue case counts if public health and social measures had been more stringent or declines in non-residential movement been more substantial.
By combining the most globally comprehensive collection of dengue and COVID-19 response data, we show that the sudden decline in dengue cases in April, 2020 is associated with the imposition of restrictions and changes in human movement behaviours. We show that school closures and declines in non-residential trips have the strongest association with reduced dengue risk. Combined, we estimate that 0·72 million (95% CI 0·12–1·47) fewer dengue cases occurred in 2020 than would have occurred in the absence of COVID-19-related disruption.
It remains to be seen how many of these 0·72 million cases are truly averted or just delayed until later years as pre-COVID-19 human movement behaviours re-establish. By using distributed lag non-linear models, we were able to show that public health and social measures and human movement behaviours confer both short-term (0–1 month) and medium to long-term (2–3 months) protective effects. Continued observation and re-analysis will be needed to assess longer term effects. Disruption to routine vector control (eg, household larval inspections and community clear-up campaigns) could suffer long-term effects that are not observable until the next dengue epidemic.
WHO Informing vaccination programs: a guide to the design and conduct of dengue serosurveys.
and a better understanding of how treatment-seeking behaviour changes at different phases of dengue and COVID-19 epidemics (as access to care and rapid diagnostics changes) will be important to interpret changes in reported caseloads. Continued monitoring of dengue trends in 2021 and beyond will be key, including the continued collection of human movement data, better data on adherence to public health and social measures,
Dataset on SARS-CoV-2 non-pharmaceutical interventions in Brazilian municipalities.
and the use of disease forecasting systems to detect and respond to dengue epidemics when they do occur.
Theoretically, COVID-19-related disruption could increase or decrease dengue transmission through mechanisms such as mosquito control disruption, reduced human movement restricting geographical spread, and reduced time spent in high risk non-residential environments.
These hypothetical changes in risk would probably act over different timescales, with reducing time in high-risk environments leading to the most immediate reductions, whereas restricting spread and disruption to mosquito control could take 1–3 months to have substantial effects. This mixture of effects might explain why we estimate varying levels of protection at different lags.
Although we caution against overinterpretation of the selection, magnitude, lag, or direction of specific variables in our analysis, some consistent trends could guide further studies. Reductions in non-residential movement and closure of schools had the strongest evidence of an association with reduced dengue risk among the variables analysed. Understanding where dengue transmission occurs in different settings (eg, home, workplace, or school) remains a major knowledge gap. Targeting mosquito control measures to households of individuals with dengue has long been recommended and practiced
under the assumption that mosquito exposure within, or in close proximity to, the home drives transmission. Despite this, household cluster studies rarely identify strong clustering of transmission around houses,
Households as foci for dengue transmission in highly urban Vietnam.
and a competing theory has emerged that transmission occurs in shared spaces away from the home or is driven by the movement of people that allows the virus to expand into new pockets of human susceptibility.
Disease-driven reduction in human mobility influences human-mosquito contacts and dengue transmission dynamics.
By showing that dengue risk is more closely associated with reduced time spent in public areas, we add evidence to this theory. Our findings imply that schools and other commonly visited public areas (or travel between home and these places) could be dengue transmission hotspots. These findings are consistent with the apparent concentration of symptomatic cases in children younger than 15 years and the main vector of dengue, Aedes aegypti, preferentially biting during the day. If supported by further outbreak investigation studies, this finding would suggest a greater emphasis is needed on dengue control in public places, and in schools in particular. These findings might also be of relevance to the dynamics of other arboviral diseases (eg, Zika and chikungunya) and to infectious diseases more generally, and serves as an example that COVID-19-related disruption does not always result in adverse effects.
Our findings have several limitations. First, owing to data availability, we did not include information on dengue serotypes or genotypes, which are well known drivers of dengue outbreaks.
Dengue virus surveillance in Singapore reveals high viral diversity through multiple introductions and in situ evolution.
Such switches might explain outliers, such as Singapore, where a sustained switch in the predominant serotype from DENV-1 or DENV-2 to DENV-3 could have led to the observed increases in incidence in 2020.
Pandemic-associated mobility restrictions could cause increases in dengue virus transmission.
Third, we were unable to control for potential changes in dengue reporting that might have occurred due to COVID-19 disruption. By showing that case fatality rates in 2020 were not abnormal, we provide evidence against the theory that reduced dengue incidence in 2020 is due solely to underreporting. Some countries, such as Sri Lanka, have also reported undertaking additional community outreach activities for dengue during the COVID-19 pandemic to restrict the effects of any reduced treatment-seeking behaviours.
The impact of COVID-19 lockdown on dengue transmission in Sri Lanka; A natural experiment for understanding the influence of human mobility.
Effects on case presentation, diagnosis, and reporting are likely to be complex, country-specific, and delayed. Additionally, if dengue cases were substantially underreported then we would expect a rapid rise in reported cases as COVID-19 restrictions are lifted, as opposed to a more gradual rise due to resurgence of dengue transmission. Despite countries in Asia relaxing domestic COVID-19 restrictions in late 2020, we did not observe rapid rises in reported dengue cases. A more detailed temporal analysis of fatal and non-fatal cases of multiple acute conditions would give more insight into how disease surveillance has changed during the COVID-19 pandemic.
Fourth, we were not able to include all countries seriously affected by dengue in our analysis because publicly available monthly case reports could not be found for some countries. Indonesia reports the highest number of dengue cases in southeast Asia and, with equatorial seasonality, would have improved our historical and intervention models. However, the reported annual case decline in Indonesia (138 127 in 2019 and 108 303 in 2020)
Indonesia Ministry of Health Dengue data January–December 2020: surveillance of Ministry of Health Indonesia.
is within the range of other countries in the region and is unlikely to change our main findings.
Fifth, in this study we use penalised complexity priors for estimating parameters using integrated nested laplace approximation. Penalised complexity priors are well suited for penalising more complex models with multiple variables; however, model fit and predictions of the number of averted cases might differ with different prior specifications.
Lastly, our analysis was restricted to national-scale dengue and movement dynamics. There is probably substantial sub-national heterogeneity in the size and strength of association between movement restrictions and dengue risk that will be important to quantify. One priority for research is measuring how this association varies between urban and rural areas, with urban areas typically having much higher baseline movement than rural areas.
In summary, this study is the most geographically comprehensive study to date to show that the substantial reduction in dengue cases seen in 2020 is potentially attributable to COVID-19-related disruption. Although it remains unknown what effect these restrictions will have on dengue dynamics in the long term, the unique circumstances of the COVID-19 pandemic might give new insights into the development and targeting of new and existing interventions for dengue.
HT and OJB designed the study. YC and NL collected the data, accessed and verified the data, and did the statistical analyses. YC, NL, JL, YL, MJ, NIB, SA, HT, and OJB designed the statistical analysis. YC, NL, HT, and OJB wrote the first draft of the manuscript. LW, BC, LD, BL, RV, and AW-S provided important comments on the draft manuscript. All authors read and approved the manuscript. The corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.