key: cord-0850376-fl4jornf authors: Ong, Song-Quan; Ahmad, Hamdan; Ngesom, Ahmad Mohiddin Mohd title: Implications of the COVID-19 lockdown on dengue transmission in Malaysia date: 2020-09-18 journal: bioRxiv DOI: 10.1101/2020.07.21.214056 sha: 8e4b35126f3065dba4d7cceb65fb2697d6f44908 doc_id: 850376 cord_uid: fl4jornf The impact of movement restrictions during the COVID-19 lockdown on the existing endemic infectious disease dengue fever has generated considerable research interest. We compared the Malaysia weekly epidemiological records of dengue incidences during the period of lockdown to the trend of previous years (2015 to 2019) and a simulation at the corresponding period that expected no movement restrictions. We found that the dengue incidence declined significantly with a greater magnitude at phase 1 of lockdown, with a negative gradient of 3.2-fold steeper than the trend observed in previous years and 6.5-fold steeper than the simulation, indicating that the control of population movement did reduce dengue transmission. However, starting from phase 2 of lockdown, the dengue incidences demonstrated an elevation and earlier rebound by at least 4 weeks and grew with an exponential pattern compared to the simulation and previous years. Together with our data on Aedes mosquitoes from a district of Penang, Malaysia, we revealed that Aedes albopictus is the predominant species for both indoor and outdoor environments. The abundance of the mosquito was increasing steadily during the period of lockdown, and demonstrated strong correlation with the locally reported dengue incidences; therefore, we proposed the possible diffusive effect of vector that led to a higher acceleration of incidence rate. These findings would help authorities review the direction and efforts of the vector control strategy. Lockdown has been commonly implemented all over the world to halt COVID-19 transmission [1] . This is to control human movement, as physical proximity is a key risk factor for the transmission of SAR-CoV-2 [2] . Although the physical proximity of humans shapes the spatial spread of a pathogen, it refers mostly to directly transmitted infectious diseases [3] such as COVID-19, but the effect on indirectly transmitted infectious diseases such as dengue fever (DF) remains unclear. Dengue fever (DF) is the most prevalent mosquito-borne disease in the world [4] , understanding the impacts of human (host) movement on the transmission can improve the prevention system. Falcón-Lezama et al. [5] used a mathematical model to evaluate the effect of people's day-to-day movement on the dengue epidemic and concluded that the vector-host's spatial connectivity posted epidemic risk. To simulate the actual situation, Stoddard et al. [6] used contact-site cluster investigations in a case-control design to review the risk of dengue infection by human movement and argued the importance of movement restriction in managing the spatiotemporal dynamics of dengue virus. However, the previous experimental configurations were far from the real situation, especially when the mobility of the population on a large scale is not feasible to demonstrate the direct effect on dengue transmission. With the imposition of the COVID-19 partial lockdown in Malaysia on 13 th to 20 th weeks of 2020 (18 March to 12 May 2020), about 90% of people were restricted to their homes, and 10% essential workers were allowed to carry out their daily activities for the whole country [7] ; thus, the unprecedented large-scale movement restriction of host provides opportunities to evaluate its direct impact on the dengue transmission. To understand the direct impact of the large-scale movement restriction on Malaysia dengue endemic, comparison between the actual incidences trend with a simulation trend that expect no interference could be one of the approach as demonstrated by other studies on infection diseases [8] [9] [10] . To construct a best-fitting simulation, we have to understand the dengue endemic pattern in Malaysia, in which the dengue endemic in Malaysia is seasonal, with variable transmission and prevalence patterns affected by the large diversity in rainfall and spatial variation [11] . The major transmission periods of DF occur from June to September, following the main rainy seasons; the minor transmission period is from September to March, following monsoon rains sessions that bring higher precipitation and lead to the greater potential breeding ground for the vector [12] . The end of the minor peak and the start of major peaks of DF transmission often coincides with the duration of the COVID-19 lockdown ( Figure 1 ). We postulate the trend of dengue incidences could be altered due to the changes of host movements, and seasonal time series analysis such as SARIMA is a suitable approach to construct the simulation as the method usually used to predict or forecast the infection incidences that were seasonal [8] [9] [10] . One of the key element in dengue transmission is the abundance and distribution of the vectors -Aedes aegypti and Aedes albopictus, in which the spatial distribution is also potentially affected by movement restrictions of humans. By using an agent-based transmission model, Reiner et al. [13] indicated that the socially structured movement of humans caused a significant influence on dengue transmission, and the infection dynamics were hidden by the diffusive effect of the vectors. The implementation of lockdown in Malaysia changed the host's daily activities, most of the time people were contained in their own housing area, and also to both Ae. aegypti and Ae. albopictus that are highly anthropophilic, in which Ae. aegypti almost exclusively rely on human blood and Ae. albopictus is an aggressive and highly adaptive species that can easily colonize the habitat of other mosquitoes in urban areas [14] . In addition, the spatial distribution of the host also influenced the behavior of the vectors, and previous studies [14] [15] [16] [17] [18] identified a shift of the Ae. albopictus habitat to an indoor environment where the species usually inhabit the forest or are mostly vegetative and cause interspecies competition with other existing mosquito species, such as Ae. aegypti. Therefore, when the COVID-19 partial lockdown restricts humans in mostly indoor environments with minimum outdoor activities, we are interested in the distribution of the mosquitoes as well. We study the effect of the physical proximity restriction on humans during the COVID-19 lockdown in Malaysia on two variables as shown in the paradigm in Fig. 2 . Country dengue incidences trend. This study aims to understand the impact of movement restrictions on the dengue transmission by statistically comparing the trend of actual reported dengue incidences with a SARIMA model simulation and previous years of incidences. We establish the SARIMA model simulation that expects no MR practice from the dataset of Malaysia weekly reported dengue incidences of previous years (2015 to 2019). We evaluate the level of heterogeneity of actual trend from simulation and previous years according to the phases of lockdown that implemented by Malaysia Government. To rationale the changes of country dengue incidences trend, we aim to obtain data on the abundance and distribution of Aedes mosquitoes during the period of lockdown by conducting sampling in the indoor and outdoor environments of areas from a district of Penang Malaysia. We started the analysis from the data of 2015, which recorded the second-highest dengue incidence after 2019 [19] . We retrieved the data from the official press statement and the We observed the previous temporal pattern of dengue incidences and found that the period of the COVID-19 partial lockdown [13 th week (March 18) to 24 th week (June 9) of 2020] coincided between the end of minor (March-May) and the start of major (June-Sept) fluctuations of dengue transmission (Fig. 1) . Therefore, to understand the heterogeneity of the trend of Malaysia weekly incidences due to the partial lockdown, we compared the actual trend with three reference trends, namely, a simulation, mean weekly incidence of year 2015-2019, and the years that recorded the highest incidences trend (2019). The simulation was assumed to be without the interference of city lockdown and population movement control. To construct a simulation, we applied seasonal autoregressive integrated moving average (SARIMA) models, which are advantageous for modeling the seasonal and time-based dependent configuration of a time series and are commonly applied for epidemiological surveillance [8] [9] [10] 21] . We trained the models by using Malaysia weekly dengue incidences from 1 st week of 2015 to 52 nd week of 2019 from the dataset, and forecasted the incidences on 1 st to 35 th weeks of 2020, including the phases of partial lockdown in Malaysia (Table 1) . We constructed and selected the best SARIMA model (p,d,q) x (P,D,Q) [p is the autoregressive lags, d is the degree of differencing, q is the moving-average lags, P is the seasonal autoregressive lags, D is the seasonal degree of differencing and Q is the seasonal (Table 1 ) and compared the actual weekly dengue incidences during the period of partial lockdown with the those of the simulation and previous years (2015-2019) using two-way ANOVA with two independent variables, namely, trends (actual/simulation/previous years of incidences), and phases. Because the trends follow an open-up parabolic pattern, we further distinguish the pattern by comparing the slopes for the phases (based on incidences in three weeks) to study the rates of decline and endemic incline. We assessed the abundance and distribution of Aedes mosquitoes during the movement restriction of human in particular the indoor and outdoor environments due to the possibility of a decrease in artificial breeding sites reported at the beginning of lockdown [24] and the host contained in their housing area. The sites were selected from three residential areas, namely, Taman Bukit Jambul (5°20'06.6"N 100°17'18.7"E), Taman Permai Indah Most of the studies on the growth of dengue incidences in Malaysia have focused on the total cases reported annually, in which the proposed model may not be sensitive and flexible enough to predict growth and propose necessary management. To our knowledge, this study is the first to report a simulation model with weekly dengue incidences in 2020, including the period of COVID-19 partial lockdown in Malaysia. To select the best-fitting simulation model, Table 2 shows the comparison of the NBIC, RMSE, and MAPE of nine SARIMA models, and the best-fitting forecast model for dengue incidences is SARIMA (1,1,0) (1,1,1) model due having the lowest NBIC and RMSE, strongest correlation (r=0.718, p<0.05) between the forecasted and actual incidences trend before partial lockdown (1 st to 12 th weeks of 2020), and the model also passed the Ljung-Box Q Test (z = 19.782, p = 0.180). As seen in Fig. 4 , when we generated a simulated model on the trend of 13 th to 35 weeks of 2020, which coincided with the COVID-19 partial lockdown period in Malaysia, the actual dengue incidence trend was significantly diverted and demonstrated a stronger negative steepness compared with the simulated trend. The actual trend is lower than the lower confidence level of forecasted trend (Fig.4) , which implied that the actual dengue incidence trend significantly disobeyed the simulation, which presumed no lockdown, indicating that movement control greatly impacted dengue transmission in Malaysia. To further analyze the changes in dengue incidence trends due to the partial lockdown, we divided the time series of endemical weeks into eight stages as described in Table 1 Table 3 showed the slopes of the eight stages for previous years (mean weekly incidence of 2015-2019), the SARIMA model simulated trends, year 2019 and actual dengue incidences during the COVID-19 partial lockdown. Some researchers have proposed that the dengue incidences in 2020 were lower than those in 2019 [16] , but when we averaged the weekly incidences of the previous five years (2015-2019), 2020 had significantly higher dengue incidences at prelockdown 1 and 2 even though the slopes between the dengue incidences of the previous years and those of the year 2020 during the period of prelockdown (1 and 2) are fairly the same. When Malaysia imposed phase 1 of the partial lockdown, the slope declined dramatically, which was 319% steeper than in previous years and 650% steeper than SARIMA simulated trend. These provide a strong implication that movement control during partial lockdown significantly reduced the reported dengue incidences. Although at phases 2 to 4, the incidences in 2020 were significantly lower than those in previous years, when we compared the stages for the slope to become positive (which indicates an upsurge in dengue incidences), this change in slope occurred in 2020 two stages (4 weeks) earlier than in previous years; specifically, a positive slope was obtained at phase 3 in 2020 compared to in previous years, in which a positive was obtained at phase 5. Furthermore, at phase 5, the steepness of the slope of the year 2020 spiked from phase 4 to phase 5 compared to previous years, with the steepness increasing from 22% to 227%, suggesting a significant increase in dengue transmission. To study one of the factors that contributes to the spread-out of dengue transmission during the COVID-19 partial lockdown, we assessed the abundance and distribution of vectors during the partial breakdown. Fig. 6 shows the total numbers of Ae. albopictus collected from the outdoor area of the sampling locations during the period of partial lockdown. Ae. albopictus is the predominant species in the outdoor area, with no Ae. aegypti was caught. The abundance of Ae. albopictus showed slight fluctuation patterns during the partial lockdown but still demonstrated a strong linear increment throughout the eight phases of partial lockdown, and showed strong correlation with Penang reported dengue incidences (r=0.952, p<0.001). In general, the total number of mosquitoes caught indoors was significantly lower than that outdoors, and both Ae. aegypti and Ae. albopictus were caught indoors, with the abundance of Ae. aegypti being relatively low and plateauing throughout phase 3 to 6. However, Ae. albopictus demonstrated higher abundance and exponential growth with the population during the same corresponding period (Table 4 ). The factors that contribute to dengue transmission are multifaceted, and the spatial variation in the contact rates of the host and vector are probably the most important factors for the dynamics of DENV [28] . With the movement restrictions (MRs) imposed in Malaysia due to the COVID-19 pandemic, we can investigate the effect of the large-scale MRs of the host on two interrelated variables: dengue transmission and Aedes mosquitoes occurrences. We analyze the dengue incidence trends by comparing their significant differences among the phases before and during the lockdown to those of the same corresponding periods for previous years and simulation. We first reported evidence that the MRs of the COVID-19 partial lockdown significantly influenced the weekly dengue incidence trend in Malaysia. Our findings provide direct evidence from analysis and extend the studies of Reiner et al. [13] and Falcón-Lezama et al. [5] , which demonstrated that people's movement affected dengue transmission by using a simulation model. The early decline in dengue incidences was also reported in India, with dengue cases dropping by 50% compared to previous years. The incidences. We also suggest that Ae. albopictus could be the key substitution vector that contributes significantly to dengue virus circulation, and therefore, the vector control direction and strategies should be redesigned. With no specific previous entomological data about female adult Ae. albopictus in the corresponding period with lockdown, we refer to Rozilawati et al. [14] and Rahim et al. Their results demonstrated that the indexes of Ae. albopictus for the corresponding period of phase 2 to 4 of lockdown should be lower, in contrast to our finding that the abundance of Ae. albopictus increased steadily from phases 1 to 5. There are several reasons for the increase in Ae. Albopictus. First, as proposed by the WHO [34], the upsurge of Aedes mosquitoes may be due to the southwestern monsoon (end of May to September), which brought a higher frequency of precipitation and higher humidity and temperature, and therefore, a higher breeding rate for the mosquitoes. Second, a minimum centralized vector control program can be conducted due to the stay-at-home policy. It is relevant to any method that is intended to reduce dengue incidences by reducing, but not eliminating, Aedes mosquito populations. Before that, researchers [35] have associated the index of the temporal vector with dengue occurrence, and the relationships between Aedes mosquito density and DENV transmission indexes for Ae. aegypti density are correlated with the prevalence of human dengue infections but are relatively weakly correlated with the incidences, indicating that other factors were involved in determining the incidence pattern. This is supported by the participant during the Aedes survey when a fogging activity was observed on May 28, 2020 (22 nd weeks of 2020, phase 5 -conditional movement control order), and total Ae. albopictus was significantly lower on May 29, 2020, but the mosquitoes caught afterward remained elevated in general. 1 3 5 7 9 1 1 1 3 1 5 1 7 1 9 2 1 2 3 2 5 2 7 2 9 3 1 3 3 3 5 3 7 3 9 4 1 4 H u m a n m o v e m e n t r e s t r i c t i o n s d u e t o C O V I D -1 9 l o c k d o w n i n M a l a y s i Pre -Pre-lockdown; P -phase Why lockdowns can halt the spread of COVID-19. World Economic Forum Modes of transmission of virus causing COVID-19: implications for IPC precaution recommendations Social contacts and mixing patterns relevant to the spread of infectious diseases World Health Organization (WHO) Day-to-Day Population Movement and the Management of Dengue Epidemics House-to-house human movement drives dengue virus transmission Operating businesses during MCO must be classified as offence Comparative study of four time series methods in forecasting typhoid fever incidence in China A SARIMA forecasting model to predict the number of cases of dengue in Campinas, State of São Paulo, Brazil. Revista da Sociedade Brasileira de Predicting Seasonal Influenza Based on SARIMA Model Research on Climate and Dengue in Malaysia: A Systematic Review Studies on the distribution and abundance of Aedes aegypti (L.) and Aedes albopictus (Skuse) (Diptera: Culicidae)in an endemic area of dengue/ dengue hemorrhagic fever in Kuala Lumpur Socially structured human movement shapes dengue transmission despite the diffusive effect of mosquito dispersal Urbanization Increases Aedes albopictus Larval Habitats and Accelerates Mosquito Development and Survivorship Seasonal abundance of Aedes albopictus in selected urban and suburban areas in Penang, Malaysia Density of Aedes aegypti and Aedes albopictus and its association with number of residents and meteorological variables in the home environment of dengue endemic area Surveillance of Aedes aegypti indoors and outdoors using Autocidal Gravid Ovitraps in South Texas during local transmission of Zika virus Indoor-Breeding of Aedes albopictus in Northern Peninsular Malaysia and Its Potential Epidemiological Implications 000 Dengue Cases In Male-female differences in the number of reported incident dengue fever cases in six Asian countries We extend our deepest gratitude to the Ministry of Health Malaysia for providing the data.