Zika, chikungunya and co-occurrence in Brazil: space-time clusters and associated environmental–socioeconomic factors

Chikungunya and Zika have been neglected as emerging diseases. This study aimed to analyze the space-time patterns of their occurrence and co-occurrence and their associated environmental and socioeconomic factors. Univariate (individually) and multivariate (co-occurrence) scans were analyzed for 608,388 and 162,992 cases of chikungunya and Zika, respectively. These occurred more frequently in the summer and autumn. The clusters with the highest risk were initially located in the northeast, dispersed to the central-west and coastal areas of São Paulo and Rio de Janeiro (2018–2021), and then increased in the northeast (2019–2021). Chikungunya and Zika demonstrated decreasing trends of 13% and 40%, respectively, whereas clusters showed an increasing trend of 85% and 57%, respectively. Clusters with a high co-occurrence risk have been identified in some regions of Brazil. High temperatures are associated with areas at a greater risk of these diseases. Chikungunya was associated with low precipitation levels, more urbanized environments, and places with greater social inequalities, whereas Zika was associated with high precipitation levels and low sewage network coverage. In conclusion, to optimize the surveillance and control of chikungunya and Zika, this study’s results revealed high-risk areas with increasing trends and priority months and the role of socioeconomic and environmental factors.

clinical-epidemiological criteria and distributed across 3042 and 2050 municipalities of residence, respectively (Fig. 1 and see Supplementary and Table S1a online).Of these, 850 (chikungunya) and 74 (Zika) patients distributed across 229 and 44 municipalities of residence, respectively, died during the study period (see Supplementary Fig. S1 and Table S1a and S2 online).
The incidence rates per 100,000 inhabitants-year in Brazil were 42.90 and 11.40 for chikungunya and Zika, respectively.Incidence rates were higher among women than among men, and in the age groups>15 years for chikungunya and>10 years for Zika, with differences that persisted to approximately 79 years (Fig. 2).There was also a predominance of black and pardo race/color for chikungunya (53.73%) and Zika (37.15%).Approximately 60% of the records presented no data on the educational level (Table 2).

Purely spatial analysis
Univariate (Chikungunya and Zika considered individually) and multivariate (evaluating the co-occurrence of both diseases) purely spatial analyses identified 38, 53, and 20 significant clusters, respectively (Fig. 4, see Supplementary Tables S3 and S4 online).The Gini index for all the above analyses was 4% for the population of the scan windows.Chikungunya clusters were distributed across more municipalities (707) than those of Zika (520) and co-occurrence of both diseases (186).www.nature.com/scientificreports/Regarding chikungunya clusters, the Brazilian region with the most municipalities was the northeast (552), followed by the southeast (113), north (39), central-west (2), and south (1).The distribution of Zika clusters had another pattern, with more municipalities in the central-west (284), southeast (117), northeast (68), and north (21); however, no municipality was identified in the south.Co-occurring clusters were distributed across 103 municipalities in the central-west, 63 in the southeast, 11 in the north, eight in the northeast, and one in the south.
Spatial Zika clusters with the highest RR were identified in municipalities located in the state of Bahia (BA), one in Itabuna (cluster 1, RR = 118.04)and another in Uibaí (cluster 8, RR = 56.00).In contrast, the two spatial chikungunya clusters with the highest RR were identified in Itabuna and Barro Preto in BA (cluster 4, RR = 23.06) and Várzea Grande in Mato Grosso (cluster 7, RR = 15.74).In Itabuna and Barro Preto in BA, co-occurrence analysis also revealed the cluster with the highest RR for chikungunya and Zika (cluster 1, RR = 23.06 and 115.69, respectively) (Fig. 4, see Supplementary Tables S3 and S4 online).

Space-time analysis
The space-time analysis for chikungunya, Zika, and co-occurrence revealed 24, 27, and 13 high-risk clusters, respectively.This analysis showed that most clusters began between January and April throughout the study period, except for chikungunya (cluster 6) and co-occurrence (cluster 5), which began in November 2015, chikungunya (cluster 21) in June 2019, and Zika (cluster 21) in December 2016 and (cluster 20) in May 2020 (Fig. 5, see Supplementary Tables S5 and S6 online).For chikungunya, clusters that started in 2015-2017 predominated in the north and northeast regions, whereas in other regions, the clusters were identified predominantly between 2018 and 2021.However, the most recent cluster was located in the northeast region between April and August 2021 (cluster 8, RR = 17.08) (Fig. 5).For Zika, clusters that started in 2016-2017 predominated in all regions of Brazil, except for three clusters that occurred between 2019 and 2021 in Pernambuco (cluster 25), Roraima (cluster 20), and Paraíba (cluster 14).In multivariate analysis, clusters that started between 2015 and 2017 predominated, except for three clusters that started in 2021 in the northeast: Paraíba and Pernambuco (cluster 6) and Bahia (cluster 8), and in the south: Rio Grande do Sul (cluster 12) (Fig. 5, see Supplementary Tables S5 and S6 online).

Spatial variation analysis in temporal trends
Spatial variations in the temporal trends for chikungunya and Zika decreased across the country by 13% and 40%, respectively, with 47 and 28 significant clusters identified for chikungunya and Zika, respectively.Despite the decreasing temporal trend of chikungunya throughout the country, spatial variation clusters with a growing internal trend predominated in practically all states, with annual growth of 0.85%-96.56%.Only seven of the 47 clusters of chikungunya demonstrated a decreasing trend.Spatial variation analysis of temporal trends for Zika identified 16 clusters with an increasing annual trend and 12 with a decreasing trend, with internal trends ranging from −32.96% to 53.03%.Clusters with annual increases were located in the north and northeast, in www.nature.com/scientificreports/ the states of Rio Grande do Sul, Mato Grosso, Minas Gerais, Bahia, and Espírito Santo, and on the coast of São Paulo (Fig. 6, see Supplementary Table S7 online).

T-test of socioeconomic and environmental variables
The t-test for statistical comparison of socioeconomic and environmental variables between municipalities included in purely high-risk spatial clusters and those not included showed significant differences between groups.Municipalities included in high-risk clusters for both diseases had higher temperatures than those not included.
Municipalities included in high-risk clusters for chikungunya had lower precipitation levels and a lower percentage of households with a sewage system or septic tank, running water, and garbage collection by cleaning services than those not included.These municipalities also had lower Normalized Difference Vegetation Index (NDVI) values in urban areas and higher Brazilian Deprivation Index (BDI) measurements.
In contrast, Municipalities included in high-risk clusters for Zika had higher precipitation levels and a lower percentage of households with a sewage system or septic tank than those not included.These municipalities had lower BDI measurements, a higher percentage of households with garbage collection by cleaning services, and lower NDVI values in urban areas.All analyses resulted in significant p values, except for the percentage of households served piped water for Zika (Table 3, see Supplementary Fig. S2 online).

Discussion
Scan statistics were used to identify high-risk areas for chikungunya, Zika, and their co-occurrence in Brazil between 2015 and 2021.The results were consistent with those of previous studies, indicating northeast Brazil as the region with the highest incidence of these diseases between 2015 and 2016 53,57,58 .These risk areas remained  S3 and S4.S5 and S6.active in 2017 for chikungunya.The results also showed a decline in the incidence of both conditions in Brazil between 2018 and 2021.
The spatial and temporal patterns of the two arboviruses have been changing, with the dispersion of their occurrence from the northeast to the central-west region, a change already reported in the literature for Zika 58,59 .A resurgence in the occurrence was identified in the northeast between 2019 and 2021 for Zika and in 2021 for chikungunya.Dispersion and an upward trend were also observed for chikungunya along the coast of São Paulo, especially in Cubatão, Guarujá, Santos, and São Vicente in 2021 60 and in Rio de Janeiro between 2018 and  S7.Out = Outside the cluster.In = Inside the cluster.www.nature.com/scientificreports/2021.However, some hypotheses for these pattern changes include the degree of susceptibility of populations in different Brazilian regions 58,61 and climate change associated with greenhouse gas emissions 62 .For example, the highest temperature increase in the state of São Paulo over the last few decades was observed in municipalities located on the São Paulo coast 63 .Seasonal scan analysis revealed that the greatest risk of occurrence of these diseases in the first months of the year was during summer when higher temperatures occur in the Brazilian regions, consistent with the results of our statistical analyses.This was related to the increased Ae. aegypti infestation levels due to the decreased time for larval development and increased proportion of infectious mosquitoes, given the decreased intrinsic incubation periods of the viruses in the vector 58,64 .Studies conducted in China, the United States, and the states of Rio de Janeiro and São Paulo also indicated that temperature influenced the distribution patterns of Ae. aegypti and Ae.albopictus, consequently influencing the incidence of diseases they transmit [65][66][67][68][69] .The small but significant differences (from 0.7 to 2.6 • C) in the average temperature between the high-risk and no-risk areas for both diseases are worth consideration.Banu et al. 70 showed that an increase of 1 • C could be related to a future rise in arbovirus cases.
In recent decades, consistent and widespread warming has been observed throughout Brazil, with greater extreme heat occurring during spring and summer 71 .Increased temperatures are also expected in the coming years in Brazil, mainly in the Amazon, Tocantins, and Paraná River basins, based on future climate change scenarios and considering increased greenhouse gas levels 72 .This may indicate a greater risk of the occurrence of arboviruses in the future.
Furthermore, several factors may be associated with the inverted results obtained for precipitation behavior in clusters at high risk of chikungunya and Zika.One of the hypotheses is that the risk may increase after extreme conditions, either in places with a lot of drought or rain, as reported in dengue cases in Brazil 73 .In addition, these authors related extreme precipitation levels to urbanization and socioeconomic factors.They concluded that the risk of dengue was higher under extremely wet conditions in more rural areas and after extremely dry conditions in highly urbanized areas with a higher frequency of water shortage 73 .This is consistent with the present study because the areas at high-risk of chikungunya were more socioeconomically unfavorable and had lower precipitation levels and a lower percentage of water supply.Lower proportions of households with water supply, sewage systems, garbage collection, and higher BDI values (a larger percentage of the population with per capita income below half of the minimum wage, illiterate people, and inappropriate households) in high-risk clusters for chikungunya indicate that these are vulnerable areas from a socioeconomic point of view 74 .In addition to the lower levels of precipitation identified in these clusters, these could lead to increased breeding sites and an abundant mosquito population owing to inadequate water storage and waste disposal, thereby raising the incidence of the disease 1,7,53 .The risk of Zika in places with higher precipitation levels, as highlighted by Lowe et al. 73 regarding dengue, also indicates that these areas are suitable for developing Aedes by providing an increased number of artificial and/or natural breeding sites.
Despite the significant difference, the similarity between BDI means for areas at high-risk of Zika and other diseases may indicate that socioeconomic variables had little influence on their distribution in the Brazilian territory.The differences between worse sewage collection indicators and better garbage collection indicators in areas at risk of Zika may be associated with a large proportion of municipalities in high-risk clusters located in the central-west region of Brazil, a region that, despite having good piped water and garbage collection coverage, has precarious sewage system coverage 75,76 .According to the 2017 Brazilian National Basic Sanitation Survey (PNSP), the central-west has, second to the north region, the lowest proportion of municipalities with sewage system services (38.1%) 76 .
The positive relationship with garbage collection in municipalities with high-risk clusters for Zika can be a confusing factor.Municipalities have cleaning services; however, this does not assure proper solid waste disposal, which may end in dumping grounds (open pits).In 2008, the central-west, north, and northeast regions ofBrazil had the highest proportions of municipalities with this type of disposal.In addition, the central-west and northeast regions had the highest proportion of municipalities with waste pickers in dumps or landfills at 46.6% and 43.1%, respectively 75 .These conditions increase the number of possible vector-breeding sites in urban areas.
Brazil has continental dimensions, with social inequalities, precarious socioeconomic levels, heterogeneous basic sanitation services, and climatic differences between the five regions, mainly regarding precipitation.A study that analyzed the Brazilian regions' seasonal and climatic trends between 1961 and 1981 reported that the northeast region underwent changes during a drier climate, mainly in the summer.In the southern region and southern parts of the central-west region, the climate became more humid; however, there were no significant changes in the other regions during this period 71 .Therefore, in addition to studies considering climate change scenarios with greater greenhouse gas emissions, more studies should be conducted at different geographic scales 58 .
A lower amount of vegetation in the urban area was statistically significant for chikungunya and Zika risk areas; nonetheless, their average values were similar.High-risk areas for Zika were associated higher precipitation levels, which could be associated with the greater presence of Ae. albopictus in these areas.This vector is most likely found in colder and wetter suburban and rural areas36; therefore, this could be a hypothetical explanation for the differences between diseases.Ae. albopictus naturally infected with ZIKV has already been reported in Brazil 77 .In addition, rapid index surveys for Aedes (LIRA) have shown that Ae. albopictus expanded its geographic distribution in Brazil between 2015 and 2020 and was registered in 37.4% of the surveyed municipalities.The central-west region demonstrated the most significant increase in this species among all the Brazilian regions 78 .However, previous studies reported that Ae. aegypti has a higher transmission rate and is more easily infected by ZIKV than CHIKV 79 .The interactions between the vectors (Ae.albopictus and Ae.aegypti) and viruses are subject to constant mutations; consequently, vector competence may change over time 59 .
The absence of a perfect overlap between spatial and space-time clusters for Zika and chikungunya in the present study is consistent with the results of other studies 41,59,80  www.nature.com/scientificreports/diseases could be associated with the abovementioned factors.However, simultaneous risk areas for both diseases stand out, such as Itabuna, BA, a place with a high incidence of dengue 81 .This study had some limitations, including the use of socioeconomic data from the 2010 census and PNSP data on solid waste, which was updated in 2008, and the use of secondary data on Zika and chikungunya, with a significant percentage of cases confirmed using clinical-epidemiological criteria.The MH recommends these criteria, which are used after confirmed sustained transmission in a certain geographic area 9 .
In conclusion, this study's results are useful for selecting areas at greatest risk, areas with increasing trends, and the months with the highest incidence, to optimize surveillance and control actions for Zika and chikungunya.This is necessary, considering the high costs associated with arboviruses (dengue, chikungunya, and Zika).In Brazil, costs for combating the vector and medical costs (direct and indirect) accounted for approximately 2% of the planned budget for the entire health area in 2016, showing a relevant impact on society 82 .The indirect costs of productivity loss during periods of medical leave for chikungunya and Zika accounted for a reduction of approximately 429 and 48 million dollars (mean BRL/USD exchange rate for the year 2016), respectively, compared with the 2016 GDP.The direct medical costs reached USD 291 million for both diseases 82 .This indicates that directing and optimizing the implementation of surveillance and control measures can provide a more rational use of available resources.It is also important to consider socioeconomic, basic sanitation, and climatic factors in decision-making.These can also help implement surveillance and control activities to avoid or minimize analyzed diseases.

Methods
The study area covers the Brazilian territory and its 5570 municipalities as units of analysis, which were grouped into five regions (Fig. 1), with a territorial area of 8,510,820,623 km² and an estimated population of 213,317,639 people in 2021 83 .
Data on the date of symptom onset, patient age, sex, race/color, disease progression, municipality of residence, classification, epidemiological profile, and confirmation criteria for the Zika and chikungunya cases notified in Brazil between 2015 and 2021 were obtained from the databases of the Notifiable Diseases Information System (SINAN) using the website of the Department of the Unified Health System of the Brazilian MH 84 .The spatial analysis included cases with laboratory and clinical-epidemiological confirmations.Imported cases, cases with no information on sex, and patients not residing in Brazil were excluded from the scan analyses in this study.
Incidence and mortality rates (per 100,000 inhabitants-year) were mapped by the municipality of residence between 2015 and 2021 and by each condition using Quantum Geographic Information System (QGIS) software version 3.22 85 .In addition, graphs (based on age group and sex) were created for the absolute numbers and incidence rates of each disease in Brazil during the study period using R software version 4.1.0 86.Estimates considered the mid-period population (2018) obtained from the Brazilian Institute of Geography and Statistics (IBGE) and cartographic materials (municipal and regional meshes) 87,88 .
High-risk areas for the occurrence of chikungunya and Zika were identified based on scan statistics; relative risks (RR), which correspond to the ratios between the incidence rates inside and outside the clusters, were obtained.Three tables were built: a) chikungunya and Zika cases based on the municipality of residence, age group (classified into 11 groups: 0-4, 5-9, 10-14, 15-19, 20-29, 30-39, 40-49, 50-59, 60-69, 70-79, and >= 80 years), and sex for each symptom onset date; b) 2018 population based on age group (11 groups described above) and sex; and c) centroid coordinates of Brazilian municipalities 88 .Scan statistics were performed using the SatScan software version 10.0.02 89 .
Univariate scan analyses were performed for chikungunya and Zika cases (considered individually), and multivariate analyses were performed for both diseases (together) to assess co-occurrence.These analyses compared the number of observed and expected cases inside and outside possible clusters in multiple window sizes.The windows are circles (in the space and spatial variation in temporal trend analysis) or cylinders with a circular base and a time interval as the height (in the space-time analysis).The expected cases were obtained through an indirectly standardized method considering sex and age 90 .The SaTScan considers, for the analyses carried out in this study, the discrete Poisson model, where the number of cases in each location is Poisson-distributed.This probability distribution is well-suited for analyzing event count data, such as disease occurrences.This model was used under the following conditions: circular shape clusters, no geographical overlapping, adjustment for age and sex, and using the Monte Carlo method with 999 repetitions to estimate probabilities.Purely spatial, purely temporal, seasonal, and space-time analyses were performed to find high-risk rates, and spatial variation in temporal trend analysis was performed to find clusters with high and low temporal trends 90 .The RR for each cluster is the estimated risk within it divided by the estimated risk outside it, as presented in the following formula 90 : where, C and E[C] are respectively, the observed and expected number of total cases, and c and E[c] are, respectively, the observed and expected number of cases within the possible cluster 90 .The Gini index was used in the univariate and multivariate purely spatial analyses were used to determine the maximum population size of the scan windows.This Gini index was used to optimize the size of the population included in a specific cluster, to avoid finding the big ones only 91 .The same value obtained for purely spatial analysis was considered in spacetime and spatial variation in the temporal trend analyses.The month range was used as the aggregation time for purely temporal, seasonal, and space-time analyses, whereas the year range was used for spatial variation analysis in temporal trends..In this last analysis, the scan statistics estimate the annual percentage increase (or decrease) of temporal trends inside and outside the possible clusters.It identifies clusters where the inside temporal trends are statistically different from the outer ones 90 .In addition, multivariate analyses (purely spatial and space-time)

Figure 1 .
Figure 1.Map of South America, Brazil, and its regions (A), map of Brazil and its states, with the incidence rate (100,000 inhabitants-year) of confirmed chikungunya (B) and Zika (C) cases based on the municipality of residence, with symptom onset between 2015 and 2021.(B) Legend with the names of the states, (C) legend with the abbreviations of the states.

Figure 2 .
Figure 2. Graph of absolute numbers based on age group and sex (A) and incidence rates (per 100,000 inhabitants-year) (B) of chikungunya and Zika cases in Brazil, with symptom onset between 2015 and 2021.

Figure 3 .
Figure 3. Purely temporal analysis based on month of the distribution of chikungunya (A) and Zika (B) cases, seasonal analysis of the distribution of chikungunya (C) and Zika (d) cases, and multivariate analysis (red lines) in Brazil between 2015 and 2021.

Figure 4 .
Figure 4. Spatial analysis of chikungunya (A) and Zika (B) cases, and multivariate (C) analysis in Brazil between 2015 and 2021.N = clusters' identification number.See Supplementary TablesS3 and S4.

Figure 5 .
Figure 5. Space-time analysis of chikungunya (A) and Zika (B) cases, and multivariate (C) analysis in Brazil between 2015 and 2021.N = clusters' identification number, see Supplementary TablesS5 and S6.

Figure 6 .
Figure 6.Spatial variation analysis in temporal trends of chikungunya (A) and Zika (B) cases in Brazil between 2015 and 2021.N = clusters' identification number, see Supplementary TableS7.Out = Outside the cluster.In = Inside the cluster.

Table 1 .
Table with previous studies considered cluster analysis of the co-occurrence of Zika and chikungunya analyzed simultaneously.*All incidence = Incidence of the sum of three diseases cases: dengue, CHIKV and ZIKV.

Table 2 .
Number of confirmed cases and epidemiological profile of chikungunya and Zika in Brazil, with symptoms onset between 2015 and 2021.

Table 3 .
Statistical comparison of the means (t-test) of environmental and socioeconomic variables of municipalities included in the highest risk cluster with those not included in the purely spatial clusters of chikungunya and Zika in Brazil between 2015 and 2021.*p value < 0.001 .**p value < 0.01 .NS Not significant.Numbers in parentheses indicate the standard deviation.