Space-time dynamics of the dengue epidemic in Brazil, 2024: an insight for decision making

Abstract

Background

Dengue is a vector-borne viral infection caused by the dengue virus transmitted to humans primarily by Aedes aegypti. The year 2024 has been a historic year for dengue in Brazil, with the highest number of probable cases ever registered. Herein, we analyze the temporal trend and spatio-temporal dynamics of dengue cases in Brazil during the first nine epidemiological weeks (EW) of 2024.

Methods

This is an ecological study, including all probable cases of dengue in Brazil during the period, carried out in two steps: time series analysis to assess the temporal trend and spatial analysis to identify high-risk clusters.

Results

1,345,801 probable cases of dengue were reported. The regions with the highest increasing trend were the Northeast with an average epidemiologic week percent change (AEPC) of 52.4 (95% CI: 45.5–59.7; p < 0.001) and the South with 35.9 (95% CI: 27.7–44.5; p < 0.001). There was a statistically significant increasing trend in all states, except Acre (AEPC = -4.1; 95% CI: -16.3–10; p = 0.55), Amapá (AEPC = 1.3; 95% CI: -16.2–22.3; p = 0.9) and Espírito Santo (AEPC = 8.9; 95% CI: -15.7–40.6; p = 0.5). The retrospective space-time analysis showed a cluster within the Northeast, Central-West and Southeast regions, with a radius of 515.3 km, in which 1,267 municipalities and 525,324 of the cases were concentrated (RR = 6.3; p < 0.001). Regarding the spatial variation of the temporal trend, 21 risk areas were found, all of them located in Southeast or Central-West states. The area with the highest relative risk was Minas Gerais state, where 5,748 cases were concentrated (RR = 8.1; p < 0.001). Finally, a purely spatial analysis revealed 25 clusters, the one with the highest relative risk being composed of two municipalities in Acre (RR = 6.9; p < 0.001).

Conclusions

We described a detailed temporal-spatial analysis of dengue cases in the first EWs of 2024 in Brazil, which were mainly concentrated in the Southeast and Central-West regions. Overall, it is recommended that governments adopt public policies to control the the vector population in high-risk areas, as well as to prevent the spread of dengue fever to other areas of Brazil.

Text box 1. Contributions to the literature
– Identifying vulnerable areas: This study shows that the dengue epidemic in Brazil was concentrated in the Central-West and Southeast regions.
– Understanding epidemic dynamics: This study revealed which states and regions had the highest average weekly percentage change during the first nine epidemiological weeks of the 2024 dengue outbreak in Brazil.
– Targeting high-risk areas: Our study highlights the importance of rapid and effective government action to prevent the spread of dengue to other areas of Brazil and to subsidize specific interventions in high-risk areas.

Background

Dengue is a vector-borne viral infection caused by the dengue virus (DENV), which is transmitted primarily by Aedes aegypti mosquitoes [1]. Dengue is a dynamic, systemic, acute febrile illness, with a broad clinical spectrum, which can be severe and fatal in some individuals [2]. Dengue is more common in tropical and subtropical regions [1] and Brazil is an endemic region [2].

Worldwide, dengue is endemic in approximately 100 countries. The Americas, Africa, the Eastern Mediterranean, the Western Pacific and Southeast Asia are the most affected regions, with the latter accounting for 70% of the global disease burden [1]. In 2019, there were an estimated 56.8 million new cases worldwide, of which approximately 5 million occurred in Latin America and the Caribbean, including 2.1 million in Brazil [3].

The incidence of dengue cases and fatalities emerges from a multifaceted interplay of viral, host, and environmental elements [4, 5]. In addition, dengue has been associated with urbanization and the presence of water bodies [6], poor basic sanitation and inadequate garbage collection [7], and more recently with climate change [8]. In Brazil, the unequal access to clean water, basic sanitation, garbage collection, malnutrition, low socio-economic development and poor public health structure create the perfect scenario for Aedes spp. to breed and spread dengue, making the North and especially the Northeast the most vulnerable regions [9].

The year 2024 has been a historic year for dengue in Brazil, with the highest number of probable cases ever registered [10]. Between epidemiological weeks 1 and 31 of 2024, there were approximately 6.4 million probable cases and 5,067 deaths in Brazil [11]. One of the factors influencing this recent outbreak may be the current trend of expansion and exacerbation of dengue incidence in previously unaffected regions of Brazil [12]. In the southern region of the country, for example, which reported few cases of dengue in previous years [5, 6], an increase in cases was observed in 2024, width municipalities that had never reported cases now doing so [5, 6].

It is also known that in January and February 2024 some state capitals in Brazil, such as Brasília/DF, Porto Alegre/RS and Palmas/TO, have registered above average precipitation rates [13]. The increase in precipitation rates in those regions may have also increased the breeding of mosquitoes; as a study conducted in five Latin American cities has shown, average pupae population was 800% higher during the wet season in the study area in comparison to the dry season [14]. The amplitude of daily temperature changes, the mean temperature, the virus dose and also the interplay between the vector and dengue genotypes influence vector breeding and competence [15].

In the past few years, spatial analyzes of dengue had been conducted in Brazil. Most of them, however, encompassing smaller areas, such as an investigation of dengue spatial dynamics from 2009 to 2018 in the state of Bahia [16], a similar study conducted from 2014 to 2017 in the Northeastern region of Brazil [17], and a third one, from 2001 to 2019, in the state of Ceará [18]. All of them have found clusters of high relative risk in their respective study areas.

Considering the WHO’s road map for neglected tropical diseases 2021–2030, robust surveillance systems are a key aspect of a government plan to control and eliminate endemic diseases such as dengue [19]. The early detection of hot spots for dengue outbreaks and understanding of the epidemic dynamics is important to coordinate actions of vector control [19]. This study can contribute to the identification of such areas and to raise the awareness of the government for similar events in the future, which hasn’t yet been provided.

Based on the above, our objective was to analyze temporal trend and spatio-temporal dynamics of dengue in the country during the first nine epidemiological weeks of 2024.

Methods

Study design and area

This is an ecological study that includes all probable cases of dengue in Brazil during the first nine epidemiologic weeks (EWs) of 2024, when the data for this study were collected.

Brazil is a country in South America with a population of about 203 million people and a surface area of 8.5 million Km², which means a demographic density of 23.9 inhabitants/Km² [20, 21]. Brazil is divided into five regions: North, Northeast, South, Southeast and Central-West, within which are the 26 states and the Federal District that make up its territory (Fig. 1) [22]. The country has 5570 municipalities [20, 21].

Fig. 1

Geographical Regions of Brazil within South America

In 2022, Brazil’s Human Development Index (HDI) was 0.760 [23], and its latest Gini index was 52.9 [22]. Approximately 64.7% of households in Brazil are connected to some kind of sewerage system, 83.9% have running water, and 91.7% are served by garbage collection systems [21]. In Brazil, 15% of the population lives under a roof without access to improved water, access to improved sanitation, adequate living space, housing durability, or tenure security [23].

Data source

All probable cases of dengue in the first nine EWs of 2024 (from January 1 to March 2 2024) in Brazil were included in the present study. The data were extracted from the Notification Disease Information System (Sistema de Informação de Agravos de Notificação, SINAN, acronym in Portuguese). SINAN is a health information system that aims to collect, transmit and disseminate epidemiologic data from the surveillance services of municipalities and states in Brazil [24].

From the absolute number of probable cases, the incidence rate was calculated as follows:

$$\eqalign{& Dengue\,Epidemio\,logical\,Week\,Incidence\,Rate\,(IRW)\, \cr & \,\,\,\,\,\,\,\,\, = \,{\matrix{ Number\,of\,new\,probable\,dengue \hfill \cr cases\,in\,a\,place\,and\,week \hfill \cr} \over {Population\,living\,in\,the\,place\,and\,week}}\, \times \,100,000 \cr}$$

(1)

In addition, population data by state and municipality were extracted from the 2022 Census information [25].

The use of probable cases is recommended by the Brazilian Ministry of Health, mainly due to the country’s limited capacity to offer serological tests for all suspected cases. To reduce the bias of overestimating the collective risk, cases with a negative test are excluded from this contingent. Thus, probable cases correspond to suspected cases minus cases ruled out by serological testing [11].

Study steps

Step 1 Time series analysis

For the temporal trend, a time series of the first nine EWs of 2024 was adopted. The trend was analyzed at two levels: region and state.

To perform the temporal analysis, we used the joinpoint regression model. This model tests whether a line with multiple segments is statistically better at describing the temporal evolution of the data than a straight line or one with fewer segments. Thus, the model allows us to determine whether the temporal trend in the data is stationary, increasing or decreasing; its statistical significance through the slope of the regression line; the points where there is a change in this trend (joints); the epidemiologic week percent change (EPC); and the average change for each period (average epidemiologic week percent change, AEPC) [25].

The joinpoint regression model for the observations: (𝑥1, 𝑦1), (𝑥𝑛, 𝑦𝑛), where 𝑥1 ≤ … ≤ 𝑥𝑛 represents the time variable, and 𝑦𝑖, 𝑖 = 1, 2,…, n is the response variable, being given by:

$$\sum {{y_i}}\, |{x_i}] = {\beta _0} + {\beta _1}{x_1} + {\gamma _1}({x_i} - {\tau _1}) + \,...\, + {\gamma _n}({x_i} - {\tau _n})$$

(2)

where β0, β1, y1,…, γn are regression coefficients and yk, K = 1, 2,…, n, n < N, is the k-th unknown jointpoints where.

$$({x_i} - {\tau _k}) = ({x_i} - {\tau _k})\,if({x_i} - {\tau _k}) > 0$$

(3)

= 0, otherwise.

Parameters used in the joinpoint analysis: minimum: 0; maximum: 4; model selection: test with 4,499 permutations, 5% significance, 95% confidence interval and date-based autocorrelation of errors. It is importante to note that this model has been used in several studies reagarding dengue occurance in Brazil [16,17,18]. These analyses were performed in the Joinpoint Regression software (version 4.5.0, National Cancer Institute – USA).

Step 2 spatial analysis

The following spatial scan analysis techniques were used to identify high-risk clusters: retrospective space-time, spatial variation in temporal trend and purely spatial. Spatio-temporal statistics is defined by a cylindrical window with a cylindrical or elliptical geographic base and with the height corresponding to time. Spatial variation in temporal trend, on the other hand, includes the calculation of the temporal trend inside (inside time trend) and outside (outside time trend) the scan window. On the other hand, purely spatial scanning does not take trend variations into account [26].

The Poisson discrete probability model was used. This model allows the identification of spatial clusters and the calculation of the relative risk (RR) of each cluster. The RR represents how much more common the disease is in that location and time period compared to the baseline, i.e., it is the estimated risk within the cluster divided by the estimated risk outside the cluster. The test to identify clusters is based on the maximum likelihood method, with the alternative hypothesis that there is a high risk inside the window compared to outside [26]. The scanning statistic creates a circular window on the map, which is positioned in each of the different centroids and whose radius is set to 50% of the total population at risk and the minimum temporal cluster size equal to 1 (Generic) and the minimum cases in cluster for high rates equal to 2 [26]. The flexibility of the window is justified by the fact that the size of the spatial-temporal clusters are not known a priori, given that the population at risk is not geographically homogeneous. Monte Carlo simulations (999 permutations) were used to obtain p-values, where clusters with a p-value < 0.05 are significant. The software used to run the spatial analyses was SaTScan (version 10.1.3, National Cancer Institute, Bethesda, MD, USA) and the maps were created using QGis (version 2.14.11, Open Source Geospatial, Foundation (OSGeo), Beaverton, OR, USA).

Results

During the study period, 1,345,801 probable cases of dengue were reported. The eighth EW had the most records, 270,509 cases (20.10%). The incidence of dengue in Brazil increased from 12.62/100,000 to 125.36/100,000 during the study period. The regions with the highest increasing trends were the Northeast, with an AEPC of 52.4% (95% CI: 45.5–59.7 p < 0.001); the South, with 35.9% (95% CI: 27.7–44.5; p < 0.001); and the Southeast, with 33.0% (95%| CI: 22.7–44.2; p < 0.001). Regarding absolute frequency, the Southeast region was the one with more cases with 848,369 cases (63.4%), followed by the Central-West region with 222,510 cases (16.76%) and the South region with 188,412 (14.0%) (Fig. 2).

Fig. 2

Incidence rate, number of cases and average percentage change (AEPC) per epidemiologic week by region and country. Brazil, epidemiological weeks 1st to 9th, 2024 Legend: AEPC: average percentage change; CI: Confiance Interval

A statistically significant increasing trend was observed in all states, except Acre (AEPC = -4.1; 95% CI: -16.3–10; p = 0.55), Amapá (AEPC = 1.3; 95% CI: -16.2–22.3; p = 0.9) and Espírito Santo (AEPC = 8.9; 95% CI: -15.7–40.6; p = 0.5). The highest increasing trends were observed in Maranhão (AEPC = 65.4; 95% CI: 48.6–84.1; p < 0.001), Rio Grande do Sul (AEPC = 59.8; 95% CI: 51.6–68.5; p < 0.001) and Roraima (AEPC = 59.5; 95% CI: 43.6–77.3; p < 0.001). The highest incidence rates were observed in Distrito Federal, which reached 720.5/100,000 during the fifth EW, followed by Minas Gerais (323.3/100,000) and Goiás (133.2/100,000) (Table 1).

Table 1 Incidence rate (IRW) and average percentage change (AEPC) per epidemiologic week by state. Brazil, epidemiological weeks 1 to 9, 2024

The retrospective space-time analysis showed a cluster within Bahia, Espírito Santo, Goiás, Minas Gerais, Rio de Janeiro and São Paulo states, with a radius of 515.3 km, in which 1,267 municipalities and 525,324 cases were concentrated (RR = 6.3; p < 0.001). Regarding the spatial variation of the temporal trend, 21 risk areas were found, all within the Minas Gerais, São Paulo or Goiás states. The area with the highest relative risk was an area in Minas Gerais (cluster 13), where 5,748 cases were concentrated (RR = 8.1; p < 0.001). Finally, a purely spatial analysis revealed 25 clusters, the one with the highest relative risk being composed of two municipalities in Acre (RR = 6.9; p < 0.001), followed by cluster 1, within the borders of Bahia, Espírito Santo, Goiás and Minas Gerais (RR = 6.8; p < 0.001), and cluster 2, in Goiás, Minas Gerais and São Paulo (RR = 5.9; p < 0.001) (Table 2; Fig. 3).

Table 2 Retrospective analyses space-time and spatial variation in the temporal trend of the incidence of dengue in Brazil, 1st to 9th epidemiological week of 2024

Fig. 3

Restrospective space-time, spatial variation in temporal trend and purely spatial scanning statistics of dengue occurrence in Brazil, 1st to 9th epidemiologic weeks of 2024 Legend: In the figure, non-significant (p > 0.05) and one-municipality clusters were not included. For this reason, there’s a numeric variation in the order of the clusters. A- Retrospective space-time; B- Spatial variation in temporal trend and C- Purely spatial analysis. RR = Relative Risk

Discussion

The aim of this study was to analyze the temporal trend and spatiotemporal dynamics of dengue cases in Brazil during the first nine EWs of 2024. The results showed an increasing trend of dengue cases between the 1st and 9th EWs in most states and in all regions, as well as a heterogeneous spatial and temporal distribution during this period. Cases were concentrated in the Southeast and Central-West regions, which showed more clusters and higher relative risks in the spatial analyses.

Dengue epidemics are somewhat common in Brazil, there had been around 1.4 million cases in 2013, 1.6 million in 2015, 1.2 million in 2016 and 1.5 million in 2019, with 674, 986, 701 and 840 deaths respectively [27]. The latter epidemic was mainly caused by DENV-2 [27]. From 2008 to 2019, dengue incidence was higher in the states of Acre, Goiás, Mato Grosso do Sul, Minas Gerais and Espírito Santo; and mortality in the states of Minas Gerais, Goiás, Rio de Janeiro and São Paulo [28].

Initially, it is important to note that dengue is a seasonal disease. In Brazil, the increase in cases occurs between October of one year and May of the following year [2, 8]. However, the increase in cases in the first few months of 2024 was intense and much higher than in previous years [5]. The current dengue epidemic in Brazil began and was in the Central-West and South-east regions, which are suffering from warming and changes in the rainfall regime [12]. A study of the thermal optima of several mosquito-borne diseases has shown that climate warming may expand the geographic and seasonal ranges of vector-borne diseases, such as dengue and Zika [29].

A cluster of high relative risk was found in the space-time analysis during the nine-week period. This cluster consisted mainly of Central-West and Southeast municipalities, except for some municipalities in Bahia. A study in Brazil comparing dengue incidence from 2002 to 2010 showed the same pattern. In 2002, the areas of concentration were Rio de Janeiro, Pernambuco and Bahia, with 35.3%, 13.7% and 10.9% of cases, respectively. In 2010, the distribution was closer to the one we found, with the incidence concentrated in Minas Gerais, São Paulo and Goiás (21.1%, 20.3% and 10.3%, respectively) [30].

Although the incidence of dengue was not concentrated in the Northeast, it’s important to note that the region was found to have the highest increasing trend. This is a serious concern, especially since a study conducted in Brazil between 2000 and 2014 found that the risk of hospitalization was higher for those living in the Northeast (OR = 1.38; 95% CI: 1.11–2.10) [9]. Our results and these findings underscore the importance of government policies to prevent morbidity and mortality in the region.

The spatial variation analysis in temporal trend showed clusters in Goiás, Minas Gerais and São Paulo states. A similar profile has also been described for other arbovirus in previous years in Brazil [17, 31]. Another spatial analysis study carried out with data from 2015 to 2021 in Brazil showed that Chikungunya incidence was concentrated in the northeast and some Southeast states, and the incidence of Zika occurred was highest in the Central-West [31].

Despite not having analyzed the influence of the social determinants of health in this 2024 dengue epidemic, it is relevant to highlight the known influence of the living conditions among the population at risk and the occurrence of the disease. A study conducted in 2018 in the Brazilian legal Amazon found an association between inadequate or insufficient knowledge about the Aedes aegypti vector with illiteracy (p < 0.001) or low education (p < 0.001), and with not cleaning the water tank (p = 0.002) and not using insecticides at home (p = 0.007) [32].

Not only health education, but many other factors seem to influence dengue incidence. It is well known that poor sanitation is a major problem in Brazil and strongly influences the incidence of dengue in the country [7]. To make matters worse, a study of arbovirus prevention policies in Brazil concluded that basic sanitation is not sufficiently addressed in the tools used to prevent vector-borne diseases such as dengue, which may contribute to their generally low effectiveness [33].

A study conducted from 2014 to 2017 in Northeast Brazil found associations between dengue incidence and socioeconomic indicators [17]. The percentage of households with access to piped water was negatively correlated with dengue incidence [17]. These results are consistent with the idea that there’s a link between basic sanitation and dengue prevention.

In addition, a study conducted in Bahia from 2009 to 2018 to analyze the impact of the Social Vulnerability Index, the Municipal Human Development Index and other socioeconomic indicators concluded that there were significant correlations between these factors and dengue incidence. Illiteracy rates were positively correlated with dengue, as was the proportion of people with a per capita income of less than half the minimum wage [16].

Despite the methodological care taken, the present study has some limitations. The first is related to the use of secondary data, which is influenced by the quality of the city health surveillance systems. It is well known that in small municipalities, the information on notifiable diseases faces even greater challenges, especially regarding adequate confirmatory testing, underreporting, and monitoring the quality of records. In addition, another limitation concerns the use of probable dengue cases, which may somehow overestimate the risk. However, to minimize this bias, cases with a negative test are excluded. The use of this methodology in Brazil is due to regional inequalities in the supply/performance of serological tests, especially in the more remote areas of the country [11].

Conclusion

In conclusion, an increasing trend in dengue cases was observed during the first nine EWs in Brazil, and the spatial analysis showed that it was mainly concentrated in the Southeastern and Central-Western areas.

It is recommended that governments adopt public policies capable of controlling the proliferation of mosquitoes in high-risk areas to prevent the spread of dengue to other areas of the country, as well as subsidize specific interventions in high-risk areas, thereby effectively reducing dengue morbidity and mortality. This study provides public health policy makers with important evidence to inform their decisions.