Determining the relationship between dengue and vulnerability in a Brazilian city: a spatial modeling analysis

ABSTRACT

Dengue is a viral infection transmitted by the Aedes aegypti mosquito. This study aimed to assess the distribution of cases and deaths from dengue and severe dengue, and its relationship with social vulnerability in Belo Horizonte, State of Minas Gerais, Brazil, from 2010 to 2018. The incidence and lethality rates of dengue and their relationship with sex, age, education, skin color, and social vulnerability were studied using chi-square tests, Ordinary Least Squares (OLS), and Geographically Weighted Regression (GWR) analyses. The number of cases of dengue in Belo Horizonte during the study period was 324,044 dengue cases, with 1,334 cases of severe dengue and 88 deaths. During the past few decades, the incidence rate of both dengue and severe cases varied, with an average incidence rate of respectively 1515.5 and 6.2/100,000 inhabitants. The increase in dengue cases was directly related to areas with higher social vulnerability areas and more working-age people. Also, the disease is more severe in people self-declared as black, elderly, and male. The findings of this study might provide relevant information for health services in the organization of control and prevention policies for this problem, emphasizing the most vulnerable urban areas and categories.

Introduction

Dengue, an arboviral disease, is a neglected and emerging disease that is transmitted by the Aedes aegypti mosquitoes [1]. The risk of infection extends to more than half of the world’s population, primarily in tropical and subtropical regions [2,3,]. The Americas registered more than 3 million cases of dengue in 2019, surpassing the 2.5 million cases reported in 2015 [4]. The growing incidence of dengue cases in Brazil is a severe public health problem, generating costs for the Brazilian public health system [5–7. Over the past ten years, Brazil has faced major epidemics, including a spike in severe cases and deaths from the disease [8]. In 2019, there were over 1.5 million dengue cases, particularly in the states of the Midwest and Southeast, which constituted 67.9% of probable cases. Of these, more than 18,000 (1.2%) people were diagnosed with severe dengue or dengue with warning signs, resulting in 732 deaths [8].

Unlike rural areas, urban areas are complex mosaics of heterogeneous environmental and socioeconomic conditions, making them hotspots for dengue. The high population density in cities also increases exposure to mosquitoes and the chances of dengue transmission [9]. The increase in urban conglomerates, the generation of waste related to precariousness in basic sanitation systems, climate change, and deforestation [10] are some of the social and environmental factors that promote the rapid proliferation and dissemination of this vector [11]. These determinants affect the spread of diseases and the occurrence of epidemics in different parts of the same municipality differently in time and space [12].

Therefore, as dengue is one of the most relevant neglected diseases in Brazil and evaluating its risk is a way to help the city’s health system in the development of specific control policies for the most vulnerable population, this study aimed to describe the factors associated with the incidence and lethality of dengue in Belo Horizonte and to investigate the association with social and environmental vulnerability indicators used in routine health surveillance, using spatial modeling. Firstly, we aimed to evaluate the usefulness of this index in the planning and prioritization of areas for dengue control actions in Belo Horizonte. We hypothesized that more vulnerable areas would be at greater risk due to lower environmental quality, which could be a better habitat for the vector. Furthermore, the findings are intended to inform urban planners and public health professionals about vulnerability management as a tool to mitigate the transmission of dengue and other arboviruses in Brazil.

Data and methods

This work is a collaboration between the Federal University of Pelotas and the health surveillance system responsible for controlling dengue in Belo Horizonte and the Urban Health Observatory of Belo Horizonte (OSUBH) of the Federal University of Minas Gerais (UFMG). An ecological examination was performed based on data associated with dengue in Belo Horizonte, MG, Brazil, between 2010 and 2018. The spatial and temporal distribution was examined. Further, the association between social determinants (age, sex, education level, and skin color) and the occurrence of severe dengue and death was analyzed. The study also explored the relationship between intra-urban indicators of social vulnerability used in routine health surveillance and dengue incidence in the diverse municipal coverage areas.

Study area

Belo Horizonte is the sixth most populous municipality in Brazil. Located in the southeastern part of Brazil between the latitude of 19°49’01” S and longitude of 43°57’21” W, it is the capital of the state of Minas Gerais. In 2020, its estimated population is 2,521,564 inhabitants. It has an area of 331.4 km² with a population density of 7,615.53 inhabitants/km² [13]. It has a tropical climate with an average annual temperature of around 21°C and annual rainfall exceeding 1,300 mm [14].

The city has a high literacy rate (97.6%). According to statistical data from 2017, the per capita gross domestic product was R$ 35245.02, and 96.2% of the municipality has adequate sanitary sewage[13].

The municipal territory is divided into nine administrative regions, which are further subdivided into 152 coverage areas assigned to health centers, composed of sets of contiguous census sectors, thereby allowing local health teams to have information about the population living in the area [15].

Data collection

The outcome variable is dengue incidence (cases per 100.000 populations) from 2010 to 2018. Dengue cases are reported to the Notifiable Diseases Information System, and for Belo Horizonte, we obtained the information through the City Health Department. As part of the BH-VIVA database Project [16], the Urban Health Observatory of Belo Horizonte carried out a data cleaning and consistency check, and it georeferenced each case to a census tract using the coordinates of the patient’s residence [9]. The cases included in this study were validated by clinical epidemiological or laboratory criteria as described by [17]. Data regarding age, sex, education, skin color, dengue classification, and death in the municipality were analyzed.

We further used the Health Vulnerability Index (HVI) to measure socioeconomic vulnerability. The HVI, developed by the Municipal Health Department of Belo Horizonte (SMSA/BH) and utilized in the routine of municipal health surveillance to characterize health risk areas, was updated in 2012. HVI is a measure of the health vulnerability in Belo Horizonte census tracts based on sanitation and socioeconomic indicators. Therefore, the sanitation index was built from the following indicators: i. % permanent private households with inadequate or no water supply, ii. % permanent private households with an inadequate or no sanitary sewage system, iii. % permanent private households with inadequate or no waste destination. The socioeconomic index was constructed from the following parameters: I. The ratio of residents per household, II. % illiterate people, III. % private households with per capita income up to half the minimum wage, IV. Average monthly nominal income of responsible persons (inverted) and V. % people of black, brown, and indigenous race/color. HVI is measured from 0 to 10, with higher values indicating more vulnerable areas.

Statistical analysis

For the consolidated period between 2010 and 2018, descriptive analyses were conducted on gender, age, race, and education related to dengue cases; dengue with warning signs and severe dengue, classification performed according to [18]based on the clinical signs of the disease. The morbidity and lethality of dengue, dengue with warning signs, and severe dengue in Belo Horizonte were characterized by the following parameters: (i) incidence rate (IR), and (ii) lethality rate (LR).

$$\mathrm{IR}=\left(\mathrm{denguecases}/\mathrm{population}\right)\times 100,000$$ (i)

$$\mathrm{LR}=\left(\mathrm{denguedeaths}/\mathrm{denguecases}\right)\times 100.$$ (ii)

Bivariate statistical analysis was conducted on dengue cases and dengue deaths and their social determinants, including age, gender, education, and skin color, using chi-square or Fisher’s exact tests, when necessary. The variables used in these analyzes were available in the Notifiable Diseases Information System database (SINAN). According to the standardization method used in the 2010 Brazilian Institute of Geography and Statistics (IBGE) census, age categorization was performed at intervals of 5 and 10 years [19]. A minimum confidence level of 95% (p < 0.05) was employed for all tests, including the relative risk-RR. The SPSS version 20.0 was used to perform all the statistical analyses.

Generalized additive models (GAMs)

Non-linear relationships without a defined shape can be modeled using GAMs [20]. They are based on non-parametric functions, called smoothing curves, in which the association shape is determined by the data [20,21,]. This smoothed curve is nothing more than some kind of Y_i average value in the vicinity of a given x_i value, which allows for describing the shape, and even showing probable nonlinearities in the studied relationships since it does not have the rigid structure of a parametric function [20].

We followed [22]to select the best GAM and used a Poisson likelihood and cubic splines with 5 knots on all predictors. We used a Poisson likelihood (we trained equivalent models with negative binomial and Gaussian likelihoods) to tune the type of spline (shrinkage cubic or cubic) and the upper limit on the degrees of freedom (df) associated with the spline (k = df-1 = {3,4,5,6,7,8}) [22].

The general formula of GAM with Poisson likelihood is as follows:

$$y_i\mathit{Poisson}\left({\mathrm{Z}}_i\right)$$

$$\mathit{log}\left({\mathrm{Z}}_i\right)=b_0+\sum s_j\left(x_{\mathit{ij}}\mathit{k}\right),$$

where $y_i$ is the observation i, ${\mathrm{Z}}_i$ is the linear predictor for the observation i, $b_0$ is the intercept, $s_j$ is the spline for predictor $x_{\mathit{ij}}$ and $\mathit{k}$ is the number of knots [22].

The response variable was the incidence of the confirmed response variable and linkage functions to measure the predictor variables’ effects on the dependent regressors, as stated by [23] and [21]. Therefore, GAMs were adjusted to verify the relationship between dengue occurrences over time. Gross rates in Belo Horizonte were adjusted. The response variables were then observed, reported, and confirmed cases with a Poisson distribution. The notification year variable formed the linear predictor models with a smoothing function (spline), and the offset term represented the natural logarithm of the population exposed in each notification year. The analyses were conducted using R software and the ‘mgcv’ and ‘tidyverse’ packages.

Spatial modeling

The terrestrial system of geographic coordinates was used to carry out the spatial analysis according to ordered pairs of coordinates (x, y) for the incidence of dengue. Hence, annual thematic maps depicting the spreading of dengue cases in the city were constructed. The maps were acquired from the IBGE cartographic database in shapefile format (.shp) and subsequently analyzed using the QGIS 3.4.7 software. Furthermore, the association between the indicators that make up the HVI (Table 1) was evaluated as predictors of the monthly incidence rate of dengue in each area covered by Belo Horizonte over the years by analysis of Ordinary Least Square linear regression (OLS) and the Geographically Weighted Regression (GWR). Firstly, multiple OLS regression models were built to get global regression equations, and the statistically significant independent variables were selected to construct thematic maps of the GWR regression.

Table 1.

Indexes that make up the Health Vulnerability Index (HVI) and its indicators.

Indicator	Indexes
Sanitation	Percentage of permanent private households within an adequate or no water supply
	Percentage of permanent private households with inadequate or no sanitary sewage
	Percentage of permanent private households with garbage destination inappropriately or absent
Economic partner	The ratio of residents per household
	Percentage of illiterate people
	Average monthly nominal income of responsible persons (inverted)
	Percentage of black, brown, and indigenous peoples

OLS is a traditional regression method that evaluates a global regression coefficient, which is constant over space (Eq. (1)).

$$\mathit{Y}={\beta }_0+\sum _{\mathit{i}=1}^n{\beta }_i{X}_i+\mathit{\epsilon }$$ 1

where Y is the dependent variable, β₀ is the intercept, X_i is the i^th independent variable, β_i is the i^th regression coefficient, ε is the error term, and p is the number of independent variables.

The following assumptions were considered to perform the OLS regression: i. Multicollinearity, verified by the variance inflation factor (VIF); ii. Stationarity through Koenker’s studentized Breusch-Pagan tests, and iii. Normality in the distribution of errors by the Jarque-Bera test. Furthermore, the representativeness of the independent over the dependent variable was evaluated by calculating the adjusted R² for each model built [24]. The R² represents the prediction ability of a regression model to fit the measured values of the dependent variable [25]. Additionally, global Moran’s I statistics were used to verify whether the assumption of the regression models that the residuals show a random spatial distribution is satisfied with our OLS and GWR findings. Moran’s I is a measure of spatial autocorrelation widely used to confirm the suitability of regression models [26–28].

Following the selection of the significant independent variables by the OLS regression, GWR regression models were built to determine differences in the relationship between the indexes and indicators and the incidence rate of dengue in the different areas within the municipality.

GWR is an extension of OLS regression in which locally varying parameters are considered in a spatially nonstationary sample [25,28,]. In Eq. (2), the location of the sample (u, v) is added to the regression equation to calculate the local parameters.

$$y_i={\beta }_0\left(u_J,v_i\right)+\sum _{\mathit{i}-1}^P\mathit{\beta k}\left(\mathit{ui},\mathit{vi}\right)\mathit{xik}+\mathit{\epsilon i}$$ 2

where y_i is the dependent variable for location i; u_i and v_i are the coordinates of location i; β₀(u_i, v_i) is the intercept at location i; β_k (u_i, v_i) is the local parameter estimate for an independent variable x_ik at location i, and ε_i denotes the error term.

Therefore, thematic maps of the coefficient and R² were constructed for each variable tested and the error map to verify the randomness in the distribution of the error of the models between the municipal coverage areas. The cross-validation Bandwidth method [25] was used to select the number of neighbors to be considered in the elaboration of regionalized regression equations. All spatial statistical analyses were executed using GeoDa1.8.10 [31] (Spatial Data Science Center, University of Chicago, Chicago, Illinois, IL, U.S.A.) and ArcGis 10.3 (Open-Source Geospatial Foundation, Beaverton, Oregon, OR, U.S.A.) software) [32].

This study was approved by the Ethics Committee of the Municipality of Belo Horizonte (CAAE 11,548,913.33001.5140), and by the Federal University of Pelotas (CAAE35798920.8.0000.5317) and follows all ethical principles and current legislation for research involving human subjects. This guarantees the confidentiality of the information and its use only for research.

Results

From 2010 to 2018, there were 325.378 confirmed dengue cases in Belo Horizonte, with 324.044 dengue cases and 1.334 dengue cases with severe signs, with 88 deaths (Incidence Rate (IR) mean/period = 1.52/100.000 inhabitants and Average Lethality Rate (LR)/period = 1.3%). Using GAM, a non-linear regression model for dengue cases, it was possible to observe an increase in the disease occurrence risk from 2010 to 2016 and a decline between 2017 and 2018, with peaks in 2010, 2013, and 2016, with a decline in the two immediately subsequent years, demonstrating the cyclical nature of the disease (Figure 3a). There were 16 deaths in 2010 and 62 deaths in 2016. It is important to note that all peaks observed in TL were due to severe dengue cases (7.7%) (Figure 1).

Figure 3.

Spatial distribution of dengue cases in Belo Horizonte from 2010 to 2018.

Figure 1.

Dengue and severe dengue temporal distribution between 2010 and 2018 in Belo Horizonte, Minas Gerais, Brazil, estimated by Generalized additive models (GAM) with 95% confidence interval (95% CI) (dengue: p < 0.01, R^{2 =} 99.4%; severe dengue: p < 0.01, R^{2 =} 97.9%).

Individuals affected by dengue were mostly females (57.7%; 187,431; IR = 1854.5 cases/100,000 inhabitants), aged 20–30 years (55.7%; 180,958), with incomplete primary education (23.3%; 10703), followed by incomplete secondary education (13.8%; 6,347), and of brown color (51.3%; 36226). Figure 2 shows the distribution of dengue cases according to gender, race, age, and education in the nine health districts of Belo Horizonte. Generally, information about skin color (254,313; 78.2%) and education level (279,418; 85.9%) was not available from the information system.

Figure 2.

Incidence of dengue and severe dengue based on sex, skin color, and age in the nine districts of Belo Horizonte.

The differences in morbidity and mortality rates from dengue among different age groups and genders were evaluated (Table 2). It was found that the IR was higher in adults aged 15 to 19 years, followed by 20 to 30 years for both genders. With the Chi-square test, it was verified that males (p = 0.014, RR = 1.63, 95% CI = 1.07–2.48) and people of older age have a higher risk of dying from dengue in Belo Horizonte (p < 0.05).

Table 2.

Epidemiological indicators of dengue in Belo Horizonte, based on age and sex (2010–2018).

	Dengue cases		Deaths		Popa		IRb		LRb
AGE	F	M	F	M	F	M	F	M	F	M
0–4	3642	3761	0	0	66.522	67.689	684.4	694.5	0.0	0.0
5–9	6634	6893	0	0	71.221	73.647	1164.3	1169.9	0.0	0.0
10–14	12187	12985	0	1	85.153	86.338	1789.0	1880.0	0.0	0.0
15–19	18558	16257	3	1	91.815	90.895	2526.5	2235.7	0.0	0.0
20–30	44162	33960	2	2	232.593	216.947	2373.4	1956.7	0.0	0.0
31–40	34368	23211	2	1	206.008	186.635	2085.4	1554.6	0.0	0.0
41–50	27945	17312	4	5	180.419	153.830	1936.1	1406.7	0.0	0.0
51–60	22031	12550	7	6	148.479	118.388	1854.7	1325.1	0.0	0.0
>60	17800	10394	22	32	180.428	119.144	1233.2	1090.5	0.0	0.0
Total	187431	137218	40	48	1262638	1113513	1854.5	1541.6	0.0	0.0

a Population according to the 2010 census of IBGE. b Average values of IR and LR.

(i) IR = (dengue cases/population) × 100.000.

(ii) LR = (Deaths of dengue/dengue cases) × 100.

F = Female.

M = Male.

Source: Notifiable Diseases Information System (SINAN), Municipal Secretary of Health.

Self-reported yellow skin color individuals (1,097 dengue cases; 18 severe dengue cases) found to be at higher risk of having the most severe form of the disease than the white (RR = 2.2; CI 95% = 1.3–3.5), black (RR = 2.4; CI 95% = 1.4–4.1), brown (RR = 2.4; CI 95% = 1.5–3.6) and indigenous (RR = 5.5; CI 95% = 1.4–41.7) skin colors. According to the statistical analysis using the Chi-square test, the distribution of deaths from dengue was similar among the education categories (p ≥ 0.05). The spatial distribution of the incidence of dengue in the 152 areas within the municipality from 2010 to 2018 is represented in Figure 3. In Table 3, we described the HVI and the disaggregated sanitation, sewage, and race indicators, which were positively correlated to the dengue incidence rate in Belo Horizonte (p < 0.05). Figure 4 represents the findings of the geographic regression of these indices and indicators with the incidence of dengue in different areas of the city of Belo Horizonte. These indicators were observed to be more representative and strongly related to the incidence rate of dengue in the coverage areas located in the north of the city.

Table 3.

Estimation of ordinary least squares of independent variables showing significance with dengue incidence rate.

Independent variables	Constant	Coefficient	P-value	R2
HVI	−0.03	6.85	.001	0.14
Human race	3.72	5.66	.026	0.15
Basic sanitation	1.59	4.25	.030	0.03
General socio-economic	−0.49	5.64	.001	0.15

Figure 4.

Regression maps of the independent variables selected for modeling dengue incidence rates (A: HVI; B: human race; C: sanitation; D: general socioeconomic), using Geographically Weighted regression (GWR) models.

Discussion

Our results suggested a high incidence of dengue cases in Belo Horizonte, with statistically significant emphasis on the most vulnerable populations. According to other authors, dengue outbreaks in Brazil and other countries in the southern hemisphere and their relation to vulnerability indicate the need to invest in public policies that improve the living conditions of low-income populations, in addition to measures for controlling and preventing the disease [29–31].

GAM models are frequently used in association analysis that does not show a linear pattern. Usually, this is the case with the temporal evolution of infectious diseases. Therefore, in this study, the GAM model comprises non-linear effects of time, and dengue incidence allowed better visualization of the temporal dynamics of the disease in Belo Horizonte, based on identifying temporal trends and significant occurrence peaks in the city, primarily in 2010, 2013, and 2016. The Generalized Additive Model (GAM) has been employed by several authors to assess the pattern of dengue. In Paraguay [32],) utilized it to investigate the association between climatic variables and dengue cases. In Costa Rica, the model was used to evaluate the behavior of dengue cases and exhibited a good fit to the data [33]. Furthermore, this technique is also employed to assess the pattern of other diseases, such as leptospirosis [34] and leishmaniasis [35].

A cyclic pattern of epidemic years is a recurring characteristic of dengue. In the Americas, this pattern is strongly marked, as the region experienced epidemics in the years 2010, 2013, 2016, 2019, and 2022 [36]. Furthermore, such a pattern is also observed in Asian countries, such as Vietnam, where 90,844 cases of dengue occurred annually from 2007 to 2016, with epidemics recorded in 2010, 2015, and 2016 [37]. Understanding this behavior is a way to devise prevention and preparedness strategies for epidemic years [38].

The highest rates of dengue were found in self-declared brown skin color women and among young adults, which favors the occurrence of future reinfections, predisposing to severe cases and a possible increase in the fatality rate [17]. The results obtained regarding age and sex are similar to those observed in other studies conducted in Brazil [5,39–42,,,,]. However, in Colombia, a study conducted by [38]identified that the highest incidence rate of dengue was in male children. In Argentina [43], found no correlation between age, sex, and dengue incidence. Meanwhile, in Mexico, the highest incidence was reported in young people aged 15 to 19 years [44]. This difference in the notification profile reinforces that dengue affects the general population and is more associated with the environment than with the individual [45].

When analyzing factors associated with dengue mortality in [35] reported that the fatality rate was higher in men and the elderly. The increase in severe cases may also be associated with diagnostic challenges, as dengue symptoms are similar to other diseases like the flu, and often people are unaware of their initial contact with the virus, thereby increasing the risk of developing severe dengue [29]. Dengue causes social and economic burdens and strains the healthcare system due to the high number of cases. According to [46]arboviruses directly impact the economy, affecting the distribution of resources for health. Therefore, these results highlight the importance of directing prevention programs to the population at higher risk to reduce the impact of this disease on the municipal healthcare system.

Over 30 years ago, the Unified Health System (SUS) was implemented in Brazil, a public system that aims to serve everyone regardless of their age, gender, race, or social status [47,48,]. However, education level and skin color are factors that predispose to social inequalities in the country, becoming a socioeconomic determinant of diseases, particularly infectious ones [49,50,].

It was found that health and socioeconomic vulnerability indices, along with sanitation, sewage, and race indicators, are related to a higher incidence rate of the disease among the different areas within the municipality. Furthermore, it was found that the city’s northern region has a stronger relationship between social vulnerability indices and the incidence of dengue cases over the years. Other studies already performed in Belo Horizonte concluded that the North, Northeast, Northwest, East, and Venda Nova districts comprised the largest number of infectious diseases cases, including leptospirosis, leishmaniasis, and dengue [49,51,52,,]). Similarly [53], and [54]concluded that the incidence of dengue cases was higher in places with high population density and lower income, and in its vicinity when evaluating the spatial distribution of dengue cases in the municipalities of Niterói and Rio de Janeiro, State of Rio de Janeiro.

The increase in vector-borne disease notifications can be attributed to unplanned urban expansion and climate change [55]. Due to the lack of access to services, such as basic sanitation, water supply, and waste collection; as a result, the vector easily finds favorable places to lay its eggs in areas with high levels of social vulnerability. On the other hand, climatic factors, such as humidity and temperature, contribute to the rapid infestation and dissemination of the transmitting agent [56]. It is important to carry out studies assessing social organization in various locations in order to achieve effective vector control; improvements in urban infrastructure and reduction of social inequalities are necessary, as well as the support and participation of the population [41,57,].

It is essential to acknowledge certain limitations of this study so that they can be explored in future research. Although this study observed a correlation between dengue cases and vulnerability factors, we did not investigate the influence of climatic and environmental variables on the occurrence of dengue. Assessing the association between environmental factors in vulnerable areas can significantly contribute to a better understanding of the disease’s behavior.

In this study, we analyzed secondary data reported by health professionals, health services, and the population. Therefore, this case report was subject to the bias of incorrect classification. Dengue cases were notified to the Disease Notification Information System, where they may have been misdiagnosed [58,59,]. An analysis can be weakened by incomplete records and underreporting of cases [17,60,]. It is estimated that around 70% of dengue cases result in mild symptoms or are asymptomatic, so many people do not seek health services, leading to a large number of underreporting cases, which hinders the epidemiological characterization of the disease [61–66]. However, despite the potential notification biases, this data provides extremely valuable information for health organizations as it allows for the analysis of the behavior of various diseases and, therefore, directs efforts and resources toward more effective surveillance and control.

As dengue is a serious health problem in Belo Horizonte, assessing the risk for the disease is a way to help the city’s health system in the development of specific control policies for the population at risk. In this study, data analysis indicated that the disease lethality rate was higher in men, the elderly, and black people, while the incidence rate predominated in women from 20 to 40 years old self-reported yellow skin color individuals. Furthermore, it was found that the HVI, as well as the socioeconomic index and the sanitation, sewage, and race indicators, are positively related to the incidence rate of dengue, suggesting that this index could be useful in planning and prioritizing control actions in Belo Horizonte. In addition, these results reinforce the severity of the disease and the need to develop effective and targeted public policies for populations in high-risk areas to reduce the effects of this disease in the city.

Funding Statement

The work was supported by the CAPES - Coordination for the Improvement of Higher Education Personnel-Brazil [001]; CNPQ - Conselho Nacional de Desenvolvimento Científico e Tecnológico [433418/2018–4].

Disclosure statement

No potential conflict of interest was reported by the author(s).