Analysis of the spatial distribution of cases of Zika virus infection and congenital Zika virus syndrome in a state in the southeastern region of Brazil: Sociodemographic factors and implications for public health

Abstract Objective To perform spatial distribution analysis of reported cases of Zika virus and congenital Zika syndrome (CZS) in the state of Espírito Santo, Brazil, by neighborhood, and relate the results to sociodemographic indicators and implications for the health process. Methods An ecological study using data from the 2016 National Notifiable Diseases Surveillance System, epidemiological records, and information on neighborhoods of families confirmed with CZS from qualitative field research. Results Sociodemographic indicators were analyzed in three distinct groups: general population with Zika virus, pregnant women with Zika virus, and cases of CZS visited. For the three groups, average literacy rates were 71.1%, 71.0%, and 68.3%; the average income per minimum wage was 1.4, 1.1, and 1.4; sanitary sewage coverage was 75.6%, 76.1%, and 71.4%; garbage coverage was 90.8%, 91.2%, and 89.2%; and water supply was 93.8%, 94.1% and 93.8%, respectively. Socioeconomic indicators showed no significant differences between groups, although they were above the national average. A nonsignificant variation of 68.3%–71.1% was seen in the average literacy level above 15 years of age. Conclusion Socioeconomic and demographic indicators of cases of Zika virus infection and CZS may indicate that the outbreak had different impacts according to class, social group, or gender, reflecting the persistence and social geography of inequality in Brazil.

the epicenter of an epidemic, with confirmed exposure in all regions of the country. 9 Since then, Zika virus infections have been reported in epidemic proportions, affecting more than 80 countries and subsequently causing a pandemic. 10,11 Countries most affected were those that had never had circulation of Zika virus (and thus the population was susceptible to infections), were endemic for dengue (Ae. aegypti was found), and contained areas with conditions often associated with arbovirus transmission (e.g. high population density, ideal climatological conditions, lack of infrastructure). It has also been speculated that extreme climatic conditions may have been a contributor to Zika virus spread. 12 Zika virus symptoms (similar to dengue) include fever, skin rash, conjunctivitis, muscle and joint pain, and malaise or headache, with a duration of 2-7 days. In Brazil, Zika virus infection was marked by a finding never before described in the scientific literature: fetal microcephaly. 13 The perception of the increase in the number of microcephaly cases in severe malnutrition; or exposure to harmful substances (alcohol, certain drugs, or toxic substances). 16 Microcephaly was found to be just one of the manifestations of an anomaly called congenital Zika syndrome (CZS). 17,18 Conditions associated with CZS include eye damage, joint problems, excessive muscle tone, and seizures, among other signs and symptoms. 1,19 In view of the harmful consequence of microcephaly, in November In February 2016, Zika virus infection became a disease of mandatory reporting, leading to increased investment in research on the disease. 1,23,24 Following improvements in epidemiological surveillance, research on the association between Zika virus infection and dissemination patterns of the disease has become more robust. 25 In addition, areas were affected in an unequal way, with a higher incidence of CZS infection in places of social vulnerability.
Accordingly, important government measures were implemented to address issues surrounding poverty and inequity to access, namely the supply of insect repellent to pregnant women in the public service and monthly payment of the Continuous Cash Benefit, corresponding to approximately 1-month earnings at minimum wage (US$ 247.96) for children diagnosed with CZS. 26,27 According to the Ministry of Health, between epidemiological The United Nations Food and Agriculture Organization (FAO) maps critical zones and groups more prone to virus infection, and maps zones as they are likely to control infection. FAO can also make use of models with meteorological, socioeconomic, environmental information and data, and provide an efficient and intersectoral response to fight infection. 29,30 Mapping of diseases helps to describe their spatial behavior, creating clues for the development of public practices and policies focused on health care and surveillance. The maps also allow correlation of various factors that are present in the territory such as infrastructure, living conditions, and inequalities. Analysis of the spatial pattern of Zika virus makes it possible to identify the most affected areas and the main social determinants of the disease. 31 Information gaps for the country, and the state of Espírito Santo in particular, remain for the Zika virus epidemic and socio-environmental factors. The aim of the present study was to describe the spatial distribution of Zika virus and CZS cases in the state of Espírito Santo according to neighborhoods, and to describe the sociodemographic indicators and their implications in the health-disease process.

| MATERIALS AND METHODS
We conducted an ecological study with a descriptive approach using secondary data from the National Notifiable Diseases Surveillance Regarding mapping of cases on a national scale, Zika virus notification data in the general population and in pregnant women were obtained, spatially aggregated per unit of the federation, from the SINAN database. For analysis at a regional scale, Zika virus cases were mapped in the general population using the rate per 100 000 inhabitants, and in the case of pregnant women, absolute numbers were used due to the absence of a denominator for the calculation of rates; the figures were spatially aggregated per neighborhood. The neighborhood of residence field was used as a spatial aggregation unit for the reported cases of Zika virus infection. This unit was available on the basis of the most disaggregated and was an intermediary between the municipality and the place of residence.
ArcGIS (ESRI, Redlands, CA, USA) software was used for georeferenced data, with the federal unities' cartographical bases, municipalities, and census tracts available from the Brazilian Institute of Geography and Statistics (IBGE). The neighborhoods were obtained from the Jones of Santos Neves Institute, a state-level research institute. The reported cases were added to the base in a georeferenced environment, distinguishing the cases of pregnant women from the others. This spatial unit aggregation allows a higher level of scale refinement than municipality, indicating possible spatial clusters within municipalities. Sociodemographic variables such as general water supply network, garbage collection service, sanitary coverage through general sewage network, total monthly income, and literacy over 15 years of age were obtained from the 2010 census. We used the geometric growth method to correct the population (from 2010 to 2016), applying the verified growth of the municipalities in the neighborhoods. 33 For this aggregation unit, there were no datasets available with the same territorial delimitation, except for census tract data (grouping of 300 households on average), which did not hold exact geographic correspondence with the neighborhoods. There were cases where some census tracts crossed two or more neighborhoods, and in other places there was a total absence of census tracts in some areas of the neighborhoods. To adjust this situation and minimize data loss we added in a geographical information system (GIS) layer of spatial information containing sociodemographic variables in a raster format. This kind of file format is a data matrix in which data are stored in each pixel with a determined size (smaller than a census tract), allowing topological and spatial operations with fewer errors. In this operation we divided a census tract area into smaller pieces (20 × 20 meters), which allowed us to recompose the socioeconomic data in the neighborhoods, merging the smaller pieces of data using means and standard deviations.
For the total population, however, the sectors were proportionally added in the neighborhoods by means of operation between layers with zonal statistics also using the raster. The overall incidence rate was calculated with the estimated population by neighborhoods.
CZS cases in Espírito Santo were georeferenced by municipalities due to the incompleteness of the information in the neighborhood field (about 93%). However, it was possible to map CZS cases by neighborhood that were visited during the field research cited above.
The sociodemographic information of the neighborhoods was collected for these interviewees, using the census variables.
Among the total number of reports, 342 did not indicate gender.  (Fig. 1).
The spatial distribution of pregnant women was mainly concentrated in the states of Rio de Janeiro (n=5722, 23.7%), Mato Grosso (n=1778, 7.4%), Amazonas (n=1293, 5.4%), and Tocantins (n=428, 1.8%) (Fig. 1)  Of these 49 confirmed cases of CZS, 10 died (Fig. 4). A total of four deaths occurred in the city of Aracruz, two in Cariacica, and one case in each of the municipalities of Cachoeiro de Itapemirim, Serra, Viana, and Vila Velha.   The results of the present study also demonstrate the predomi-  considered. 30 Health professionals should expand information to both men and women, thus avoiding gender inequality in responsibility for combating the epidemic.

| DISCUSSION
A limitation of the study is that the results do not consider the social heterogeneity of the neighborhoods given the use of averages for comparison. Due to the lack of spatial adequacy between layers (census tract and neighborhoods) we had to adapt the data per neighborhood that were homogenized using averages, especially for population. In addition, the transformation of census tract information into neighborhoods can have an effect known as the "modifiable areal unit problem" (MAUP) whereas the neighborhood boundaries can distort the census data.
Another important limitation concerns the demographic data for race/ color and education, which had a high proportion of nonresponses. These data suggest that the underprivileged may be more susceptible to exposure to the virus, which can help to trigger action to promote public policies that support the reduction of social and gender inequalities, such as investment in education for vulnerable populations, as well as promotion of health education, employment opportunities, and income for women. In addition, healthy cities should be focused on urban planning and environmental sanitation, particularly regarding access to clean water, garbage collection, and sanitation.
Large intersectoral investments aimed at improving the living condi-

CONFLICTS OF INTEREST
The authors have no conflicts of interest.