Spatial distribution of the relative risk of Zika virus disease in Colombia during the 2015–2016 epidemic from a Bayesian approach
MetadataShow full item record
Knowledge of the Zika virus was scarce in many high-resource countries until an epidemic began in Brazil in 2015.1 A public health emergency of international importance was declared in the Americas after 48 countries had confirmed autochthonous cases of Zika by vector transmission; furthermore, the Emergency Committee of the World Health Organization (WHO) reported cases of microcephaly and other neurological disorders in disease endemic areas.2 In Colombia, information provided by the National Public Health Surveillance System (SIVIGILA) through the epidemiological bulletin of the Colombian National Institute of Health (NIH) reported that, since the disease was first identified in epidemiological week 29 of 2015 up to the endemic phase, close to 105 000 symptomatic cases of Zika virus infection were identified, including more than 19 000 among pregnant women. The incidence reported during the epidemic phase was of 377.7 cases per 100 000 inhabitants.3 Several studies have been carried out to investigate the causes and behavior of Zika virus disease in Colombia. Pacheco et al.4 estimated the risk of the disease in the country and found that most provinces had an accumulated incidence of between 0.1 and 129.7 per 100 000 habitants. Rojas et al.5 analyzed data to assess the incidence of the virus in Girardot and the island of San Andrés, and estimated attack rates of 18.43 and 12.13 per 1000 inhabitants, respectively. Although the results of previous studies provide important information on the use of crude data, they show great variability and consider spatial relationships between observations only according to the geographical limits of each province. These studies do not consider the influence of other factors, such as settlement size, size of neighboring settlements, and the relationship between the risk area and neighboring areas, which can all affect the spatial and temporal risk of contracting the Zika virus. Health data commonly consist of aggregated counts of disease within administrative units (small areas) such as departments and municipalities. To estimate the risk of disease, maximum likelihood estimators are typically used, such as standardized morbidity rate (SMR). These rates are variable because they depend on the expected values, which in turn depend on the size of the population. When mapping the geographical distribution of a disease, the aim is to discover spatial patterns that help explain behavior and enable hypotheses about its etiology. The present study used Bayesian smoothing methods to estimate risk to spatially review the geographical structures of disease behavior. A wide range of models in disease mapping have been developed to offer appropriate relative risk estimates. Taking into account area information, these models provide smoothed risk maps and improve the estimates. One of the most important studies in risk estimation is that of Besag et al.6 Risks are estimated using a model that captures the risk structure through incorporation of information from neighboring areas. The aim is to identify any spatial relationship that reveals the behavior of the disease, how it is distributed, and whether different factors explain this behavior. Besag et al.6 proposed a Bayesian hierarchical model that models risk, incorporating two random factors: one that explains the spatial dependence between neighboring areas and one that explains the residual effects. Some extensions and contributions to the Besag model are the works of Bernardinelli et al.,7 Best et al.,8 and Lawson.9, 10 The aim of the present study was to determine the spatial distribution of the risk of Zika virus disease in each region of Colombia during the 2015–2016 epidemic using a Bayesian hierarchical model.