CIENCIASMEDICASNEWS: Ahead of Print - Joint Estimation of Relative Risk for Dengue and Zika Infections, Colombia, 2015–2016 - Volume 25, Number 6—June 2019 - Emerging Infectious Diseases journal

We jointly estimated relative risk for dengue and Zika virus disease (Zika) in Colombia, establishing the spatial association between them at the department and city levels for October 2015–December 2016. Cases of dengue and Zika were allocated to the 87 municipalities of 1 department and the 293 census sections of 1 city in Colombia. We fitted 8 hierarchical Bayesian Poisson joint models of relative risk for dengue and Zika, including area- and disease-specific random effects accounting for several spatial patterns of disease risk (clustered or uncorrelated heterogeneity) within and between both diseases. Most of the dengue and Zika high-risk municipalities varied in their risk distribution; those for Zika were in the northern part of the department and dengue in the southern to northeastern parts. At city level, spatially clustered patterns of dengue high-risk census sections indicated Zika high-risk areas. This information can be used to inform public health decision making.

Dengue and Zika virus disease (hereafter referred to as Zika) are infectious diseases caused by arboviruses in the family Flaviviridae. Dengue virus has 4 serotypes (1–4); serotypes 2 and 3 are associated with severe disease in persons with second dengue infections. Zika virus infection is associated with congenital malformations in babies born from women infected during pregnancy and with Guillain-Barré syndrome in infected adults (1).

Colombia is highly affected by vectorborne diseases. Villar et al. (2) reviewed the dengue epidemic in this country for 2000–2010 and reported an increasing epidemic trend for the period; outbreaks occurred in 2001, 2003, and 2010. In 2016, health authorities in Colombia reported >101,016 dengue cases that resulted in 289 deaths (3) and 9,799 Zika cases that were laboratory confirmed and 96,860 suspected Zika cases diagnosed by clinical signs (4).

For this study, we concentrated on the spatial patterns assessment of risk for dengue and Zika; in particular, we focused on the relative risk (RR) estimation for areal data by using hierarchical Bayesian models for these infections. RR represents the excess (or lack) of risk in a small area compared with the background risk. RR is mostly based on models and supported by Bayesian estimation methods (5). We used the following as study regions: the municipalities in the department of Santander, Colombia (1 of the departments where incidence of dengue and Zika for 2015–2016 was highest) and the city census sections belonging to the capital city of Santander (1 of the cities most affected by dengue and Zika for the same period).

Racloz et al. (6) and Louis et al. (7) reviewed the spatial patterns assessment of dengue risk; specifically for RR estimation of dengue, Ferreira and Schmidt (8) and Martínez-Bello et al. (9) estimated RR for dengue on a local spatial scale; and Restrepo et al. (10) and Martínez-Bello et al. (11) applied methods for the spatiotemporal assessment of dengue risk. Examples for the spatial patterns assessment of Zika risk run from merely descriptive methods to model-based approaches. For instance, in Colombia, descriptive risk maps associating Zika incidence rates with environmental and sociodemographic factors have been produced in the departments of Sucre (12), Tolima (13), Guajira (14), Santander, and Norte de Santander (15) and in the city of Pereira in the department of Risaralda (16). Model-based spatial patterns assessment of Zika risk have been developed for the 33 departments in Colombia by using Poisson models for the RR for Zika (17). The distribution of risk for Zika transmission among counties/districts in Guangdong Province, China, was assessed by using analytic hierarchy process models (18).

Zika, dengue, and chikungunya are also jointly studied by using the spatial patterns assessment of risk because the viruses share similar transmission routes (Aedes mosquitoes). On a large scale in Brazil, ecological studies have explored the risk factors for unusual spatial patterns of microcephaly, including dengue, Zika, and chikungunya data (19). Also estimated is the potential spatial risk for Zika and chikungunya according to socioenvironmental factors, estimating the size of the populations at risk for both diseases (20). On a small geographic scale, the risk factors for cocirculating arboviruses (dengue, Zika, chikungunya) at the community level have been evaluated (21).

In Colombia, Krystosik et al. (22) generated city-level risk maps of chikungunya, Zika, and dengue supporting vector-control strategies; Martínez-Bello et al. (23) estimated the RR for dengue and Zika by using spatiotemporal interaction effects models for 1 department and 1 city in Colombia. Riou et al. (24) assessed the spatial patterns of risk for the 2013 Zika and chikungunya outbreaks in the French Polynesia islands, and Funk et al. (25) jointly modeled Zika and dengue time series data from the Zika outbreak in the Yap Island in the Pacific Ocean.

Our aim with this study was to jointly estimate the disease- and area-specific RRs for dengue and Zika by using hierarchical Bayesian joint models accounting for the spatial association between both diseases. We used data from the 2015–2016 Zika outbreak in Colombia and analyzed 2 levels of spatial data aggregation: the department level (disease counts aggregated per municipality) in the department of Santander and the city level (disease counts aggregated per census section) in the city of Bucaramanga (Santander).