A nationwide joint spatial modelling of simultaneous epidemics of dengue, chikungunya, and Zika in Colombia

Abstract

Background: Chikungunya, and Zika emerged in the 2010s in the Americas, causing simultaneous epidemics with dengue. However, little is known of these Aedes-borne diseases (ABDs) joint patterns and contributors at the population-level.

Methods: We applied a novel Poisson-multinomial spatial model to the registered cases of dengue (n = 291,820), chikungunya (n = 75,913), and Zika (n = 72,031) by municipality in Colombia, 2014-2016. This model estimates the relative risk of total ABDs cases and associated factors, and, simultaneously, the odds of presence and contributors of each disease using dengue as a baseline category. This approach allows us to identify combined characteristics of ABDs, since they are transmitted by the same mosquitoes, while also identifying differences between them.

Results: We found an increased ABDs risk in valleys and south of the Andes, the Caribbean coast, and borders, with temperature as the main contributor (Relative Risk 2.32, 95% Credible Interval, CrI, 2.05-2.64). Generally, dengue presence was the most probable among the ABDs, although that of Zika was greater on Caribbean islands. Chikungunya and Zika were more likely present than dengue in municipalities with less vegetation (Odds Ratio, OR, 0.75, 95%CrI 0.65-0.86, and 0.85, 95%CrI 0.74-0.99, respectively). Chikungunya tended to be present in more socially vulnerable areas than dengue (OR 1.20, 95%CrI 0.99-1.44) and Zika (OR 1.19, 95%CrI 0.95-1.48).

Conclusions: Important differences between the ABDs were identified and can help guide local and context-specific interventions, such as those aimed at preventing cases importation in border and tourism locations and reducing chikungunya burden in socially vulnerable regions.

Fig. 1

Aedes-borne diseases (dengue, chikungunya, and Zika) total incidence (per 100,000 inhabitants) and for each disease by municipality, Colombia, EW 01/2014 to 39/2016

Fig. 2

Posterior mean (A) of the relative risk (RR) for the total cases of Aedes-borne diseases (dengue, chikungunya, and Zika) and (B) of probabilities of presence of each disease, given the total, by municipality, Colombia, epidemiological week 01/2014 to 39/2016. The solid circle and the solid triangle are the locations of two municipalities (Puerto Gaitán and Cumaribo, respectively) to serve as examples of the results interpretation in the main text

Fig. 3

Posterior mean of the shared spatial effect for Aedes-borne diseases (dengue, chikungunya, and Zika) (A) and of the latent effects associated with the total cases (equation [2]) (B) and with the presence of chikungunya (C) or Zika (D) compared to dengue (equation [4]) by municipality in Colombia, epidemiological week 01/2014 to 39/2016. Municipalities for which the 95% credible interval of the effect included 0 are shown in light grey

Fig. 4

Posterior distribution of the relative risks associated with each of the covariates and the total number of cases of Aedes-borne diseases (dengue, chikungunya, and Zika), and of the odds ratios associated with each of the covariates and the odds of a municipality having chikungunya or Zika in comparison to dengue, and Zika in comparison to Chikungunya. NDVI = Normalized Difference Vegetation Index, HC = Healthcare