Spatial Distribution of Dengue in a Brazilian Urban Slum Setting: Role of Socioeconomic Gradient in Disease Risk

Abstract

Background: Few studies of dengue have shown group-level associations between demographic, socioeconomic, or geographic characteristics and the spatial distribution of dengue within small urban areas. This study aimed to examine whether specific characteristics of an urban slum community were associated with the risk of dengue disease.

Methodology/principal findings: From 01/2009 to 12/2010, we conducted enhanced, community-based surveillance in the only public emergency unit in a slum in Salvador, Brazil to identify acute febrile illness (AFI) patients with laboratory evidence of dengue infection. Patient households were geocoded within census tracts (CTs). Demographic, socioeconomic, and geographical data were obtained from the 2010 national census. Associations between CTs characteristics and the spatial risk of both dengue and non-dengue AFI were assessed by Poisson log-normal and conditional auto-regressive models (CAR). We identified 651 (22.0%) dengue cases among 2,962 AFI patients. Estimated risk of symptomatic dengue was 21.3 and 70.2 cases per 10,000 inhabitants in 2009 and 2010, respectively. All the four dengue serotypes were identified, but DENV2 predominated (DENV1: 8.1%; DENV2: 90.7%; DENV3: 0.4%; DENV4: 0.8%). Multivariable CAR regression analysis showed increased dengue risk in CTs with poorer inhabitants (RR: 1.02 for each percent increase in the frequency of families earning ≤1 times the minimum wage; 95% CI: 1.01-1.04), and decreased risk in CTs located farther from the health unit (RR: 0.87 for each 100 meter increase; 95% CI: 0.80-0.94). The same CTs characteristics were also associated with non-dengue AFI risk.

Conclusions/significance: This study highlights the large burden of symptomatic dengue on individuals living in urban slums in Brazil. Lower neighborhood socioeconomic status was independently associated with increased risk of dengue, indicating that within slum communities with high levels of absolute poverty, factors associated with the social gradient influence dengue transmission. In addition, poor geographic access to health services may be a barrier to identifying both dengue and non-dengue AFI cases. Therefore, further spatial studies should account for this potential source of bias.

Fig 1. (A) Pau da Lima study site in Salvador, Brazil and spatial distribution of the estimated risk of (B) dengue and (C) non-dengue acute febrile illness (AFI) in the 98 census tracts that comprise the study site.

Risks (per 10,000 population) were estimated for the two-year study period from January 1, 2009 and December 31, 2010.

Fig 2. Enrollment of acute febrile illness (AFI) patients and dengue detection through enhanced surveillance in the Pau da Lima community, Salvador, Brazil, from January 1, 2009 to December 31, 2010.

Fig 3. Standardized morbidity ratios (SMRs) for dengue and non-dengue acute febrile illness (AFI) in the Pau da Lima community, Salvador, Brazil, from January 1, 2009 to December 31, 2010.

Non-adjusted SMR for (A) dengue and (B) non-dengue AFI; SMR adjusted by the final Poisson-log normal model for (C) dengue and (D) non-dengue AFI; SMR adjusted by the final conditional auto-regressive model for (E) dengue and (F) non-dengue AFI.