Environmental and socio-demographic determinants of severe malaria risk in urban Kisumu, Kenya.

Abstract

Urban malaria is understudied, although it represents a growing problem in urbanizing sub-Saharan Africa. Research is particularly needed on risk factors for severe malaria and on standardized quantitative classification systems for urban areas. Accordingly, we conducted a study on risk factors for severe malarial anemia (SMA) in Kisumu, Kenya, a city undergoing urban population growth typical of many sub-Saharan African cities. First, we developed a novel map-based strategy to sample population and environment simultaneously and representatively, and interviewed 4,336 caregivers of children <10 years of age with regard to their knowledge, attitudes and practices (KAP) related to malaria. Concurrently, we identified and geolocated the households of 80 children presenting at Kisumu District Hospital with SMA and interviewed their caregivers. Environmental and demographic information was used to classify the study area into urban, peri-urban and semi-rural zones using a principal components analysis and a k-means clustering algorithm, demonstrating KAP variations among these zones. The clustered classification identified such variation better than two traditionally-used classifications and suggested that peri-urban areas are more similar to semi-rural than to urban areas. Using a spatially-constrained clustering algorithm, we further classified the study area into eleven neighborhoods, demonstrating intra-urban variation in incidence of self-reported malaria and SMA. Risk factors for SMA were assessed via an unmatched case-control study, comparing cases with 826 healthy community controls from among KAP survey respondents. In bivariate analyses, household size, length of residence, house construction, caregiver education, land ownership, travel, use of mosquito coils, distance to hospital, presence of domestic animals, and NDVI and other local environmental characteristics were significantly associated with SMA. Multivariate analysis confirmed recent travel to a rural area was an extremely significant risk factor (OR = 9.33, p < 0.0001). Unexpectedly, use of bednets or ITNs was not related to case status. We further examined neighborhood-level associations between environmental or socio-demographic risk factors and malaria incidence or SMA. While several factors were correlated with self-reported malaria, none were associated with SMA, nor were the distributions of the two outcomes intercorrelated, potentially indicating different etiologies. Results of demonstrate the growing importance of urban malaria and suggest intervention strategies.