Quantifying Climatic and Socio-Economic Influences on Urban Malaria in Surat, India: A Modelling Study

BackgroundCities are becoming increasingly important habitats for mosquito-borne infections. The pronounced heterogeneity of urban landscapes challenges our understanding of the spatio-temporal dynamics of these diseases, and of the influence of climate and socio-economic factors at different spatial scales. Here, we quantify this joint influence on malaria risk by taking advantage of an extensive dataset in both space and time for reported Plasmodium falciparum cases in the city of Surat, Northwest India. MethodsWe analyzed 10 years of monthly falciparum cases resolved at three nested spatial resolutions (for 7 zones, 32 units and 478 workers units subdivisions, respectively). With a Bayesian hierarchical mixed model that incorporates effects of population density, poverty, humidity and temperature, we investigate the main drivers of spatio-temporal malaria risk at the intermediate scale of districts. The significance of covariates and the model fit is then examined at lower and higher resolutions. FindingsThe spatial variation of urban malaria cases is strongly stationary in time, whereby locations exhibiting high and low yearly cases remain largely consistent across years. Local socio-economic variation can be summarized with two main principal components, representing poverty and population density respectively. The model that incorporates these two factors together with local temperature and global relative humidity, best explains monthly malaria patterns at the intermediate resolution. The effects of local temperature and population density remain significant at the finest spatial scale. We further identify the specific areas where such increased resolution improves model fit. InterpretationMalaria risk patterns within the city are largely driven by fixed spatial structures, highlighting the key role of local climate conditions and social inequality. As a result, malaria elimination efforts in the Indian subcontinent can benefit from identifying, predicting and targeting disease hotspots within cities. Spatio-temporal statistical models for the mesoscale of administrative units can inform control efforts, and be complemented with bespoke plans in the identified areas where finer scale data could be of value. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSUrban areas have become the new dominant ecosystem around the globe. Developing countries comprise the most urbanized regions of the world, with 80% of their population living in cities and an expected increase to 90% by 2050. The large and heterogeneous environments of today challenge the understanding and control of infectious disease dynamics, including of those transmitted by vectors. Malaria in the Indian subcontinent has an important urban component given the existence of a truly urban mosquito vector Anopheles stephensi. A literature search in Mendeley of "urban malaria" and "India" returned 161 publications, in their majority on diagnostics or brief reports on the disease, and on cross-sectional rather than longitudinal studies addressing the spatio-temporal variation of disease risk for a whole city, the subject of our work. A relevant exception is a study for the city of Ahmedabad; this not address multiple seasons across different spatial scales, and climatic conditions are not considered jointly with socio-economic drivers in the modeling. A second Mendeley search on A. stephensi returned 11 publications into two distinct groups: early entomological studies for India and recent reports of the mosquito in the Horn of Africa. This geographical expansion makes the specter of urban malaria a future possibility for the African continent where the disease remains so far rural and peri-urban. Added value of this studyThis paper relies on an extensive surveillance data set of Plasmodium falciparum cases for Surat (India) to investigate the variation and drivers of malaria risk in an heterogenous urban environment. A statistical model for the spatio-temporal variability of cases is developed, which includes both climatic and socio-economic drivers, with the latter summarized into two major axes of variation. Model fits are compared across three spatial resolutions, ranging from a few zones to a few hundred units. Seasonal hotspots are shown to be largely stationary in time, which allows identification of dominant drivers, including population density and local temperatures, whereas humidity acts globally modulating year-to-year burden. More granular statistical models and datasets like the one analyzed here are needed to capture the effects of socioeconomic and climatic drivers, and to predict current and future malaria incidence patterns within cities. Implications of all the available evidenceThe analysis identifies relevant resolution which can vary across the city for targeted intervention, including vector control, that would focus on reducing and eliminating transmission hotspots. The modeling framework, incorporating predictors representing climate at local vs. aggregate levels, and major axes of socio-economic variation, should apply to other vector-borne diseases and other cities for which surveillance records are available. The importance of spatially-explicit and sustained surveillance data for informing these models cannot be overstated.