The study area is a geographically continuous area comprising 50 villages in the Muzaffarpur District of Bihar State, India, a district where VL is highly endemic. The 50 villages can be further subdivided into 200 hamlets, also known as tolla. We conducted 3 annual surveys in September and October of 2008, 2009, and 2010. In each survey, we visited all households in the study area, collected demographic information, and asked whether the house had been covered by indoor residual insecticide spraying in the year preceding the survey. In each survey, we also collected information about VL in the household since the previous survey. For the first survey, we used a recall period of 1.5 years. A case of VL was defined as the combination of a clinical history typical for VL (fever of >2 weeks’ duration, lack of response to antimalarial drug treatment); a positive result by the rK39 rapid diagnostic test; and a good response to specific VL treatment, with or without confirmation of parasites. Each case reported was verified from medical records by a study physician. At the time of the second survey in 2009, we also collected information about assets owned by each household, including domestic animals, and we recorded characteristics of the structure of the house and the surrounding vegetation. Information about assets owned (other than domestic animals) was used to subdivide the study population into 5 quintiles of socioeconomic status. To study potential associations with household environment, remote sensing (15-m resolution ASTER images), and ground data were combined to derive a map of the study area showing 7 types of land cover. Data were analyzed in a binomial multilevel model with tolla as a random effect.

We enrolled a study population of 81,210 persons, divided over 13,416 households (92% of all households in the study area). During the study period, we registered 207 VL cases, equivalent to an average annual incidence of 72.8/100,000 population. Cases were strongly clustered at the tolla level, with an intraclass correlation of 32%. None of the types of land cover was significantly associated with disease. With the villages spread out along roads and farming land split up into small plots, the environment experienced by persons from different villages or from different locations did not vary greatly.

VL was strongly associated with age; the odds of having VL was lowest for children <5 years of age and highest for children 5–14 years of age (odds ratio [OR] 2.5, 95% CI 1.5–4.0). Higher socioeconomic status was associated with reduced risk; comparing the wealthiest to the poorest quintile, we observed an OR of 0.5 (95% CI 0.3−1.0). Having at least 1 bed net per 3 household members was protective on univariate analysis, but the effect was weaker and no longer statistically significant when we controlled for confounding (OR 0.8, 95% CI 0.5–1.3). This finding can be explained by a strong association between socioeconomic status and ownership of bed nets.

Of ownership of all the domestic animals investigated, only ownership of goats was weakly, but significantly, associated with VL (OR 1.4, 95% CI 1.0–1.8). Other factors at household level that were statistically significant in multivariate analysis were the following: belonging to the Musahar caste (OR 2.9, 95% CI 1.3–6.8); presence of a bamboo tree (OR 1.5, 95% CI 1.2–2.0); and type of walls (OR 1.8, 95% CI 1.0–3.3 for unplastered brick walls and OR 2.5, 95% CI 1.3–4.6 for thatched walls, with plastered brick walls as reference for both). (Table) Thatched walls and presence of bamboo trees are likely to provide favorable breeding conditions for the sand fly vector (13).

Indoor residual insecticide spraying coverage was poor. In 2009 (the last year for which data were collected for the full year), only 12% of all households had reportedly been sprayed at least once.

In this large cohort study, controlled for potential confounding by socioeconomic status and other contextual factors, we identified several factors associated with VL. Ownership of goats and presence of bamboo trees near the house are risk factors, but are not strong enough to warrant specific interventions. Poor housing is a stronger risk factor; thus, housing plans launched by the Indian government may positively affect control of VL. Persons in the Musahar caste were at increased risk; they made up 2.4% of the study population but had >15% of VL cases. The Musahars are known to be among the poorest of the poor, but even after we controlled for confounding by socioeconomic status, the association remained statistically significant. Some residual confounding cannot be ruled out, but other factors probably play a role. One such factor could be long delays in seeking health care by Musahars, which was documented in another recent study (14). When devising improved VL control strategies, it would certainly be justified to pay special attention to the hamlets inhabited by the Musahar caste. Overall, however, the most benefit can be expected from strengthening vector control efforts. In the 1960s, as a byproduct of intensive indoor residual insecticide spraying for malaria eradication campaigns, VL was all but eliminated from the area, and biannual indoor residual insecticide spraying is one of the cornerstones of the regional VL elimination strategy (15). Thus, it defies imagination that in this highly VL-endemic area, the annual indoor residual insecticide spraying coverage can be as low as 12%.