Modeling snakebite risk in Sri Lankan community

Abstract

Background and rationale Snakebite is considered as a neglected tropical disease (WHO, n.d.). Sri Lanka has one of the highest snakebite incidence (SBI) rates in the world and according to hospital statistics about 37,000 patients are admitted to government hospitals annually as a result of snakebite (Kasturiratne et al., 2005; Kasturiratne et al., 2008). Incidence data are usually modeled with the use of generalized linear models. The aim of the present study is to develop a snakebite risk model for Sri Lanka. Methodology Incidence data was obtained from “National Snakebite Study”. Generalized additive and linear models were considered to model snakebite incidence. Individual-level variables namely gender, age, ethnicity, religion, income, education and employment were considered as explanatory variables. The goodness of fit statistics and standardized residuals of the fitted model were used to assess the model fit. Empirical variogram was calculated on standardized residuals to determine spatial dependence. Statistical analysis used the R programming language. Results In the fitted generalized linear model, age was considered in quadratic form; gender, employment, ethnicity, and income were considered as factor variables, with significant interaction was noted between ethnicity and income groups. The goodness of fit statistics showed fitted model is adequate to represent snakebite incidence data. According to the fitted model, highest snakebite risk groups were 30 to 59 years age group, males, field workers, low-income non-Sinhalese and Sinhalese compared to middle-income and high-income non-Sinhalese. Map of the standardized residuals of the fitted model, at each sampled locations showed, visually identifiable geographical pattern in the standardized residuals of the fitted model. Empirical variogram calculation on the standardized residuals showed the presence of spatial dependence. Discussion and conclusions Although the fitted model was good enough to represent snakebite incidence data, underline geographical distribution of snakebite incidence was not represented. Further spatial analysis is needed to be done to model snakebite incidence in Sri Lanka community.