Abstract
Leishmaniasis is a vector-borne disease of which the transmission is highly influenced by climatic factors, whereas the nature and magnitude differ between geographical regions. The effects of climatic variables on leishmaniasis in Sri Lanka are poorly investigated. The present study focused on time-series analysis of leishmaniasis cases reported from Sri Lanka with selected climatic variables. Variance stabilized time series of leishmaniasis patients of entire Sri Lanka and major districts from 2014 to 2018 was fitted to autoregressive integrated moving average (ARIMA) models. All the possible models were generated by assigning different values for autoregression and moving average terms using a function written in R statistical program. The top ten models with the lowest Akaike information criterion (AIC) values were selected by writing another function. These models were further evaluated using RMSE and MAPE error parameters to select the optimal model for each area. Cross-autocorrelation analyses were performed to assess the associations between climate and the leishmaniasis incidence. Most associated lags of each variable were integrated into the optimal models to determine the true effects imposed. The optimal models varied depending on the area. SARIMA (0,1,1) (1,0,0)(12) was optimal for the country level. All the forecasts were within the 95% confidence intervals. Humidity was the most notable factor associated with leishmaniasis, which could be attributed to the positive impacts on sand fly activity. Rainfall showed a negative impact probably as a result of flooding of sand fly larval habitats. The ARIMA-based models performed well for the prediction of leishmaniasis in the short term.