Prediction of Influenza-Like Illness Incidence in Cheras, Malaysia based on Environmental Data using the Generalised Additive Model

Abstract

Influenza-like illness (ILI) accounts for 500,000 fatalities worldwide yearly. Environmental factors contribute to spread of respiratory infections. This study describes the epidemiological trends of ILI in Cheras, Kuala Lumpur, Malaysia and seeks scientific evidence on ILI activity and environmental factors. Cheras Health District collected ILI surveillance data and the nearest weather station which provided environmental data on mean temperature, humidity, wind speed, cumulative rainfall, particulate matter (PM)10 and PM2.5 levels. A total of 51,245 ILI cases were reported from 1st December 2021 to 30th April 2023. The non-linear relationship between ILI incidence and environmental variables was modelled using the Generalised Additive Model (GAM). The lag time for mean temperature, mean humidity, cumulative rainfall, mean wind speed, PM 10 level, and PM 2.5 level was 2, 3, 3, 3, 0, and 3 days, respectively. The model with the optimal lag better describes ILI case variance (R2=0.5, explained deviance=58%) than the model without lag selection (R2=0.5, explained deviance=57.2%). The Lag Model indicates a significant p-value for PM10 but no significant concurvity between predictor variables. Thus, the final model (R2=0.5, explained deviance=59.2%) has k=15. Higher rainfall, relative humidity, colder temperature, and decreased wind speed increased ILI incidence. PM2.5 and PM10 also contributes to ILI.