Effect of climate factors evaluation on hand-foot-and-mouth disease in malaysia using generalized models

Abstract

Hand, foot, and mouth disease, also known as HFMD, affects millions of people worldwide and has been a significant public health issue as well as a substantial burden since 1969. According to a conclusive evidence, the combined impact of rapid population growth, climate change, socioeconomic, and other changes in living ways are the critical determinants of this disease. In Asian countries, this disease has marked an increasing trend of outbreaks leading to fatal complications since the late 1990s. This disease exhibits a cyclic seasonal pattern that can be clarified by climate. Numerous experts from various countries revealed that climate factors play a critical role in predicting HFMD cases. However, the findings vary across countries, implying that the climate of each country has a notable influence on HFMD incidence. In Malaysia, there is indeed a paucity of research and empirical evidence related to these issues. Hence, this study aims to examine the behaviour of the HFMD series in Malaysia and its association with climate factors. Various statistical modelling approaches have been extensively used in previous studies, for example, the Generalized Linear Model (GLM) and the Generalized Additive Model (GAM). However, GLM often fails to account for the non-linearity effect of the variables. In the meantime, GAM demands that each observation be independently distributed. This assumption can be violated when dealing with time-series data comprising clustered variables, such as HFMD data with clustered states. A flexible statistical approach is needed to better understand and interpret the incidence of HFMD and climate change phenomena. Following that, this research proposed a new framework by incorporating a mixed effect into GAM and autoregressive terms, named a generalized additive mixed model (GAMM) with autoregressive terms. Apart from that, the issue of overdispersion was taken into consideration as the type of count data, such as HFMD cases, often displays an overdispersion problem. This is due to the high variability in the data sets. Therefore, to solve this issue, a GAMM Negative Binomial with autoregressive terms is proposed in this study. Besides, a rolling basis cross-validation approach was used to validate the best model between the GLM, GAM, and GAMM, with and without autoregressive terms, in describing the association between HFMD incidence and climate factors in Malaysia. This study found that the HFMD incidence in Malaysia increased during the inter-monsoon and southwest monsoon seasons but decreased during the northeast monsoon. The cross-validation findings imply that the proposed model, GAMM Negative Binomial with autoregressive terms, is the best model to describe the impact of climate factors on HFMD incidence in Malaysia. The model demonstrated that the risk of HFMD increased in the following two weeks with a rainfall of less than 60 mm and decreased with more than 60 mm. The risk of HFMD is decreased when wind speed at a two-week lag is less than 3.5 m/s, and increases when wind speed exceeds 3.5 m/s. The findings can be used as an early risk indicator, assisting local health authorities in developing a simple climate-based disease early warning system to help minimize outbreaks of HFMD.