Exploring Usefulness of Meteorology Data for Predicting Malaria Cases

Introduction

Malaria and dengue fever are among the most important vector-borne diseases in the tropics and subtropics. Our study objective was to find out if meteorology can be used to identify areas of risk and predict incidence of malaria cases.

Methods

We obtained district level weekly reported malaria cases from Integrated Disease Surveillance Program (IDSP), Department of Health and Family Welfare, Andhra Pradesh, for three years, 2014 – 16. The average weekly meteorological parameters, specifically, minimum temperature, maximum temperature, humidity, rainfall were collected using hundred automated weather station data from Indian Meteorology Department.

We used Generalized Linear Model (GLM) with Poisson distribution and default log-link to estimate model parameters and also used quasi-Poisson method with Generalized Additive Model (GAM) that uses non-parametric regression with smoothing splines.

To identify areas and create a risk map we used GIS as a tool for malaria risk assessment. This risk mapping was based on metrological (humidity, rainfall, minimum temperature, maximum temperature) and satellite derived physical (Vegetation and Water Index, and Elevation) variables for risk assessment. This spatial risk mapping used Analytical Hierarchical Process (AHP) to predict areas of high risk.

Results

Our analysis shows, that higher minimum temperatures (e.g.>24⁰C) tend to lead to higher malaria counts but lower values do not seem to impact the malaria counts. On the other hand, higher values of maximum temperature (e.g.>32⁰C) seem to negatively impact the malaria counts. The relationships with rainfall and humidity appears not so strong once we account for smooth (weekly) trends and temperatures as both smooth curves seem to hover around the value of zero across all its value. We noted a daily rainfall amount of 4 to 5cm to have positive impact on malaria counts and excessive or much lower amount of rainfall appear to have very little effect on malaria counts.

High risk areas of malaria with in districts (smaller administrative units in a state) of the state of Andhra Pradesh were identified, these results were validated using the malaria occurrence data for the year of 2014-2016 and a statistically significant correlation of 0.68 was obtained.

Conclusion

Our analyses shows, the incremental increase in meteorology parameter does not lead to increase in reported malaria cases in the same manner for all the districts within the same state. The results of our study suggest for effective spatial targeting and control of malaria factors such as of vegetation; elevation and water index along with meteorological variables are useful.