Integrating Climate Information into Meningitis Early Warning Systems

By Yang Liu, Climate and Society ’13

For this summer, I am working with Dr. Madeleine Thomson from the International Research Institute of Climate and Society (IRI), and Dr. Carlos Pérez García-Pando from the NASA Goddard Institute of Space Studies (GISS) on climate and health focusing on epidemic meningitis in the Sub Saharan region, an area that is historically known as the Meningitis Belt. Meningitis is an infectious disease caused by bacteria Neisseria Meningitides and is transmitted between humans through respiratory droplets or throat secretions. Existing research has confirmed the tie between environmental factors and the transmission risk of meningitis. Such factors include but are not limited to humidity, temperature and wind speed. With infectious diseases being one of the three most challenging climate sensitive health issues, the central question that the environmental science and the health communities both struggle to answer is: how to integrate climate information into public health decision making process? With this question in mind, my internship is divided into three pillars: project coordination, inter-institutional communication and model building. I am involved in coordinating a research project funded by the Research Council of Norway that studies aerosol impacts on meningitis morbidity and the NASA Development Project at IRI that tests the current disease suitability mapping tool based on satellite data, and I have drafted funding proposals to the Wellcome Trust and the Gates Foundation. However, in this blog, I would be only focusing on the model that I am currently working on with Dr. Pérez.

The purpose of this product is to make an improvement to the current meningitis epidemic early warning system (EWS) in Niger, which relies solely on historical health data, by integrating district level climate data. Raw data that I obtained from the NASA ROSES Earth Sciences Applications Feasibility Studies and the World Health Organizations (WHO) include population, weekly suspected meningitis case numbers, weekly absolute (g/m3) and relative (%) humidity. My independent variable (IV) is the weekly incidence rate (suspected cases/ 100,000 population) and my dependent variables (DV) are incidence rates of the past 0-4 weeks and relative humidity or absolute humidity. Although many other environmental factors have also been confirmed relevant to the transmission rate of meningitis, humidity has revealed the most straightforward causal mechanism and is therefore considered in the first stage. In the future, other environmental variables will also be considered. With these data, I am creating a multi-variable non-linear fitting model aiming for the best predicting power. In the process, I am testing a variety of methodologies including Logistic, Poisson and Negative Binomial regressions. Through discussion with my supervisor Dr. Pérez, we reach a consensus that using negative binomial regression to fit the non-linear model will be the most appropriate. And since there are 38 distinctive districts in Niger, 38 versions of this model are being created. In the remaining month of my internship, my task will be testing the quality of these models. I will compare the model output and the historical observation and fill in the contingency table shown below that is normally used to process bio-data:

*An epidemic is defined as 10+ suspected meningitis cases per 100,000 populations by the WHO.

Then I will be calculating the following: 

These criteria reflect whether this model is suitable for policy makers. The results will further be compared to those of Dr. Michelle Stanton from the Liverpool School of Tropical Medicine to identify the benefits of integrating climate information into Meningitis EWSs. This model will be a long term effort of IRI, EI and NASA-GISS and hopefully my work will be able to lay the ground for future research.

Feature image by: emilio labrador/Flickr