Statistical methods for investigating and controlling for weather-health dependences in time series data

Lead Research Organisation: London School of Hygiene & Tropical Medicine
Department Name: Public Health and Policy


Interest has grown recently on the effects of weather, especially temperature, on health. One cause of this is the concern over health effects of global climate change. Another is the emergence, mostly as a by-product of studies of air pollution and health, of evidence pointing to substantial importance of weather in causing poor health – in particular events such as death or admission to hospitalisation. Although variation in health according to season have long been known, the details of this relationship remain only partly understood.
Studies to estimate health effects of weather and of climate change depend heavily on complex statistical methods. These are currently poorly developed, so we propose in this research project to improve them.
This research will improve the ability of future studies to identify with confidence the effects of weather and climate. Improving this knowledge will help us develop policies to reduce effects of weather and future changes in climate on our health.

Technical Summary

There is convincing evidence that weather, primarily temperature, substantially affects several health outcomes, but many details are unclear. Concerns over effects of climate change have increased interest in this association. The primary source of evidence on weather-effects is from studies of associations between variations in health and in weather over time. Statistical methods for investigating such time series have recently been enormously developed in the context of studies of air pollution and health. However, these methods are subject to ongoing debate, and aspects of their application to weather are undeveloped. Furthermore control for weather effects is one of the most debated points of air pollution studies.

We propose to identify, clarify, and develop new methods for investigating the relationship of weather and health. In particular we will clarify best methods to control confounding, model the typically non-linear and multi-lag association of outcomes with temperature, estimate the extent to which excess outcomes are due to short term displacement (?harvesting?), and investigate modifiers of weather effects. We will address these objectives by exploratory analyses on real data, analytic clarification of model properties, and simulations. Completing this work will improve capacity to clarify weather-health effects and control confounding by them in pollution studies.


10 25 50