Environment Related Mortality in the Netherlands

By dr. Maciek Strak – National Institute for Public Health and the Environment (RIVM)

In their daily lives, people are exposed to an accumulation of environmental risks, such as air pollution and noise, but also to environmental factors with a possible positive effect on health, such as greenery in the living environment. This project aims to evaluate the individual and combined effects of various environmental factors (especially air pollution, noise, and greenery) on the health of adults based on country-wide questionnaires and registrations. Studying the combined effects of those factors is highly complex: it requires large data sets to ensure the necessary statistical power, and consequently, also a great deal of computing power to process the data. 

RIVM studies the environment-related mortality in the Netherlands by combining various sources of data, including the Statistics Netherlands’ register data for causes of death and environmental data from the RIVM. Linking data from various files and sources within the secure CBS Remote Access environment and at an individual level allows the researchers from RIVM to establish exposure-effect relationships for the entire Dutch population. This data is unique worldwide, and the Netherlands has the largest cohort in which these relationships can be examined at such a low spatial aggregation level. 

When working with complex research questions combined with highly granular data, the researchers face empirical challenges and require a great deal of computing power. The analyses that are being carried out in this project include both classical statistical methods (e.g., survival analysis, non-parametric models) and also the application of machine learning techniques (e.g., random forest, neural networks). The computing power (CPU, RAM) and parallelization possibilities that are offered by the ODISSEI Secure Supercomputer, a subset of the Snellius national supercomputer, designated to perform the analysis of highly sensitive register data, allows the analyses to be performed much more efficiently.

Relevant links

Photo by Jw. on Unsplash