Dutch statistics agency tests method to group people using BMI-related data

Statistics Netherlands (CBS) has tested a method for identifying groups of people with similar patterns in body-mass-index-related data, combining health survey results with information on nearby amenities and air pollution. The agency says the resulting groups are not ready for clinical use, but could support further research into personalised prevention and targeted interventions. [1]

Researchers selected 47 BMI-related variables using a technique called Statistically Equivalent Signature, then applied k-means clustering to principal components derived with PCAMIX. A panel of health experts assessed whether the findings were practically relevant.

The work, by Marieke Beuken, K. van Berkel, M. Kleynen, I. Kanera and J. van den Brakel, is published as a CBS discussion paper. The study presents the approach as a way to identify population subgroups from varied health data; it does not establish that the clusters can guide individual diagnosis or treatment. [1]