C I Neutel1. 1. Health Canada, Ottawa, Ontario, Canada.
Abstract
PURPOSE: Simpson's paradox is a type of severe confounding wherein a confounding variable changes the direction of an association. METHODS: This article demonstrates Simpson's paradox with three cohorts of naproxen users (new users, chronic users, and combined users) who were compared on the age/sex distribution of further naproxen use. Hypothetical new and chronic user populations were constructed with the same proportions for further naproxen use as their original counterparts. RESULTS: The hypothetical combined population showed an age/sex distribution opposite to that of the original combined population. CONCLUSIONS: This example of Simpson's paradox is a significant finding as many drug utilization studies do not distinguish between component populations.
PURPOSE: Simpson's paradox is a type of severe confounding wherein a confounding variable changes the direction of an association. METHODS: This article demonstrates Simpson's paradox with three cohorts of naproxen users (new users, chronic users, and combined users) who were compared on the age/sex distribution of further naproxen use. Hypothetical new and chronic user populations were constructed with the same proportions for further naproxen use as their original counterparts. RESULTS: The hypothetical combined population showed an age/sex distribution opposite to that of the original combined population. CONCLUSIONS: This example of Simpson's paradox is a significant finding as many drug utilization studies do not distinguish between component populations.