Relative and absolute exposures in multi-site and multi-cohort research

Abstract Pooling data across study sites or studies can provide opportunities to answer novel research questions, especially when exposure distributions vary across settings. The decision to use absolute (ie, on a common metric across studies) versus relative (ie, scaled within each contributing study) exposure measures is an under-appreciated but important decision point, and is relevant across heterogeneous exposure domains and disciplines. The use of absolute versus relative exposure measures can impact construct validity, study results, and the interpretation of findings. We highlight three key considerations that should drive decision-making: 1) conceptual and hypothesis-related factors (ie, construct validity), 2) confounding by place, and 3) harmonization. We argue that construct validity should be prioritized above other factors, though challenges around confounding by place and considerations related to the comparability of exposure measures across surveys may also influence decision-making. To help readers understand how these considerations may apply practically, we discuss two example associations (educational attainment and cognition, heat and mortality) that illustrate implications related to the choice of absolute versus relative exposure measures. We highlight key considerations and provide recommendations to researchers to aid decision-making and increase the quality and transparency of research.