A Simple Spatial Harmonization Technique by Aggregation for Ecological Studies: An Application Using Swiss Healthcare Regions

Geographical units are fundamental to epidemiological research, yet their boundaries frequently change over time, creating challenges for long-term analyses. We propose a simple aggregation-based framework to harmonize regions with temporally varying boundaries by constructing stable spatial units. The method represents geographic regions from different administrative versions as nodes in a graph and links regions that share underlying postal codes. Connected components of this graph define unified, time-consistent regions that can be used for downstream analyses without disaggregating outcome data. We demonstrate the approach using the Swiss MedStat Regions (SMR), whose boundaries were substantially revised in 2008, resulting in a many-to-many correspondence between old and new definitions. Postal codes, assumed to be largely stable, serve as the bridging key. The resulting harmonized regions (mSMR) were applied to a nationwide time-series study of mean daily summer temperature and emergency hospital admissions for respiratory diseases among older adults (≥65 years) in Switzerland from 1998 to 2023. The proposed approach is most appropriate when events are sparse at fine scale (e.g., daily counts), exposure can be aggregated to harmonized units, and some loss of spatial resolution is acceptable in exchange for temporal consistency and avoidance of interpolation. Limitations include potential masking of local heterogeneity and susceptibility to the modifiable areal unit problem, particularly for exposures with strong small-scale spatial gradients. This graph-based harmonization offers a practical solution for maintaining spatiotemporal consistency in longitudinal epidemiological studies where only aggregate health data are available.