Spatiotemporal dynamics of breast cancer screening across half a million invitations in Geneva, Switzerland

Background The implementation of population-based breast cancer screening programs has been pivotal for early cancer detection, yet sociospatial disparities in participation rates may remain. Understanding and monitoring these variations is essential for improving participation, enabled by modern space-time approaches. This study aimed to (1) assess the existence of spatial clustering of participation in a breast cancer screening program, (2) evaluate temporal shifts in spatial patterns, and (3) assess the relative importance of area-level determinants in predicting participation rates. Methods We used the emerging hot spot analysis to mine and visualize space-time participation patterns. We assessed the determinants of screening participation using eXtreme Gradient Boosting combined with SHapley Additive exPlanations values for model interpretation. This approach was applied to a dataset of 482,318 georeferenced invitations sent from 2003 to 2020 by the breast cancer screening program in the canton of Geneva, Switzerland. Results Here we show that the overall participation rate of 41.5% falls below the national average of 46%, despite increases across all population segments. Initial analysis shows a clear periurban-urban pattern with lower urban participation. Space-time pattern mining further delineates this pattern into 13 distinct profiles, with rates varying from 27.8% in intensifying cold spots to 49.2% in intensifying hot spots. Modeling reveals higher screening participation in socioeconomically deprived areas and a negative association between accessibility to screening centers and participation rates. Conclusions The approach applied in this study enables a more nuanced monitoring of screening participation dynamics. Our findings support targeted interventions in prioritized areas to further reduce cancer screening inequalities.