Investigating the effects of medical density on health-seeking behaviours using a multiscale approach to residential and activity spaces: Results from a prospective cohort study in the Paris metropolitan area, France

BACKGROUND When measuring neighbourhood effects on health, it is both incorrect to treat individuals as if they were static and tied to their residential neighbourhood and to consider neighbourhoods rigid places whose geographical scales can be delineated a priori. We propose here to investigate the effects of residential medical density on health-seeking behaviours, taking into account the mono/polycentric structure of individual activity space (i.e., the space within which people move in the course of their daily activities) and exploring various neighbourhood units based on administrative delineations and regular grids. METHODS We used data collected in the SIRS cohort study, which was carried out over a 5-year period (2005-2010) among a representative population living in 50 census blocks in the Paris metropolitan area. In the 662 women who lived in the same census blocks during the follow-up period and who had reported a recent cervical screening at baseline, we studied the association between residential medical density and individual activity space and the incidence of delayed cervical screening (> 3 years) in multilevel logistic regression models after adjustment for potential confounders. RESULTS Among the 662 women studied, there were 94 instances of delayed cervical screening in 2010 (14%). The women who indicated that their activity space was concentrated within their neighbourhood of residence were significantly more at risk for an incident delayed cervical screening. No significant association was found between residential medical density and the incidence of delayed cervical screening. However, we observed a significant interaction between individual activity space and residential medical density. Indeed, women living in neighbourhoods with a low medical density had a significantly higher risk of delayed screening, but only if they reported that their daily activities were centred within their neighbourhood of residence. Lastly, a sensitivity analysis exploring various neighbourhood spatial units revealed that the incidence of delayed screening was better modelled when residential medical densities were calculated from a 1400 × 1400 metre grid or from adjacent census blocks. CONCLUSION This analysis underscores the view that people and neighbourhoods should be considered interacting entities. Using unsuitable neighbourhood units or neglecting the mono/polycentric structure of activity space would result in downplaying the importance of access to local health resources when addressing inequalities in health-seeking behaviours.