COVID-19 in the neighbourhood: the socio-spatial selectivity of severe COVID-19 cases in Sweden, March 2020–June 2021

In this paper we analyse spatial and temporal variation in the risk of intensive care unit (ICU) admission for COVID-19 in Sweden. The analysis is based on geocoded and time-stamped data from the Swedish Intensive Care Registry (SIR). We merge this data with a classification of Swedish neighbourhood cluster types constructed from multi-scalar measures of socio-economic and country of birth segregation (Kawalerowicz and Malmberg in Multiscalar typology of residential areas in Sweden, 2021 available from https://doi.org/10.17045/sthlmuni.14753826.v1 ). We examine 1) if residence in more socio-economically deprived or diverse neighbourhood cluster types was associated with a higher risk of ICU admission for COVID-19, 2) if residence in more affluent neighbourhoods was associated with a lower risk of ICU admission for COVID-19, and 3) how these patterns changed over time during the three first waves of the pandemic. While the highest overall risk was associated with residence in urban disadvantage coupled with diversity, models where neighbourhood cluster types were interacted with waves reveal that the highest risk was associated with living in a neighbourhood cluster type characterised by rural town disadvantage coupled with diversity under the 3rd wave (February 2021–June 2021). Residence in such a neighbourhood cluster type was associated with a four times higher risk of ICU admission, compared to the reference category of living in a homogeneous rural neighbourhood cluster type with average levels of deprivation under wave 1. Looking at disparities within each wave we found that residence in most affluent urban areas was at first associated with a slightly higher risk of ICU admission for COVID-19 as compared with the reference category of living in a homogeneous rural neighbourhood cluster type, but under waves 2 and 3 this risk was no longer statistically significant. The largest inequalities between different neighbourhood cluster types could be seen during the 1st wave. Over time, the risks converged between different neighbourhood cluster types.