Chapter 3 72 male)(17). We used the gamma method to calculate 95% confidence intervals (CI)(18, 19). To evaluate both relative and absolute differences between groups, we computed rate differences (RD) and rate ratios (RR) to compare DSR of hospitalisation between (i) city districts, (ii) migration background (first and second generation migrants combined) and (iii) a combination of both variables, i.e. six strata of city districts (dichotomized into the central districts with higher average incomes (Central/West/South/East) and peripheral districts with lower average incomes (Southeast/North/New-West)), and migration background (none [ethnic-Dutch], European, non-European)(13). We combined first and second generation as this was highly correlated with age and a stratified analysis would result in low numbers in specific age-strata. Only individuals with available data on both city district (i.e., postal code of residence) and migration status (i.e., were able to be linked to municipality records) were included in the comparative analyses. A sensitivity analysis was performed where the DSR for confirmed COVID-19 related hospitalisations and were restricted to the population aged < 60 years. The purpose of this analysis was to evaluate the possible confounding effect of receiving formal home care or living in a nursing home, on the association between migration background and COVID-19 hospitalisation risk. Previous literature indicates that migration background is correlated with residing in peripheral city districts with lower socio-economic status than central city districts(20). We therefore wished to evaluate the independent effects of city district (peripheral, central) and migration background (none [ethnic-Dutch], European, non-European). We did so using a Poisson regression model, adjusting for age and sex, and using the log of the population size per district/migration background/sex/age stratum as an offset. In this model, we additionally evaluated possible effect modification of city district on migration background by adding an interaction term between city district and migration background, which was tested for statistical significance using a likelihood ratio test (LRT). In the epidemiological curve by date of symptom onset, missing dates were imputed, assuming that all those who were tested had symptoms. First, we created a distribution of time between symptom onset and case notification for those with a known date of symptom onset. Second, we randomly sampled from this distribution to estimate the date of symptom onset if this was missing. We assumed statistical significance at a P-value<0ยท05. We used the dsr package in R to calculate DSR, RD and RR(19). All analyses were performed in R (version 3.6.3, Vienna, Austria).
RkJQdWJsaXNoZXIy MTk4NDMw