Social disparities in the first wave of COVID-19 infections in Germany: A county-scale explainable machine learning approach

Doblhammer, G.; Reinke, C.; Kreft, D.

Background Little is known about factors correlated with this geographic spread of the first wave of Covid-19 infections in Germany. Given the lack of individual-level socioeconomic information on COVID-19 cases, we resorted to an ecological study design, exploring regional correlates of COVID-19 diagnoses. Data and Method We used data from the Robert-Koch-Institute on COVID-19 diagnoses by sex, age (age groups: 0-4, 5-14, 15-34, 35-59, 60-79, 80+), county (NUTS3 region) differentiating five periods (initial phase: through 15 March; 1st lockdown period: 16 March to 31 March; 2nd lockdown period: from 1 April to 15 April; easing period: 16 April to 30 April; post-lockdown period: 1 May through 23 July). For each period we calculated age-standardized incidence of COVID-19 diagnoses on the county level, using the German age distribution from the year 2018. We characterized the regions by macro variables in nine domains: "Demography", "Employment", "Politics, religion, and education", "Income", "Settlement structure and environment", "Health care", "(structural) Poverty", "Interrelationship with other regions", and "Geography". We trained gradient boosting models to predict the age-standardized incidence rates with the macro structures of the counties, and used SHAP values to characterize the 20 most prominent features in terms of negative/positive correlations with the outcome variable. Results The change in the age-standardized incidence rates over time is reflected in the changing importance of features as indicated by the mean SHAP values for the five periods. The first Covid-19 wave started as a disease in wealthy rural counties in southern Germany, and ventured into poorer urban and agricultural counties during the course of the first wave. The negative social gradient became more pronounced from the 2nd lockdown period onwards, when wealthy counties appeared to be better protected. Population density per se does not appear to be a risk factor, and only in the post-lockdown period did connectedness become an important regional characteristic correlated with higher infections. Features related to economic and educational characteristics of the young population in a county played an important role at the beginning of the pandemic up to the 2nd lockdown phase, as did features related to the population living in nursing homes; those related to international migration and a large proportion of foreigners living in a county became important in the post-lockdown period. Discussion In the absence of individual level data, explainable machine learning methods based on regional data may help to better understand the changing nature of the drivers of the pandemic. High mobility of high SES groups may drive the pandemic at the beginning of waves, while mitigation measures and beliefs about the seriousness of the pandemic as well as the compliance with mitigation measures put lower SES groups at higher risks later on.