Local socio-structural predictors of COVID-19 incidence in Germany

Abstract

Socio-economic conditions and social attitudes are known to represent epidemiological determinants. Credible knowledge on socio-economic driving factors of the COVID-19 epidemic is still incomplete. Based on linear random effects regression, an ecological model is derived to estimate COVID-19 incidence in German rural/urban districts from local socio-economic factors and popularity of political parties in terms of their share of vote. Thereby, records provided by Germany's public health institute (Robert Koch Institute) of weekly notified 7-day incidences per 100,000 inhabitants per district from the outset of the epidemic in 2020 up to December 1, 2021, are used to construct the dependent variable. Local socio-economic conditions including share of votes, retrieved from the Federal Statistical Office of Germany, have been used as potential risk factors. Socio-economic parameters like per capita income, proportions of protection seekers and social benefit claimants, and educational level have negligible impact on incidence. To the contrary, incidence significantly increases with population density and we observe a strong association with vote shares. Popularity of the right-wing party Alternative for Germany (AfD) bears a considerable risk of increasing COVID-19 incidence both in terms of predicting the maximum incidences during three epidemic periods (alternatively, cumulative incidences over the periods are used to quantify the dependent variable) and in a time-continuous sense. Thus, districts with high AfD popularity rank on top in the time-average regarding COVID-19 incidence. The impact of the popularity of the Free Democrats (FDP) is markedly intermittent in the course of time showing two pronounced peaks in incidence but also occasional drops. A moderate risk emanates from popularities of the Green Party (GRÜNE) and the Christian Democratic Union (CDU/CSU) compared to the other parties with lowest risk level. In order to effectively combat the COVID-19 epidemic, public health policymakers are well-advised to account for social attitudes and behavioral patterns reflected in local popularities of political parties, which are conceived as proper surrogates for these attitudes. Whilst causal relations between social attitudes and the presence of parties remain obscure, the political landscape in terms of share of votes constitutes at least viable predictive “markers” relevant for public health policy making.

Spatio-Temporal Patterns of the SARS-CoV-2 Epidemic in Germany

“…The observation that epidemic activity corresponds to the geopolitical East–West division of Germany confirms earlier findings based on regression analysis [ 12 ]. The dependence of this clustering on the chosen measure of similarity and reduction algorithm could give rise to a possible critique.…”

“…This finding is compatible with our results above-derived from cluster analyses. Our previous findings in [ 12 ] draw us to the conclusion that the well-known political and socio-structural differences between East and West Germany are proper surrogates for the underlying mechanisms.…”

“…, 16) have been calculated from the counts retrieved from the Robert Koch-Institute (RKI) database [26] and from the respective age-specific sub-population sizes retrieved from [27]. Depending on the context, age a either refers to the age classes (in years) [0, 5), [5,12), [12,18), [18,30), [30,40), . .…”

“…It is no surprise that a number of surveys and studies clearly point to a polarisation and radicalisation in public attitudes and behaviour driven by a polarisation in elite rhetoric that hinders effective responses to the COVID-19 crisis [ 8 ] and gives rise to an increasing social Darwinism [ 9 ]. Thus, health behaviour is increasingly driven by political ideology [ 10 , 11 , 12 ], such that the resulting epidemic dynamics have become almost unpredictable and uncontrollable.…”

“…Using a least absolute shrinkage and selection operator (LASSO) regression, a U.S. study [ 21 ] identified subsets of socioeconomic predictors that are indicators of vulnerable counties. Another study conducted in Germany [ 12 ] also examined the impact of socioeconomic determinants at the local level on the basis of linear mixed models, but in addition, also examined the impact of political predictors, i.e., party vote shares. However, all the mathematical modelling approaches and statistical regressions mentioned, as well as the majority of similar modelling articles impossible to list here, depend on controversial assumptions.…”