Impact of the COVID-19 pandemic on levels of deep-ocean acoustic noise.

The extraordinary circumstances of the COVID-19 pandemic led to measures to mitigate the spread of the disease, with lockdowns and mobility restrictions at national and international levels. These measures led to sudden and sometimes dramatic reductions in human activity, including significant reductions in ship traffic in the maritime sector. We report on a reduction of deep-ocean acoustic noise in three ocean basins in 2020, based on data acquired by hydroacoustic stations in the International Monitoring System of the Comprehensive Nuclear-Test-Ban Treaty. The noise levels measured in 2020 are compared with predicted levels obtained from modelling data from previous years using Gaussian Process regression. Comparison of the predictions with measured data for 2020 shows reductions of between 1 and 3 dB in the frequency range from 10 to 100 Hz for all but one of the stations.

Using Geocoding to Identify COVID-19 Outbreaks in Congregate Residential Settings: San Francisco's Outbreak Response in Single-Room Occupancy Hotels.

More than 500 single-room occupancy hotels (SROs), a type of low-cost congregate housing with shared bathrooms and kitchens, are available in San Francisco. SRO residents include essential workers, people with disabilities, and multigenerational immigrant families. In March 2020, with increasing concerns about the potential for rapid transmission of COVID-19 among a population with disproportionate rates of comorbidity, poor access to care, and inability to self-isolate, the San Francisco Department of Public Health formed an SRO outbreak response team to identify and contain COVID-19 clusters in this congregate residential setting. Using address-matching geocoding, the team conducted active surveillance to identify new cases and outbreaks of COVID-19 at SROs. An outbreak was defined as 3 separate households in the SRO with a positive test result for COVID-19. From March 2020 through February 2021, the SRO outbreak response team conducted on-site mass testing of all residents at 52 SROs with outbreaks identified through geocoding. The rate of positive COVID-19 tests was significantly higher at SROs with outbreaks than at SROs without outbreaks (12.7% vs 6.4%; P < .001). From March through May 2020, the rate of COVID-19 cases among SRO residents was higher than among residents of other settings (ie, non-SRO residents), before decreasing and remaining at an equal level to non-SRO residents during later periods of 2020. The annual case fatality rate for SRO residents and non-SRO residents was similar (1.8% vs 1.5%). This approach identified outbreaks in a setting at high risk of COVID-19 and facilitated rapid deployment of resources. The geocoding surveillance approach could be used for other diseases and in any setting for which a list of addresses is available.

The Disclosure of Personally Identifiable Information in Studies of Neighborhood Contexts and Patient Outcomes.

Clinical epidemiology and patient-oriented health care research that incorporates neighborhood-level data is becoming increasingly common. A key step in conducting this research is converting patient address data to longitude and latitude data, a process known as geocoding. Several commonly used approaches to geocoding (eg, ggmap or the tidygeocoder R package) send patient addresses over the internet to web-based third-party geocoding services. Here, we describe how these approaches to geocoding disclose patients' personally identifiable information (PII) and how the subsequent publication of the research findings discloses the same patients' protected health information (PHI). We explain how these disclosures can occur and recommend strategies to maintain patient privacy when studying neighborhood effects on patient outcomes.

Spatial Analysis of COVID-19 cases and intensive care beds in the State of Ceará, Brazil. / Análise Espacial dos Casos de COVID-19 e leitos de terapia intensiva no estado do Ceará, Brasil.

The geographical distribution of COVID-19 through Geographic Information Systems resources is hardly explored. We aimed to analyze the distribution of COVID-19 cases and the exclusive intensive care beds in the state of Ceará, Brazil. This is an ecological study with the geographic distribution of the case detection coefficient in 184 municipalities. Maps of crude and estimated values (global and local Bayesian method) were developed, calculating the Moran index and using BoxMap and MoranMap. Intensive care beds were distributed through geolocalized points. In total, 3,000 cases and 459 beds were studied. The highest rates were found in the capital Fortaleza, the Metropolitan Region (MR), and the south of this region. A positive spatial autocorrelation has been identified in the local Bayesian rate (I = 0.66). The distribution of beds superimposed on the BoxMap shows clusters with a High-High pattern of number of beds (capital, MR, northwestern part). However, a similar pattern is found in the far east or transition areas with insufficient beds. The MoranMap shows clusters statistically significant in the state. COVID-19 interiorization in Ceará requires contingency measures geared to the distribution of specific intensive care beds for COVID-19 cases in order to meet the demand.

A distribuição geográfica da COVID-19 por meio de recursos de Sistemas de Informação Geográfica é pouco explorada. O objetivo foi analisar a distribuição de casos da COVID-19 e de leitos de terapia intensiva exclusivos para a doença no estado do Ceará, Brasil. Estudo ecológico, com distribuição geográfica do coeficiente de detecção de casos da doença em 184 municípios. Construíram-se mapas dos valores brutos e estimados (método bayesiano global e local), com cálculo do índice de Moran e utilização do "BoxMap" e "MoranMap" Os leitos foram distribuídos por meio de pontos geolocalizados. Estudaram-se 3.000 casos e 459 leitos. As maiores taxas encontram-se na capital Fortaleza, região metropolitana (RM) e ao sul dessa região. Há autocorrelação espacial positiva na taxa bayesiana local (I = 0,66). A distribuição dos leitos de terapia intensiva sobreposta ao "BoxMap" evidenciou aglomerados com padrão Alto-Alto apresentando número de leitos (capital, RM, porção noroeste); porém, há o mesmo padrão (extremo leste) e em áreas de transição com insuficiência de leito. O "MoranMap" evidenciou "clusters" estatisticamente significativos no estado. A interiorização da COVID-19 no Ceará demanda medidas de contingência voltadas à distribuição dos leitos de terapia intensiva específicos para casos de COVID19 para atender à demanda.

Écart de mobilité : estimation des seuils de mobilité requis pour maîtriser le SRAS-CoV-2 au Canada.

CONTEXTE: Les interventions non pharmacologiques demeurent le principal moyen de maîtriser le coronavirus du syndrome respiratoire aigu sévère 2 (SRAS-CoV-2) d'ici à ce que la couverture vaccinale soit suffisante pour donner lieu à une immunité collective. Nous avons utilisé des données de mobilité anonymisées de téléphones intelligents afin de quantifier le niveau de mobilité requis pour maîtriser le SRAS-CoV-2 (c.-à-d., seuil de mobilité), et la différence par rapport au niveau de mobilité observé (c.-à-d., écart de mobilité). MÉTHODES: Nous avons procédé à une analyse de séries chronologiques sur l'incidence hebdomadaire du SRAS-CoV-2 au Canada entre le 15 mars 2020 et le 6 mars 2021. Le paramètre mesuré était le taux de croissance hebdomadaire, défini comme le rapport entre les cas d'une semaine donnée et ceux de la semaine précédente. Nous avons mesuré les effets du temps moyen passé hors domicile au cours des 3 semaines précédentes à l'aide d'un modèle de régression log-normal, en tenant compte de la province, de la semaine et de la température moyenne. Nous avons calculé le seuil de mobilité et l'écart de mobilité pour le SRAS-CoV-2. RÉSULTATS: Au cours des 51 semaines de l'étude, en tout, 888 751 personnes ont contracté le SRAS-CoV-2. Chaque augmentation de 10 % de l'écart de mobilité a été associée à une augmentation de 25 % du taux de croissance des cas hebdomadaires de SRAS-CoV-2 (rapport 1,25, intervalle de confiance à 95 % 1,20­1,29). Comparativement à la mobilité prépandémique de référence de 100 %, le seuil de mobilité a été plus élevé au cours de l'été (69 %, écart interquartile [EI] 67 %­70 %), et a chuté à 54 % pendant l'hiver 2021 (EI 52 %­55 %); un écart de mobilité a été observé au Canada entre juillet 2020 et la dernière semaine de décembre 2020. INTERPRÉTATION: La mobilité permet de prédire avec fiabilité et constance la croissance des cas hebdomadaires et il faut maintenir des niveaux faibles de mobilité pour maîtriser le SRAS-CoV-2 jusqu'à la fin du printemps 2021. Les données de mobilité anonymisées des téléphones intelligents peuvent servir à guider le relâchement ou le resserrement des mesures de distanciation physique provinciales et régionales.

Temporal and geographic analysis of trichinellosis incidence in Chile with risk assessment.

Trichinellosis is a foodborne disease caused by several Trichinella species around the world. In Chile, the domestic cycle was fairly well-studied in previous decades, but has been neglected in recent years. The aims of this study were to analyze, geographically, the incidence of trichinellosis in Chile to assess the relative risk and to analyze the incidence rate fluctuation in the last decades. Using temporal data spanning 1964-2019, as well as geographical data from 2010 to 2019, the time series of cases was analyzed with ARIMA models to explore trends and periodicity. The Dickey-Fuller test was used to study trends, and the Portmanteau test was used to study white noise in the model residuals. The Besag-York-Mollie (BYM) model was used to create Bayesian maps of the level of risk relative to that expected by the overall population. The association of the relative risk with the number of farmed swine was assessed with Spearman's correlation. The number of annual cases varied between 5 and 220 (mean: 65.13); the annual rate of reported cases varied between 0.03 and 1.9 cases per 105 inhabitants (mean: 0.53). The cases of trichinellosis in Chile showed a downward trend that has become more evident since the 1980s. No periodicities were detected via the autocorrelation function. Communes (the smallest geographical administrative subdivision) with high incidence rates and high relative risk were mostly observed in the Araucanía region. The relative risk of the commune was significantly associated with the number of farmed pigs and boar (Sus scrofa Linnaeus, 1758). The results allowed us to state that trichinellosis is not a (re)emerging disease in Chile, but the severe economic poverty rate of the Mapuche Indigenous peoples and the high number of backyard and free-ranging pigs seem to be associated with the high risk of trichinellosis in the Araucanía region.

Geospatial Mapping of Orthopaedic Surgeons Age 60 and Over and Confirmed Cases of COVID-19.

BACKGROUND: Although elective surgical procedures in the United States have been suspended because of the coronavirus disease 2019 (COVID-19) pandemic, orthopaedic surgeons are being recruited to serve patients with COVID-19 in addition to providing orthopaedic acute care. Older individuals are deemed to be at higher risk for poor outcomes with COVID-19. Although previous studies have shown a high proportion of older providers nationwide across medical specialties, we are not aware of any previous study that has analyzed the age distribution among the orthopaedic workforce. Therefore, the purposes of the present study were (1) to determine the geographic distribution of U.S. orthopaedic surgeons by age, (2) to compare the distribution with other surgical specialties, and (3) to compare this distribution with the spread of COVID-19. METHODS: Demographic statistics from the most recent State Physician Workforce Data Reports published by the Association of American Medical Colleges were extracted to identify the 2018 statewide proportion of practicing orthopaedic surgeons ≥60 years of age as well as age-related demographic data for all surgical specialties. Geospatial data on the distribution of COVID-19 cases were obtained from the Environmental Systems Research Institute. State boundary files were taken from the U.S. Census Bureau. Orthopaedic workforce age data were utilized to group states into quintiles. RESULTS: States with the highest quintile of orthopaedic surgeons ≥60 years of age included states most severely affected by COVID-19: New York, New Jersey, California, and Florida. For all states, the median number of providers ≥60 years of age was 105.5 (interquartile range [IQR], 45.5 to 182.5). The median proportion of orthopaedic surgeons ≥60 years of age was higher than that of all other surgical subspecialties, apart from thoracic surgery. CONCLUSIONS: To our knowledge, the present report provides the first age-focused view of the orthopaedic workforce during the COVID-19 pandemic. States in the highest quintile of orthopaedic surgeons ≥60 years old are also among the most overwhelmed by COVID-19. As important orthopaedic acute care continues in addition to COVID-19 frontline service, special considerations may be needed for at-risk staff. Appropriate health system measures and workforce-management strategies should protect the subset of those who are most potentially vulnerable. LEVEL OF EVIDENCE: Therapeutic Level IV. See Instructions for Authors for a complete description of levels of evidence.

Topological data analysis model for the spread of the coronavirus.

We apply topological data analysis, specifically the Mapper algorithm, to the U.S. COVID-19 data. The resulting Mapper graphs provide visualizations of the pandemic that are more complete than those supplied by other, more standard methods. They allow for easy comparisons of the features of the pandemic across time and space and encode a variety of geometric features of the data cloud created from geographic information, time progression, and the number of COVID-19 cases. The Mapper graphs reflect the development of the pandemic across all of the U.S. and capture the growth rates as well as the regional prominence of hot-spots.

OUTBREAK: a user-friendly georeferencing online tool for disease surveillance.

The current COVID-19 pandemic has already claimed more than 3.7 million victims and it will cause more deaths in the coming months. Tools that track the number and locations of cases are critical for surveillance and help in making policy decisions for controlling the outbreak. However, the current surveillance web-based dashboards run on proprietary platforms, which are often expensive and require specific computational knowledge. We developed a user-friendly web tool, named OUTBREAK, that facilitates epidemic surveillance by showing in an animated graph the timeline and geolocations of cases of an outbreak. It permits even non-specialist users to input data most conveniently and track outbreaks in real-time. We applied our tool to visualize the SARS 2003, MERS, and COVID19 epidemics, and provided them as examples on the website. Through the zoom feature, it is also possible to visualize cases at city and even neighborhood levels. We made the tool freely available at https://outbreak.sysbio.tools/ . OUTBREAK has the potential to guide and help health authorities to intervene and minimize the effects of outbreaks.

Mapping the human genetic architecture of COVID-19.

The genetic make-up of an individual contributes to the susceptibility and response to viral infection. Although environmental, clinical and social factors have a role in the chance of exposure to SARS-CoV-2 and the severity of COVID-191,2, host genetics may also be important. Identifying host-specific genetic factors may reveal biological mechanisms of therapeutic relevance and clarify causal relationships of modifiable environmental risk factors for SARS-CoV-2 infection and outcomes. We formed a global network of researchers to investigate the role of human genetics in SARS-CoV-2 infection and COVID-19 severity. Here we describe the results of three genome-wide association meta-analyses that consist of up to 49,562 patients with COVID-19 from 46 studies across 19 countries. We report 13 genome-wide significant loci that are associated with SARS-CoV-2 infection or severe manifestations of COVID-19. Several of these loci correspond to previously documented associations to lung or autoimmune and inflammatory diseases3-7. They also represent potentially actionable mechanisms in response to infection. Mendelian randomization analyses support a causal role for smoking and body-mass index for severe COVID-19 although not for type II diabetes. The identification of novel host genetic factors associated with COVID-19 was made possible by the community of human genetics researchers coming together to prioritize the sharing of data, results, resources and analytical frameworks. This working model of international collaboration underscores what is possible for future genetic discoveries in emerging pandemics, or indeed for any complex human disease.

[Proposal for territorial analysis to face the SARS-CoV-2 pandemic based on the Profile of pneumonia and Influenza in Ecuador 2016-2018]. / Propuesta de análisis territorial para enfrentar la pandemia por SARS-CoV-2 basado en el perfil de neumonía e influenza en Ecuador 2016-2018.

OBJECTIVE: The present study seeks to analyse sociodemographic determinants related to severe acute respiratory infections (SARI) and calculate the priorization index in the cantons of Ecuador to identify areas probably most vulnerable to COVID-19 transmission. DESIGN: This descriptive ecological observational study. SETTING: 224 cantons (geographical area) of Ecuador with secondary data sources of hospital information. PARTICIPANTS: The unit of measurement was 224 cantons of Ecuador, in which analysed morbidity and lethality rates for SARI using hospital release data (2016-2018). MAIN MEASUREMENTS: Eight sociodemographic indicators were structuralized, and correlation tests applied for a multiple regression model. The priorization index was created with criteria of efficiency, efficacy, effect size (IRR) and equity. Using the sum of the index for each indicator, the priorization score was calculated and localized in a territorial map. RESULTS: Morbidity associated factors where: school attendance years, urbanization and population density; for mortality resulted: school attendance and ethnics (indigenous) IRR: 1.09 (IC95%:1.06-1.15) and IRR: 1.024 (IC95%:102-1.03) respectively. With lethality where related cantons, with population older than 60 years, IRR: 1.049 (IC95%: 1.03-1.07); 87 cantons had high priority mostly localized in the mountain region and the Morona Santiago Province. CONCLUSIONS: Morbidity and mortality of SARI in Ecuador are associated to social and demographic factors. Priorization exercises considering these factors permit the identification of vulnerable territories facing respiratory disease propagation. The social determinants characteristic for each territory should be added to known individual factors to analyse the risk and vulnerability for COVID in the population.

Geographic information system-based analysis of COVID-19 cases in India during pre-lockdown, lockdown, and unlock phases.

OBJECTIVE: The World Health Organization formally announced the global COVID-19 pandemic on March 11, 2020 due to widespread infections. In this study, COVID-19 cases in India were critically analyzed during the pre-lockdown (PLD), lockdown (LD), and unlock (UL) phases. METHOD: Analyses were conducted using geospatial technology at district, state, and country levels, and comparisons were also made with other countries throughout the world that had the highest infection rates. India had the third highest infection rate in the world after the USA and Brazil during UL2.0-UL3.0 phases, the second highest after the USA during UL4.0-UL5.0 phases, and the highest among South Asian Association for Regional Cooperation (SAARC) countries in PLD-UL5.0 period. RESULTS: The trend in the number of COVID-19 cases was associated with the population density where higher numbers tended to be record in the eastern, southern, and west-central parts of India. The death rate in India throughout the pandemic period under study was lower than the global average. Kerala reported the maximum number of infections during PLD whereas Maharashtra had the highest numbers during all LD and UL phases. Eighty percent of the cases in India were concentrated mainly in highly populous districts. CONCLUSION: The top 25 districts accounted for 70.99%, 69.38%, 54.87%, 44.23%, 40.48%, and 38.96% of the infections from the start of UL1.0 until the end of UL phases, respectively, and the top 26-50 districts accounted for 6.38%, 6.76%, 11.23%, 12.98%, 13.40%, and 13.61% of cases in these phase, thereby indicating that COVID-19 cases spread during the UL period. By October 31, 2020, Delhi had the highest number of infections, followed by Bengaluru Urban, Pune, Mumbai, Thane, and Chennai. No decline in the infection rate occurred, even in UL5.0, thereby indicating a highly alarming situation in India.

Unsupervised cluster analysis of SARS-CoV-2 genomes reflects its geographic progression and identifies distinct genetic subgroups of SARS-CoV-2 virus.

Over 10,000 viral genome sequences of the SARS-CoV-2virus have been made readily available during the ongoing coronavirus pandemic since the initial genome sequence of the virus was released on the open access Virological website (http://virological.org/) early on January 11. We utilize the published data on the single stranded RNAs of 11,132 SARS-CoV-2 patients in the GISAID database, which contains fully or partially sequenced SARS-CoV-2 samples from laboratories around the world. Among many important research questions which are currently being investigated, one aspect pertains to the genetic characterization/classification of the virus. We analyze data on the nucleotide sequencing of the virus and geographic information of a subset of 7640 SARS-CoV-2 patients without missing entries that are available in the GISAID database. Instead of modeling the mutation rate, applying phylogenetic tree approaches, and so forth, we here utilize a model-free clustering approach that compares the viruses at a genome-wide level. We apply principal component analysis to a similarity matrix that compares all pairs of these SARS-CoV-2 nucleotide sequences at all loci simultaneously, using the Jaccard index. Our analysis results of the SARS-CoV-2 genome data illustrates the geographic and chronological progression of the virus, starting from the first cases that were observed in China to the current wave of cases in Europe and North America. This is in line with a phylogenetic analysis which we use to contrast our results. We also observe that, based on their sequence data, the SARS-CoV-2 viruses cluster in distinct genetic subgroups. It is the subject of ongoing research to examine whether the genetic subgroup could be related to diseases outcome and its potential implications for vaccine development.

Neighborhood Disadvantage Measures and COVID-19 Cases in Boston, 2020.

OBJECTIVE: Understanding the pattern of population risk for coronavirus disease 2019 (COVID-19) is critically important for health systems and policy makers. The objective of this study was to describe the association between neighborhood factors and number of COVID-19 cases. We hypothesized an association between disadvantaged neighborhoods and clusters of COVID-19 cases. METHODS: We analyzed data on patients presenting to a large health care system in Boston during February 5-May 4, 2020. We used a bivariate local join-count procedure to determine colocation between census tracts with high rates of neighborhood demographic characteristics (eg, Hispanic race/ethnicity) and measures of disadvantage (eg, health insurance status) and COVID-19 cases. We used negative binomial models to assess independent associations between neighborhood factors and the incidence of COVID-19. RESULTS: A total of 9898 COVID-19 patients were in the cohort. The overall crude incidence in the study area was 32 cases per 10 000 population, and the adjusted incidence per census tract ranged from 2 to 405 per 10 000 population. We found significant colocation of several neighborhood factors and the top quintile of cases: percentage of population that was Hispanic, non-Hispanic Black, without health insurance, receiving Supplemental Nutrition Assistance Program benefits, and living in poverty. Factors associated with increased incidence of COVID-19 included percentage of population that is Hispanic (incidence rate ratio [IRR] = 1.25; 95% CI, 1.23-1.28) and percentage of households living in poverty (IRR = 1.25; 95% CI, 1.19-1.32). CONCLUSIONS: We found a significant association between neighborhoods with high rates of disadvantage and COVID-19. Policy makers need to consider these health inequities when responding to the pandemic and planning for subsequent health needs.

Visualizing COVID-19 Mortality Rates and African-American Populations in the USA and Pennsylvania.

The Centers for Disease Control and Prevention has identified African-Americans as having increased risk of COVID-19-associated mortality. Access to healthcare and related social determinants of health are at the core of this disparity. To explore the geographical links between race and COVID-19 mortality, we created descriptive maps of COVID-19 mortality rates in relation to the percentage of populations self-identifying as African-American across the USA, by state, and Pennsylvania (PA), by county. In addition, we used bivariate and logistic regression analyses to quantify the statistical relationship between these variables, and control for area-level demographic, healthcare access, and comorbidity risk factors. We found that COVID-19 mortality rates were generally higher in areas that had higher African-American populations, particularly in the northeast USA and eastern PA. These relationships were quantified through Pearson correlations showing significant positive associations at the state and county level. At the US state-level, percent African-American population was the only significant correlate of COVID-19 mortality rate. In PA at the county-level, higher percent African-American population was associated with higher COVID-19 mortality rate even after controlling for area-level confounders. More resources should be allocated to address high COVID-19 mortality rates among African-American populations.

Research response to coronavirus disease 2019 needed better coordination and collaboration: a living mapping of registered trials.

OBJECTIVES: Researchers worldwide are actively engaging in research activities to search for preventive and therapeutic interventions against coronavirus disease 2019 (COVID-19). Our aim was to describe the planning of randomized controlled trials (RCTs) in terms of timing related to the course of the COVID-19 epidemic and research question evaluated. STUDY DESIGN AND SETTING: We performed a living mapping of RCTs registered in the WHO International Clinical Trials Registry Platform. We systematically search the platform every week for all RCTs evaluating preventive interventions and treatments for COVID-19 and created a publicly available interactive mapping tool at https://covid-nma.com to visualize all trials registered. RESULTS: By August 12, 2020, 1,568 trials for COVID-19 were registered worldwide. Overall, the median ([Q1-Q3]; range) delay between the first case recorded in each country and the first RCT registered was 47 days ([33-67]; 15-163). For the 9 countries with the highest number of trials registered, most trials were registered after the peak of the epidemic (from 100% trials in Italy to 38% in the United States). Most trials evaluated treatments (1,333 trials; 85%); only 223 (14%) evaluated preventive strategies and 12 postacute period intervention. A total of 254 trials were planned to assess different regimens of hydroxychloroquine with an expected sample size of 110,883 patients. CONCLUSION: This living mapping analysis showed that COVID-19 trials have relatively small sample size with certain redundancy in research questions. Most trials were registered when the first peak of the pandemic has passed.

The pandemic predominantly hits poor neighbourhoods? SARS-CoV-2 infections and COVID-19 fatalities in German districts.

BACKGROUND: Reports from the UK and the USA suggest that coronavirus disease 2019 (COVID-19) predominantly affects poorer neighbourhoods. This article paints a more complex picture by distinguishing between a first and second phase of the pandemic. The initial spread of infections and its correlation with socio-economic factors depends on how the virus first entered a country. The second phase of the pandemic begins when individuals start taking precautionary measures and governments implement lockdowns. In this phase, the spread of the virus depends on the ability of individuals to socially distance themselves, which is to some extent socially stratified. METHODS: We analyze the geographical distribution of known cumulative cases and fatalities per capita in an ecological analysis across local districts in Germany distinguishing between the first and the second phase of the pandemic. RESULTS: In Germany, the virus first entered via individuals returning from skiing in the Alps and other international travel. In this first phase, we find a positive association between the wealth of a district and infection rates and a negative association with indicators of social deprivation. During the second phase and controlling for path dependency, districts with a higher share of university-educated employees record fewer new infections and deaths and richer districts record fewer deaths, districts with a higher unemployment rate record more deaths. CONCLUSIONS: The social stratification of COVID-19 changes substantively across the two phases of the pandemic in Germany. Only in the second phase and controlling for temporal dependence does COVID-19 predominantly hit poorer districts.