Original research Sociodemographic determinants of intraurban variations in COVID-19 incidence: the case of Barcelona Antonio López-­Gay ‍ ‍,1,2 Jeroen Spijker ‍ ‍,2 Helen V S Cole,3 Antonio G Marques ‍ ‍,4 Margarita Triguero-­Mas ‍ ‍,5,6 Isabelle Anguelovski ‍ ‍,7,8 Marc Marí-Dell’Olmo ‍ ‍,9,10,11 Juan A Módenes ‍ ‍,2 Dolores Álamo-­Junquera,9 Fernando López-­Gallego,12 Carme Borrell ‍ ‍9,10,11 ►► Additional supplemental ABSTRACT transmission has shown that socioeconomic and material is published online Background Intraurban sociodemographic risk cultural factors at the individual, household and only. To view, please visit the journal online (http://​dx.​ factors for COVID-19 have yet to be fully understood. neighbourhood levels are essential mechanisms for doi.​org/​10.​1136/​jech-​2020-​ We investigated the relationship between COVID-19 community spread of the virus.2 3 216325). incidence and sociodemographic factors in Barcelona at Individual-­level risk factors such as gender, age a fine-­grained geography. or race/ethnicity are known to influence infec- For numbered affiliations see end of article. Methods This cross-­sectional ecological study is based tious disease incidence,4 5 including COVID- on 10 550 confirmed cases of COVID-19 registered 19.6 7 Although infection rates are similar between Correspondence to during the first wave in the municipality of Barcelona genders, men are more likely to have comorbid Dr Antonio López-­Gay, (population 1.64 million). We considered 16 variables conditions (such as hypertension, diabetes, obesity Department of Geography, on the demographic structure, urban density, household and cardiovascular diseases) that are also risk factors Autonomous University of conditions, socioeconomic status, mobility and health associated with worse COVID-19 outcomes.8 9 Barcelona, 08193 Bellaterra, characteristics for 76 geographical units of analysis Women, however, are often more exposed because Barcelona, Spain; ​tlopez@​ced.​uab.​cat (neighbourhoods), using a lasso analysis to identify the of their more frequent dedication to care profes- most relevant variables. We then fitted a multivariate sions.10 Older people are also known to be more Received 27 December 2020 Quasi-­Poisson model that explained the COVID-19 susceptible to COVID-19 and show higher fatality Accepted 30 May 2021 incidence by neighbourhood in relation to these rates.11 In contrast, the role that children play in variables. disease transmission is still unclear as they are Results Neighbourhoods with: (1) greater population rarely the index case12 and are less likely to transmit density, (2) an aged population structure, (3) a high COVID-19 to adults.13 On the other hand, school presence of nursing homes, (4) high proportions closures are likely to have led to increased child- of individuals who left their residential area during care by seniors,14 potentially increasing risk of lockdown and/or (5) working in health-­related transmission. occupations were more likely to register a higher Individual socioeconomic factors such as level number of cases of COVID-19. Conversely, COVID-19 of education, income, employment status and type incidence was negatively associated with (6) percentage of occupation are also thought to impact risk of of residents with post-­secondary education and (7) COVID-19. Although initial COVID-19 outbreaks population born in countries with a high Human emerged from international (business) travel and Development Index. winter holidays,15 subsequent trends reveal that Conclusion Like other historical pandemics, the those working in specific occupations, especially incidence of COVID-19 is associated with neighbourhood frontline, ‘essential’ jobs in health, care, retail sociodemographic factors with a greater burden faced and hospitality, are more at risk of infection.16 17 by already deprived areas. Because urban social and Individuals living in poverty and other marginal- health injustices already existed in those geographical ised populations are more susceptible to infectious units with higher COVID-19 incidence in Barcelona, the diseases.5 For instance, in the US context, racial- current pandemic is likely to reinforce both health and ised minorities (especially African Americans) are social inequalities, and urban environmental injustice all vulnerable social groups that exhibit higher than together. © Author(s) (or their average rates of infectious diseases. This has been employer(s)) 2021. No attributed to systematic and interpersonal racism, commercial re-­use. See rights and poorer access to healthcare facilities and other and permissions. Published by BMJ. INTRODUCTION health-­promoting resources.18 Global flows of people, resources, and capital Public health researchers have also long acknowl- To cite: López-­Gay A, involved in the production and maintenance of edged the importance of neighbourhood-­ level Spijker J, Cole HVS, et al. J sociodemographic and physical characteristics— urban life facilitate the spread of infectious disease Epidemiol Community Health Epub ahead of print: [please and the emergence of pandemics.1 After appearing including racial and economic residential segrega- include Day Month Year]. in China in late 2019, the first cases of COVID-19 tion, and the spatial distribution of affordable and doi:10.1136/jech-2020- were confirmed in Spain and elsewhere in Europe, fresh food, or public transport—for understanding 216325 by late January 2020. Previous research on virus health outcomes.19 20 Structural contexts and López-­Gay A, et al. J Epidemiol Community Health 2021;0:1–7. doi:10.1136/jech-2020-216325 1 Original research neighbourhood environments can therefore create uneven poor correspond to those of the LTCF which do not necessarily reflect living conditions and lasting environmental injustices for lower the socioeconomic position of the residents themselves. income or immigrant residents living in certain areas of a city,21 Our geographical unit of observation is the neighbourhood. resulting in health inequity by neighbourhood. In fact, during the We aggregated addresses of positive-­tested individuals by neigh- 1918 influenza pandemic, researchers already found a significant bourhood of residence. Although the municipality of Barcelona association between disease transmissibility and neighbourhood-­ (1.64 million inhabitants) is officially divided into 73 barris level social characteristics such as population density, illiteracy (Catalan for neighbourhood), for statistical purposes we have and unemployment.4 followed the adaptation developed by the Spanish National Emerging research on COVID-19 shows similar patterns and Statistical Office in several studies.33 This alternative division is pathways.22 For example, people living in denser neighbour- based on the official administrative division, but creates more hoods, with poor and overcrowded housing conditions have an statistically robust units in terms of population size, merging the elevated risk of infection as social contact in these living scenarios least populated with neighbouring units and splitting the most is more likely.11 23 Urban connectivity, mobility and the mode of populated ones, always according to urban and sociodemo- transport also play an important role in the spread of COVID- graphic criteria. Our final division consists of 76 units (hence- 19.24 At the neighbourhood level, greater use of private motor forth referred to as neighbourhoods). They contain an average of vehicles and less public transport mobility means less exposure 21 500 inhabitants and 1.3 km2 area. These units are very diverse to infection.25 Likewise, infection rates may be lower where in terms of wealth, housing characteristics, demographic ageing part of the (more mobile, international and national) popula- and health, factors known to be associated with the spread of tion was able to leave before movement restrictions or where infectious diseases. a higher proportion of people was able to work from home during lockdown. Conversely, rates may be higher where more Intraurban sociodemographic covariates essential workers live (occupations that are over-­represented by A total of 16 neighbourhood-­level indicators on demographic women and immigrants from low-­income countries) as they are structure, socioeconomic status, urban and household density, more likely to commute. Overall, higher mortality rates from mobility and health characteristics were initially chosen based COVID-19 are associated with poorer neighbourhood condi- on earlier established associations with COVID-19 (see table 1 tions, including a scarcity of healthcare facilities.26 The number for sources, expected association with COVID-19 and summary of nursing and retirement homes has also been associated with a statistics). Specifically, we included information on the propor- greater number of infections in the neighbourhood.27 tion of (1) young people (ages 0–15 years) and (2) elderly (70 To date, COVID-19 research on spatial variations has been years and older), and (3) the percentage of the population aged mainly set at the national or subnational levels. At this level of 70+ years who was male. Socioeconomic indicators included analysis, it is very difficult to disentangle the different inter- were (4) mean income per person, (5) age-­standardised ratio of vening factors behind risks and exposures to COVID-19 as this population with at least post-­secondary education, (6) percentage approach fails to reveal the diverse patterns within these larger of the population born in foreign countries with a high Human geographies. There is therefore a need to focus on geographically Development Index (HDI) and (7) low HDI. We also included smaller units to allow for better account of confounding factors28 (8) population density, (9) average number of persons per and enhance the predictive accuracy and interpretability of the dwelling and (10) people living alone. We obtained mobility resulting statistical model. As of late 2020, neighbourhood-­level data on: (11) the availability of private transportation and (12) studies of socio-­spatial inequality in COVID-19 infection and mobility during lockdown. We also captured the presence of (13) mortality have primarily focused on the USA and UK.29 30 Very transient populations (measured as the rate of inhabitants auto- little is known about such patterns in mainland Europe,31 espe- matically deregistered by the municipality, which occurs when cially so in much denser and mixed-­use urban environments. foreign residents fail to renew their registration), as cumulative To address these shortfalls, we investigated the relationship infection may be lower in areas with hypermobile groups (eg, between COVID-19 incidence and a comprehensive diversity of international students) that were likely to leave the city due to intraurban sociodemographic factors in Barcelona, Spain. the pandemic. We also incorporated (14) the number of LTCF beds per 1000 inhabitants and (15) the percentage of economi- cally active population in the health sector. Lastly, we included METHODS (16) the life expectancy at birth as a proxy for general health Study design and study population status. This cross-­sectional ecological study used data from the COVID-19 Register of the Barcelona Public Health Agency. During the first wave, Spain registered one of the highest per Statistical analyses capita number of cases in Europe, making analysis at the local Data transformation scale more reliable. Barcelona became one of the initial hotspots The distribution of each neighbourhood-­level sociodemographic in the country, possibly due to its international position in indicator and covariate was first assessed for normality using tourism, business, education and research.32 visual inspection of QQ plots and the Smirnov-­ Kolmogorov Our study included 10 550 laboratory-­ confirmed cases of test for normality. Accordingly, we log-­transformed: (1) young COVID-19 in Barcelona between 9 March and 3 May 2020. population, (2) income, (3) foreigners from high-­HDI countries, We selected these dates to focus on the first outbreak of the (4) foreigners from low-­HDI countries, (5) mobility during lock- pandemic. During this period, tests were essentially performed down and (6) transient populations. We also used a square root for those hospitalised or from specific at-­risk groups, especially transformation for the nursing homes variable. healthcare workers, as well as residents and workers in long-­ term care facilities (LTCFs). However, confirmed cases regis- Multiple variables model tered in LTCF were excluded, as test campaigns were unevenly To fit the total number of cases observed in each unit of anal- implemented across time and space and addresses of residents ysis, we relied on a generalised linear model (Quasi-­ Poisson 2 López-­Gay A, et al. J Epidemiol Community Health 2021;0:1–7. doi:10.1136/jech-2020-216325 Table 1 Covariates used in the study: hypothesised association with COVID-19, definitions, sources and summary statistics before transformation (when required*) Covariate Abbreviation Assoc. with Operationalisation and measurement Period Mean Min Max SD Demographic structure    1 .Percentage 0–15 P0–15 +a % of the population aged 0–15 (ln) 2019 16.2 8.3 22.8 3.1  2 .Percentage 70+ P70+ +b % of the population aged 70+ 2019 12.7 7.6 18.8 2.1  3. Old age–sex ratio OLDSEXR Men+c % of the 70+ population who is male 2019 38.8 33.5 45.7 2.3 Socioeconomic status    4. Income INCOME –d Mean income (€) per person (ln) 2016 15 199.0 7790.0 29 144.0 4633.0 d  5. Post-­secondary education UNI +/– Age-­standardised ratio of post-­secondary-­educated population† 2019 1.0 0.2 1.8 0.4  6. HDI high migrants MIGRHI +/–d % of the population born in foreign countries with a high HDI (ln) 2018 6.7 1.3 22.8 4.5 d  7. HDI low migrants MIGRLOW –/+ % of the population born in foreign countries with a low HDI (ln) 2018 17.5 6.5 46.7 7.9 Density (urban and household)    8. Population density (urban) URBDENSITY +e Inhabitants (in hundreds) per hectare, excl. forest parks 2019 2.9 0.2 5.9 1.4  9. Persons per dwelling HHSIZE1 +e Average number of persons per dwelling 2019 2.5 2.1 3.0 0.2  10. One person households 1PHH –e Age-­standardised relative ratio of persons living in one person households 2019 1.3 0.9 2.1 0.2 Mobility    11. Vehicles per inhabitant VEHICLES –f Number of registered vehicles per 100 persons aged 20+ years 2019 47.6 18.9 89.1 12.5  12. Mobility during lockdown MOBILE +g Percentage of people who left the area of residence during lockdown. 2020 10.7 5.5 19.6 2.8 López-­Gay A, et al. J Epidemiol Community Health 2021;0:1–7. doi:10.1136/jech-2020-216325 Daily average, 9 March–3 May 2020 (ln)  13. Population deregistration PADRON –h Automatic deregistrations from the municipal population register (expiration of the register)/1000 (ln) 2017–2018 27.3 7.6 81.9 14.1 Health    14. Nursing homes (LTCFs) NURSHOME +i Number of nursing and retirement homes (measured in beds) per 1000 people (square root) 2020 9.0 0.0 81.5 12.3  15. Health workers EMPLHEALTH +j Percentage of economically active population in health sector 2011 6.9 2.0 11.5 1.9 k  16. Life expectancy LE0 – Life expectancy at birth 2006–2017 84.0 80.3 86.3 1.3 Expected or observed association between covariate and mortality at the neighbourhood level (main unit of analysis): (a) the role that children play in disease transmission is still unclear as studies have shown that they are rarely the index case.12 On the other hand, school closures are likely to have led to increased childcare by elders (who are a high-­risk group)14; (b) 11; (c) although infection rates are similar, the severity of disease (hence COVID-19 reporting) and death is two times greater for men than for women41; (d) as an average for the studied period, a negative association with SES indicators is expected: higher income/educated/high-­HDI neighbourhoods lead to lower infection rates, although the association may be positive early on the epidemic15; (e) denser neighbourhoods and overcrowding are associated with more social contact11 15 23; (f) more vehicles, less mobility in public transport, less exposure to infection24; (g) more mobility in the geographical unit of analysis implies greater exposure to infection; (h) infection rates may be lower in neighbourhoods where part of the population has left; (i) more nursing and retirement homes, more infections in the neighbourhood27; (j) more health personnel, the greater the infection rate. This is for two reasons: they are more exposed to the virus and have systematically been tested. Note that there are 25 geographical units with repeated values due to census aggregation; (k) LE0 as a proxy for health status (the severity of COVID-19 is positively associated with coexisting illnesses).42 Data sources: (1, 2, 3, 5, 6, 7, 9, 10, 12, 13) Population Register (National Statistics and Barcelona City Council); (4) Atlas de distribución de renta de los hogares (National Statistics); (8) urban indicators (Barcelona City Council); (11) economic indicators (Barcelona City Council); (12) Estudio de movilidad durante el estado de alarma por COVID-19 (National Statistics); (14) location list and characteristics of nursing homes (Consorci de Serveis Socials de Barcelona); (15) 2011 Population Census (National Statistics); (16) health indicators (Agència de Salut Pública de Barcelona). *The type of transformation that was applied to make the variable values more normally distributed is mentioned in brackets. ln refers to natural logarithm. See also Data transformation subsection. †Here we first applied the age-­specific percentage of post-­secondary-­educated persons in the municipality to the residents of every neighbourhood by age group. We got, then, an estimated number of post-­secondary-­educated population for every unit of analysis, which we compared with the observed one. If the observed was greater than expected, the indicator is greater than 1; if lower, less than 1. HDI, Human Development Index; LTCFs, long-­term care facilities; SES, socioeconomic status. Original research 3 Original research Figure 1 Intraurban distribution of COVID-19 cumulative incidence in Barcelona from 9 March to 3 May 2020 (per 100 000 inhabitants). regression) that takes into account the total population as an offset as well as the sociodemographic variables. Given the rela- tively large number of covariates included in the study and the potential multicollinearity among them, we ran a lasso analysis to automatically identify the most relevant variables.34 In the context of generalised linear regression modelling and predic- tion, lasso performs both variable selection and regularisation to enhance prediction accuracy and interpretability of the statistical model. The hyperparameter of the lasso-­regularised maximum likelihood estimator was set using cross-­validation and, once lasso identified the most informative variables, we fitted the final Quasi-­Poisson model that explained the COVID-19 incidence for each unit of analysis considered. Finally, variable elasticities were calculated. This enables estimating the increase of cumula- tive incidence (and predict the total number of positive cases) for a 1% change in a particular covariate and thereby compare the effect of the different covariates. RESULTS The intraurban geography of the COVID-19 cumulative inci- dence in Barcelona during the period of study reveals a strong proximity among the units with the highest and lowest values (figure 1). Northern neighbourhoods (mainly located within the districts of Nou Barris and Horta-­Guinardó) have the highest Figure 2 Intraurban distribution of the sociodemographic covariates. incidence values, with some of them exceeding 1000 cases per HDI, Human Development Index. 100 000 inhabitants during the 8 weeks of observation. On the other hand, the incidence in the geographical units located in the southeast of the city (ie, historical centre) is less than one-­third lockdown, (6) LTCF and (7) health workers. These variables are of that in the worst-­affected neighbourhoods. mapped in figure 2. From the initial 16 variables considered, the lasso method Results of our Quasi-­Poisson model confirm that the associa- selected as meaningful to explain the observed COVID-19 levels tions between the final selection of variables and the intraurban the following seven (see also online supplemental material): COVID-19 incidence in Barcelona are all in the expected direc- (1) elderly, (2) high education, (3) foreigners from high-­HDI tion (table 2). Neighbourhoods that are densely populated, with countries, (4) population density (urban), (5) mobility during a higher number of older adults, with more numerous LTCF and 4 López-­Gay A, et al. J Epidemiol Community Health 2021;0:1–7. doi:10.1136/jech-2020-216325 Original research Table 2 Results of the generalised linear (Quasi-­Poisson regression) analysis of social and demographic factors on COVID-19 infection rates in Barcelona from 9 March to 3 May 2020 Estimated coefficients Elasticity # Change in cumulative incidence Equivalence in cases Estimate SE t-­statistic P value % 95% CI Number 95% CI (Intercept) −5.9826 0.3791 −15.7830 0.0000 Percentage 70+ 0.0228 0.0088 2.6000 0.0114 +0.29 0.07 to 0.51 28.0 7.0 to 49.0 Post-­secondary education −0.2694 0.1011 −2.6660 0.0096 −0.27 −0.47 to −0.07 −26.0 −45.0 to 6.8 HDI high migrants −0.0901 0.0693 −1.3000 0.1979 −0.09 −0.22 to 0.05 −8.6 −22.0 to 4.4 Population density (urban) 0.0295 0.0140 2.0990 0.0396 +0.09 0.01 to 0.16 8.2 1.0 to 15.9 Mobility during lockdown 0.2751 0.1182 2.3280 0.0229 +0.27 0.04 to 0.50 26.4 4.0 to 48.7 Nursing homes (LTCFs) 0.0279 0.0126 2.2160 0.0301 +0.04 0.00 to 0.08 4.0 0.0 to 7.6 Health workers 0.0260 0.0137 1.8930 0.0626 +0.18 −0.01 to 0.37 17.3 −1.0 to 35.2 Seventy-­six observations (neighbourhoods). Elasticity # is a form of standardised measure that estimates the relative change in the infection rate as a consequence of a relative change in a determinant. Here, we considered the effect of a 1% increase in the factor on the infection rate, keeping all other factors the same. The equivalence of cases is based on the number of cases for Barcelona as estimated by the model (9646) using the mean city-­level values of each covariate as shown in table 1. HDI, Human Development Index; LTCFs, long-­term care facilities. with higher proportions of individuals who left their area of resi- of general poor health and comorbidities also have implications dence during lockdown were statistically more likely to have a for risk of COVID-19 infections.8 9 higher number of cases of COVID-19 during the first outbreak of The environmental justice literature further demonstrates the pandemic. The work in health-­related occupations variable several causal pathways which may account for health differences was significant at the 0.063 level. Conversely, the association by neighbourhood socioeconomic status by showing that, for with COVID-19 cases is negative with the other two socioeco- example, neighbourhoods with high percentages of low-­income nomic factors: post-­secondary-­educated residents and popula- and non-­ university-­ educated residents historically have more tion born in high-­HDI countries, with the second one being less environmental hazards,36 putting residents at greater exposure relevant (note that while the cross-­validation analysis of the lasso-­ to risks leading to greater related health impacts. Because urban regularised 16-­variable regression deems the high-­HDI variable social and health injustices already existed in those neighbour- meaningful, the p value associated with the 7-­variable regression hoods with higher COVID-19 incidence in Barcelona, including casts doubts about its statistical significance). Considering the poor housing conditions, and at greater risk of economic disad- effect of the factors on the number of COVID-19 infections in vantage among others, the current pandemic is likely to reinforce a neighbourhood of Barcelona with average characteristics, a health and social inequalities and urban environmental injustice. 1% increase in older people or mobility during lockdown would People living in these neighbourhoods have less of a social safety lead to almost 30 extra cases, while a neighbourhood with a 1% net during times of both health and socioeconomic stress. They higher ratio of post-­secondary-­educated inhabitants leads to 26 are thus more likely to face an unjust burden in overcoming the fewer cases during the observed period according to our model. pandemic and its economic consequences. We finally ran a Global Moran’s I test to assess the potential During spring 2020, the lockdown in Spain limited mobility spatial autocorrelation of the model’s residuals, but results were strictly to those working in essential services, including low-­ not significant (see online supplemental material). wage jobs that require commuting by public transit to other parts of the city, which predicts higher COVID-19 incidence in DISCUSSION, INTERPRETATION AND IMPLICATIONS geographical units with higher numbers of commuters. In their Discussion case, additional health inequalities are likely to manifest because Our results confirm that incidence of COVID-19 is related to essential workers are often underpaid and underprotected, in several intraurban sociodemographic factors. In Barcelona, positions that require close interactions with the public. Addi- higher rates of infection were found in geographical units that tionally, they may already suffer from underlying health condi- were more densely populated, had more residents aged 70 tions due to their lower socioeconomic status, as recent research years or over, observed high levels of mobility during lock- suggests.37 As non-­ essential workers are losing their jobs or down, contained more nursing home facilities and had the facing less pay, these hardships affect lower educated (and logi- highest levels of people working in health-­related occupations. cally income) communities more, and jeopardise their ability to Conversely, neighbourhoods with relatively more residents with overcome the pandemic in the long term.38 In contrast, more high levels of education and with an immigration background privileged residents have greater ability to financially and physi- from high-­HDI countries registered fewer COVID-19 infections. cally recover. The negative association we found between infec- Our results are mostly in line with other indicators of spatial tion and neighbourhoods with high percentages of individuals health inequalities for Barcelona which indicate that residents in with post-­secondary degree and/or born in high-­HDI countries neighbourhoods located in the north of the city—generally lower can be understood from a dual perspective: first, the presence of income neighbourhoods, with lower education, denser areas and this type of residents is closely associated with neighbourhoods higher immigration from lower HDI countries (as an indicator of dominated by middle and upper socioeconomic households, ethnicity)—also have lower life expectancy and suffer more from which, in addition, were more likely to work remotely. Second, chronic diseases.35 The same exposures that put residents at risk this group is increasingly formed by young mobile and transient López-­Gay A, et al. J Epidemiol Community Health 2021;0:1–7. doi:10.1136/jech-2020-216325 5 Original research populations,39 who had the chance to return to their home coun- overcome the limitations of administrative boundaries. Another tries at the initial stage of the pandemic. limitation was that our estimates cover only the municipality of Last, results also indicate an expected structural age-­related Barcelona and do not include data from the metropolitan area. vulnerability, with neighbourhoods with a higher percentage Last, our measurement of incidence was biased toward more of residents over 70 years and/or with more nursing homes, severe patients with COVID-19 as testing procedures were predicting higher COVID-19 incidence. Those are thus intersec- restricted to hospital admissions at this stage of the pandemic. tional social vulnerabilities, particularly important for a context The seroprevalence study conducted between 27 April and 11 like Spain, which has a high ageing population and a high number May estimated that 7% of the residents in Barcelona’s province of residents in nursing homes, many of whom suffer from other had developed IgG antibodies against SARS-­CoV-2.40 Assuming comorbid conditions. this prevalence for the city, the total number of cases that we analysed represented between 10% and 15% of the people who Strengths and limitations became infected during our period of study. Therefore, our Barcelona is an excellent example to disentangle the spread of model is likely to be biased in estimating intraurban variations of the infection within dense and highly mixed-­use European urban the entire infected population, but not for predicting the most areas. Socioeconomic and urban conditions are significantly severe cases. Our results may also differ from subsequent waves different to other urban contexts where most of the research when massive and rapid COVID-19 testing became available that has been conducted. Another strength of our study is that the also detect asymptomatic cases. As the latter is more common high number of COVID-19 cases in Barcelona enabled us to test among younger people, the predictive value of the percentage various area-­level indicators. In addition, the vast availability of 70+ variable in intraurban variation of COVID-19 will likely be aggregated sociodemographic data at a fine-­grained scale allowed lower in subsequent waves. us to include many contextual factors that in other studies are often analysed separately. Nevertheless, using geographically Final thoughts aggregated data also has its limitations, as association found in Despite initial media and political narratives framing the ecological studies may not necessarily reflect those observed at pandemic as a social equaliser, our analysis shows how vulner- the individual level. An interesting future line of analysis would able groups by occupation, age and ethnicity, who reside in be to create buffer zones based on case addresses in order to Barcelona neighbourhoods with poor pre-­ existing social and environmental conditions, have statistically higher incidences of COVID-19. With the pandemic, their exposure to overlapping What is already known on this subject health risks has been compounded by new ones. The COVID-19 pandemic is therefore likely to reinforce existing health and ►► Previous research on virus transmission has shown social inequalities, and exacerbate urban environmental injustice that individual, household, and neighbourhood-­level in the city. These trends call for public policies and planning socioeconomic and cultural factors are associated with viral interventions to address neighbourhood environmental and transmission. social factors, strengthen social welfare and healthcare systems, ►► Most of COVID-19 research on spatial variations has been and improve open green and public spaces to serve as resources mainly set at the national or subnational regional level. and refuges for socially vulnerable groups. Because of the internal heterogeneity of these units, it is very Author affiliations difficult to disentangle the different intervening demographic 1 Department of Geography, Autonomous University of Barcelona, Barcelona, Spain and socioeconomic factors behind risks and exposures to 2 Center for Demographic Studies, Bellaterra, Spain COVID-19. 3 Barcelona Lab for Urban Environmental Justice and Sustainability, Autonomous ►► The limited research on the COVID-19 pandemic at University of Barcelona, Barcelona, Spain 4 the neighbourhood level (mainly in the USA and UK) Department of Signal Theory and Communications, Rey Juan Carlos University, Madrid, Spain identifies the effect of sociodemographic determinants, like 5 Institute for Environmental Science and Technology—Barcelona Lab for Urban socioeconomic status or ethnicity. Environmental Justice and Sustainability, Autonomous University of Barcelona, Cerdanyola del Vallès, Spain 6 Department of Urban Studies and Planning—Mariana Arcaya’s Research Lab, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA 7 What this study adds 8 Autonomous University of Barcelona, Bellaterra, Spain Catalan Institution for Research and Advanced Studies, Barcelona, Spain 9 Agència de Salut Pública de Barcelona, Barcelona, Spain ►► We analyse the spread of COVID-19 in Barcelona, a very 10 CIBER de Epidemiología y Salud Pública, Madrid, Spain dense and highly segregated city in Southern Europe, where 11 Institut d’Investigació Biomèdica Sant Pau (IIB Sant Pau), Barcelona, Spain 12 the first outbreak led to very high levels. Heterogeneous Biocatalysis Laboratory, CIC biomaGUNE, San Sebastian, Spain ►► We test a wide range of sociodemographic and urban characteristics, including mobility during lockdown, 16 Twitter Antonio López-­Gay @tonilopezga, Jeroen Spijker @popageing, Helen V S Cole @hvscole, Margarita Triguero-­Mas @ItaTrigueroMas, Isabelle Anguelovski @ variables in total, in order to predict intraurban variations in ianguelovski, Marc Marí-Dell’Olmo @epistatistic, Juan A Módenes @ModenesJA, COVID-19 infections at the neighbourhood level in Barcelona. Fernando López-­Gallego @flopez_gallego and Carme Borrell @carme1848 ►► The COVID-19 pandemic is likely to reinforce existing Contributors AGM, AL-­G and FL-­G conceived the study. AL-­G, FL-­G, MM-­D and health and social inequalities, and exacerbate urban JAM collected data, calculated indicators and built the final dataset. AGM, AL-­G, environmental injustice. These trends call for public policies MM-­D, JS and MT-­M conducted the statistical analyses. IA, CB, HC, AGM, AL-­G, JS and planning interventions that must address historical poor and MT-­M wrote the paper. AL-­G was the principal investigator of the study. All neighbourhood environmental and social factors, strengthen authors contributed to the interpretation of data, and read, edited and approved the final manuscript. social welfare systems, and improve open green and public spaces in cities. Funding This project has been funded by the following programmes: H2020 European Research Council (GREEN LULUs SG/GA678034 and HEALIN CoG/ 6 López-­Gay A, et al. J Epidemiol Community Health 2021;0:1–7. doi:10.1136/jech-2020-216325 Original research GA864616); Ministerio de Ciencia e Innovación (CSO2016-79142-­R; GLOBFAM/ 12 Ludvigsson JF. Children are unlikely to be the main drivers of the COVID-19 pandemic RTI2018-096730-­B-­I00I3; ’Maria de Maeztu’ Program/CEX2019-000940-­M; ’Ramón - A systematic review. Acta Paediatr 2020;109:1525–30. y Cajal’ Program/RYC-2013-14851; ’Juan de la Cierva’ Program/FJCI-2017-33842 13 Park YJ, Choe YJ, Park O, et al. Contact tracing during coronavirus disease outbreak, and IJC2018-035322-­I); Agència de Gestió d’Ajuts Universitaris i de Recerca South Korea, 2020. Emerg Infect Dis 2020;26:2465–8. (DEMFAMS/2017 SGR 1454); Talent Research Program, Universitat Autònoma de 14 Davies NG, Kucharski AJ, Eggo RM, et al. Effects of non-­pharmaceutical interventions Barcelona. on COVID-19 cases, deaths, and demand for hospital services in the UK: a modelling study. Lancet Public Health 2020;5:375–85. Map disclaimer The depiction of boundaries on this map does not imply the 15 Mogi R, Spijker J. The influence of social and economic ties to the spread of COVID-19 expression of any opinion whatsoever on the part of BMJ (or any member of its in Europe. J Popul Res 2021:1–7. group) concerning the legal status of any country, territory, jurisdiction or area or of 16 Gómez-­Ochoa SA, Franco OH, et al. COVID-19 in health-­care workers: a living its authorities. This map is provided without any warranty of any kind, either express systematic review and meta-­analysis of prevalence, risk factors, clinical characteristics, or implied. and outcomes. American Journal of Epidemiology 2020:kwaa191. Competing interests None declared. 17 The Lancet. The plight of essential workers during the COVID-19 pandemic. Lancet Patient consent for publication Not required. 2020;395:1587. 18 Aral SO, Adimora AA, Fenton KA. Understanding and responding to disparities Ethics approval No ethical approval was sought for this study as it used in HIV and other sexually transmitted infections in African Americans. Lancet aggregated, anonymous and publicly available data, collected at the neighbourhood 2008;372:337–40. level. 19 Bodor JN, Rice JC, Farley TA, et al. The association between obesity and urban food Provenance and peer review Not commissioned; externally peer reviewed. environments. J Urban Health 2010;87:771–81. 20 Diez Roux AV, Mair C. Neighborhoods and health. Ann N Y Acad Sci Data availability statement Our data are accessible to researchers upon 2010;1186:125–45. reasonable request for data sharing to the corresponding author. Our dataset has 21 Agyeman J, Schlosberg D, Craven L, et al. Trends and directions in environmental been built based on publicly available data in the referred repositories. justice: from inequity to everyday life, community, and just sustainabilities. Annu Rev Supplemental material This content has been supplied by the author(s). It Environ Resour 2016;41:321–40. has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have 22 Tai DBG, Shah A, Doubeni CA, et al. The disproportionate impact of COVID-19 on been peer-­reviewed. Any opinions or recommendations discussed are solely those racial and ethnic minorities in the United States. Clin Infect Dis 2021;72:703–6. of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and 23 Esteve A, Permanyer I, Boertien D, et al. National age and coresidence patterns shape responsibility arising from any reliance placed on the content. Where the content COVID-19 vulnerability. Proc Natl Acad Sci U S A 2020;117:16118–20. includes any translated material, BMJ does not warrant the accuracy and reliability 24 Sa F. Socioeconomic determinants of COVID-19 infections and mortality: evidence of the translations (including but not limited to local regulations, clinical guidelines, from England and Wales. Institute of Labor Economics (IZA), 2020: 159. http://​ftp.​iza.​ terminology, drug names and drug dosages), and is not responsible for any error org/​pp159.​pdf and/or omissions arising from translation and adaptation or otherwise. 25 Harris JE. The Subways seeded the massive coronavirus epidemic in New York City. National Bureau of Economic Research, Inc, 2020. http://www.​nber.​org/​papers/​ This article is made freely available for use in accordance with BMJ’s website w27021.​pdf terms and conditions for the duration of the covid-19 pandemic or until otherwise 26 Ji Y, Ma Z, Peppelenbosch MP, et al. Potential association between COVID-19 mortality determined by BMJ. You may use, download and print the article for any lawful, and health-­care resource availability. Lancet Glob Health 2020;8:e480. non-­commercial purpose (including text and data mining) provided that all copyright 27 Comas-­Herrera A, Zalakaín J, Litwin C. Mortality associated with COVID-19 outbreaks notices and trade marks are retained. in care homes: early International evidence. International Long-­Term Care Policy Network, 2020. ORCID iDs 28 Arpino B, Bordone V, Pasqualini M. No clear association emerges between Antonio López-­Gay http://​orcid.​org/​0000-​0001-​8892-​2816 intergenerational relationships and COVID-19 fatality rates from macro-­level analyses. Jeroen Spijker http://​orcid.​org/​0000-​0002-​3957-​9553 Proc Natl Acad Sci U S A 2020;117:19116–21. Antonio G Marques http://​orcid.​org/​0000-​0002-​4642-​7718 29 Maroko AR, Nash D, Pavilonis BT. COVID-19 and inequity: a comparative spatial Margarita Triguero-­Mas http://​orcid.​org/​0000-​0002-​1580-​2693 analysis of New York City and Chicago hot spots. J Urban Health 2020;97:461–70. Isabelle Anguelovski http://​orcid.​org/​0000-​0002-​6409-​5155 30 van Holm EJ, Wyczalkowski CK, Dantzler PA. Neighborhood conditions and the initial Marc Marí-Dell’Olmo http://​orcid.​org/​0000-​0003-​4345-​4988 outbreak of COVID-19: the case of Louisiana. J Public Health 2020. doi:10.1093/ Juan A Módenes http://​orcid.​org/​0000-​0003-​3049-​7495 pubmed/fdaa147. [Epub ahead of print: 03 Sep 2020]. Carme Borrell http://​orcid.​org/​0000-​0002-​1170-​2505 31 Brandén M, Aradhya S, Kolk M, et al. Residential context and COVID-19 mortality among adults aged 70 years and older in Stockholm: a population-­based, REFERENCES observational study using individual-­level data. Lancet Healthy Longev 2020;1:e80–8. 1 Mas-­Coma S, Jones MK, Marty AM. COVID-19 and globalization. One Health 32 Bakıcı T, Almirall E, Wareham J. A smart City initiative: the case of Barcelona. Journal 2020;9:100132. of the knowledge economy 2013;4:135–48. 2 Mossong J, Hens N, Jit M, et al. Social contacts and mixing patterns relevant to the 33 INE. Análisis de la movilidad de la población durante El estado de alarma POR spread of infectious diseases. PLoS Med 2008;5:e74. COVID-19 a partir de la posición de Los teléfonos móviles. Instituto Nacional de 3 Jarvis CI, Van Zandvoort K, Gimma A, et al. Quantifying the impact of physical Estadística, 2020. https://www.​ine.​es/​covid/​exp_​movilidad_​covid_​proyecto.​pdf distance measures on the transmission of COVID-19 in the UK. BMC Med 34 Tibshirani R. Regression shrinkage and selection via the LASSO. Journal of the Royal 2020;18:124. Statistical Society: Series B 1996;58:267–88. 4 Grantz KH, Rane MS, Salje H, et al. Disparities in influenza mortality and transmission 35 Agència de Salut Pública de Barcelona. La salut a Barcelona 2018. Agència de Salut related to sociodemographic factors within Chicago in the pandemic of 1918. Proc Pública de Bracelona: Barcelona, 2019. https://www.​aspb.​cat/​wp-​content/​uploads/​ Natl Acad Sci U S A 2016;113:13839–44. 2019/​12/​ASPB_​salutbarcelona2018.​pdf 5 Quinn SC, Kumar S. Health inequalities and infectious disease epidemics: a challenge 36 Evans GW, Kantrowitz E. Socioeconomic status and health: the potential role of for global health security. Biosecur Bioterror 2014;12:263–73. environmental risk exposure. Annu Rev Public Health 2002;23:303–31. 6 Bambra C, Riordan R, Ford J, et al. The COVID-19 pandemic and health inequalities. J 37 Dorn Avan, Cooney RE, Sabin ML. COVID-19 exacerbating inequalities in the US. Epidemiol Community Health 2020;74:964–8. Lancet 2020;395:1243. 7 Niedzwiedz CL, O’Donnell CA, Jani BD, et al. Ethnic and socioeconomic differences 38 Cole HV, Anguelovski I, Baró F. The COVID-19 pandemic: power and privilege, in SARS-­CoV-2 infection: prospective cohort study using UK Biobank. BMC Med gentrification, and urban environmental justice in the global North. Cities & Health 2020;18:160. 2020:1–5. 8 Bonow RO, Fonarow GC, O’Gara PT, et al. Association of coronavirus disease 2019 39 Cocola-­Gant A, Lopez-­Gay A. Transnational gentrification, tourism and the formation (COVID-19) with myocardial injury and mortality. JAMA Cardiol 2020;5:751–3. of ’foreign only’enclaves in Barcelona. Urban Studies 2020:0042098020916111. 9 Shi Y, Wang Y, Shao C, et al. COVID-19 infection: the perspectives on immune 40 Pollán M, Pérez-­Gómez B, Pastor-­Barriuso R, et al. Prevalence of SARS-­CoV-2 in Spain responses. Cell Death Differ 2020;27:1451–4. (ENE-­COVID): a nationwide, population-­based seroepidemiological study. Lancet 10 Ryan NE, El Ayadi AM. A call for a gender-­responsive, intersectional approach to 2020;396:535–44. address COVID-19. Glob Public Health 2020;15:1404–12. 41 Bischof E, Wolfe J, Klein SL. Clinical trials for COVID-19 should include sex as a 11 Dowd JB, Andriano L, Brazel DM, et al. Demographic science AIDS in variable. J Clin Invest 2020;130:3350–2. understanding the spread and fatality rates of COVID-19. Proc Natl Acad Sci U S A 42 Guan W-­jie, Ni Z-­yi, Hu Y, et al. Clinical characteristics of coronavirus disease 2019 in 2020;117:9696–8. China. N Engl J Med Overseas Ed 2020;382:1708–20. López-­Gay A, et al. J Epidemiol Community Health 2021;0:1–7. doi:10.1136/jech-2020-216325 7