Individual and neighborhood risk factors of hospital admission and death during the COVID-19 pandemic: a population-based cohort study

Globally, it is known that countries with high levels of income inequality have performed significantly worse when dealing with the coronavirus disease 2019 (COVID-19) outbreak in terms of cases and deaths [1, 2]. Moreover, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), causing COVID-19 disproportionately affects minority populations in the USA and in the UK [3‐5].

During the 2009 H1N1, influenza reports in the USA of racial ethnic disparities with regard to complications and hospitalization [6] led to a national survey documenting the sources of these inequalities [7]. Increased risk of exposure to the virus and increased susceptibility to severe consequences of the infection combined with lack of health care access were important contributors.

In the rest of Western Europe and specifically in the Nordics, there is a paucity of data regarding the socioeconomic impact on the risk of infection, hospitalization, and mortality due to COVID-19. Even though the risk of COVID-19 varies enormously between geographical areas as well as between individuals with different sociodemographic characteristics, such as sex, age, and socioeconomic status, the relationship between geographical areas and individual risk factors is unknown.

The present study sought to address this knowledge gap with a population-wide study of all adult inhabitants in the Stockholm Region, excluding older adults living in nursing homes or receiving in-home care. We used the real-time COVID-19 monitoring framework in Region Stockholm to investigate the demographic and socioeconomic impact on the risk of hospitalization and mortality.

We hypothesized that demographic and socioeconomic factors contributed to the risk of hospitalization and death independently of other co-morbidity-related risk factors. We further wanted to assess if living areas carried additional risks of COVID hospitalization or death, even when adjusting for all individual factors.

In all, 1.7 million people were followed from March 1st, 2020, to February 28th, 2021, during which time 10,495 hospitalizations and 1148 deaths due to COVID-19 were registered. Tables 1 and 2 detail the individual effect estimates on covid-19 hospital admission and mortality. Men had a higher risk of both hospitalization and death than women, with relative risk (RR) 1.6 (confidence interval (CI) 1.53–1.66) and RR 2.5 (CI 2.19–2.85), respectively. Increasing age was strongly associated with increasing risks, mainly regarding to mortality where a relative risk of 53 was seen in the ≥75 years age category. Immigrants to Sweden were generally at higher risks of both outcomes, with a RR for hospitalization of 2.11 (2.02–2.21) and RR for death was 1.68 (1.47–1.92). Lower education and lower income levels were significantly coupled with higher relative risks of mortality. The lowest income level was associated with a doubled risk of death from COVID-19, and the presence of co-morbidities more than quadrupled the mortality risk.

Table 3 shows the contextual effect estimates on hospitalization and death. This model takes all the individual effect estimates — presented in Tables 1 and 2 — into account. The geographical area factors standing out include a high percentage of people not born in Sweden; mortality in such areas was elevated by over 50% compared to reference areas. We also found high population density and the age distribution in an area to be associated with increased risk of both hospital admission and mortality, where a high proportion of younger adults and a low proportion of older adults were associated with increased risks.

Tables 4 and 5 demonstrate the interface between area factors and individual effect estimates on the risk of hospital admission and death in a stepwise fashion. This model takes all the individual effect estimates — presented in Tables 1 and 2 — into account. Hospitalization is higher in areas with many children, fewer elderly people, high density, high area deprivation, and in areas with low percentages of Swedish-born inhabitants. Mortality is to a larger extent impacted by individual risk factors (Table 5) as demonstrated by non-significant relative risks. Important exceptions include areas with high density and low percentages of Swedish-born inhabitants. Neighborhood deprivation is explained in the methods section and the first principal component explained 42.22% of the total variability and it can be interpreted directly from Additional file1: Figure S1.

Additional file 1: Table S1 shows the crude risk of death depending on where you live and the risk after multiple adjustments. As mentioned in the methods, residential areas were divided into quintiles with the 20% most afflicted, i.e., the highest number of deaths, in quintile 1 and the 20% least afflicted in quintile 5. Crude data indicate a quadrupled death toll in the worst-off areas. Even after adjustments for sex, age, country of origin, education, type of work, income, living conditions, and co-morbid conditions; mortality is twice as high for people living in the most afflicted neighborhoods.

To our knowledge, this large, population-based study is the first to assess the combination of individual- and area-level risk factors on outcomes of COVID-19 patients. Our key finding is that segregation and deprivation are strongly associated to detrimental outcomes of COVID-19.

Individual factors like lower income, lower education, and inability to work from home together with age, male sex, and co-morbid conditions increase the risk of hospitalization and death. Country of origin plays a part in the individual risk of severe outcomes, but we also demonstrate group-level effects. Residence in an area with a higher percentage of immigrants was independently associated with an increased risk of hospital admission. Living in affluent neighborhoods, with few immigrants, was associated with both lower risk of hospitalization and death. There were other signals in the contextual effect estimates; areas with low numbers of children and a high percentage of people over 70 years were associated with decreased risk of hospital admission. Living in the most deprived areas in Stockholm was an independent risk factor for hospitalization during the first year of the COVID-19 pandemic.

One ecological study from Germany, during the first wave, showed COVID-19 cases and death negatively associated with the share of schoolchildren and children in day care as well as physician density [18]. In contrast to that and other ecological studies, the present investigation combines individual- and area-related factors. A Norwegian study of immigrants indicated that crowded housing and low income at a group level were correlated with cases and hospitalization of COVID-19 [19]. A nationwide Swedish study [20], albeit lacking data on hospitalization and solely examining the first wave (less than 2 months of follow-up) of the pandemic, scrutinizes sociodemographic factors. Being male, low individual income, lower education, and not being married independently predict a higher risk of death from COVID-19 and from all other causes of death. Immigrants from low- or middle-income countries had a higher risk of COVID-19 mortality but not for other causes of death.

The Stockholm data presented here augments this base of knowledge. On top of adding very granular individual and group-level information, we include data on hospitalization with over two years of follow-up encompassing five distinct COVID-19 waves. The greater Stockholm area has been the epicenter of the pandemic in Sweden and indeed in all of Scandinavia [21]. This could partly be explained by the fact that the region shares many traits of other European large cities with regard to precarious employment, crowded housing, and foreign-born inhabitants. The fact that individual factors, such as older age, lower income, less education as well as contextual factors — the type of area one resides in — are all associated with increased risk of hospitalization or death is paramount for politicians, policy makers, and health care personnel. This knowledge allows for targeted outreach, when striving to contain potential new variants of the SARS-CoV-2 virus, or indeed other transmissible diseases. To be specific, exploring and explaining the intersection between risk factors at the geographical area level and individual risk factors has implications for our ability to intervene in an optimal way. If, for example, the individual age was the most important factor, interventions need to reach that demographic group independently of where they live. However, as clearly demonstrated here, certain geographic area traits carry independent additional risk; interventions need to be focused towards high-risk areas and especially on old people residing in such areas. Support focused on those not born in Sweden could have had significant health benefits during the first year of the pandemic and must be improved in the future.

This investigation has strengths. The Region Stockholm real-time COVID-19 monitoring framework makes a detailed analysis of risk factors for COVID-19-related mortality and hospitalization quite simple. We have complete coverage of the entire greater Stockholm area population and all deaths and hospitalizations due to COVID-19 for the whole study period. Data resolution is exceptional, with individual data on education, income, work situation, living area, and living condition as well as the country of origin and co-morbid conditions. Additionally, we present unique sociodemographic properties on an area level; including population density, average income, age groups and data on foreign-born inhabitants. The Stockholm Syndrome described here, with large swaths of low-income workers, unable to work from home, and residing in crowded households may, as mentioned above, be quite generalizable. These are conditions rampant in many larger cities across the continent and in the world. These findings and these circumstances are likely to have an impact beyond COVID-19. Yet, the findings should be interpreted with caution. Swedish data on COVID-19 deaths is considered accurate, the National Board of Health and Welfare reports all cases where the underlying cause of death was COVID-19, regardless of whether the diagnosis was laboratory-confirmed or not [22]. Persons ill enough to seek hospital care will be admitted (the Swedish health care system is not based on private insurance). However, it is impossible to rule out that inequalities in hospitalization due to COVID-19 might partly be affected by differences in health-seeking behavior. Certain individuals may have a lower bar for seeking medical attention. In contrast, others may have the ability to “navigate” the health care system more efficiently; one needs to be aware of this when using hospital admission as a proxy for the severity of the disease. With this said, it is not self-evident in which direction this would skew the risk estimates. Additionally, some findings, like the fact that areas with more elderly people had a lower risk of hospitalization may seem counterintuitive. We can only speculate, but the adherence to regional and national recommendations, such as avoiding shopping, avoiding crowds, and general stay-at-home-recommendations, is likely higher among the elderly. Dichotomizing Stockholm residents into born in Sweden vs not born in Sweden and not taking into account time spent in Sweden may blunt our ability to draw conclusions regarding this variable. Lastly, the risks of death in areas that were hit hard by the pandemic could — despite our best efforts to adjust for all individual factors — be driven by residual confounding. At the same time — for policy makers in the healthcare sector — that does not matter. Our findings clearly indicate where and to whom we should focus governmental resources.

segregation and deprivation are well-known public health issues. This study from Stockholm, Sweden shows that the risk of hospitalization and death from COVID-19 is increased by area factors, such as population density, the number of people born in Sweden on top of aspects like the inability to work from home and living in crowded housing. In combination, these add to known individual risk factors such as male sex, high age, and cardiac, renal, liver, and other comorbidities [15]. Even when correcting for all known individual risk factors, residing in the highest-risk areas is associated with a more than doubled risk of dying from COVID-19. This finding, in conjunction with the aforementioned individual risk factors, is paramount for governments, agencies, and healthcare institutions interested in targeted interventions such as vaccine outreach for the elderly residing in high-risk areas. The interaction between sociodemographic factors and the virus is very clear; an insight that can help how we deal with future strains of SARS-CoV-2 or other pandemics.