Inequalities in COVID-19 severe morbidity and mortality by country of birth in Sweden

The study investigates whether and how inequalities in severe COVID-19 outcomes (ICU admission and mortality) between migrants and Swedish-born individuals varied across the different waves of the pandemic in Sweden. Prior work has shown elevated risks of infection, hospitalization and death among ethnic minorities and migrants, but most evidence pertains to the early phase of the pandemic and does not distinguish between waves dominated by different variants. Socioeconomic status and living conditions have been suggested as key contributors to disparities, and migrants have exhibited lower vaccination uptake in later waves. The aim is to quantify disparities by region/country of birth across the full course of the pandemic and to assess the contributions of socioeconomic and living conditions and vaccination uptake to these differences.

Multiple studies in Sweden and internationally have documented higher risks of COVID-19 infection, hospitalization, and mortality among ethnic minorities and migrants. Socioeconomic disadvantages and adverse living conditions partially account for these disparities. Later in the pandemic, migrants showed lower vaccination uptake, which may have influenced severe outcomes. Evidence spanning all pandemic waves and explicitly examining how disparities evolved by wave has been limited, underscoring the need for this comprehensive analysis.

Design: Population-based cohort study of all adults residing in Sweden, with linked administrative registers. Cohort: Adults alive and resident at end of 2019; follow-up spanning March 1, 2020 to June 1, 2022. Sample size: n = 7,870,441. Data sources: Total Population Register (country of birth), LISA (sociodemographics), Dwelling Units Register (living conditions), SmiNet (notifiable COVID-19 disease), Swedish Intensive Care Register (ICU admissions), Cause of Death Register (COVID-19 deaths), National Vaccination Register (vaccinations). Records linked using pseudonymized personal ID. Exclusions: 1,131 unknown country of birth; 13,275 missing DeSO neighborhood (73 overlap); 83 missing income; total excluded 14,416 (48% foreign-born; 80% lowest income quartile). Variables: Region/country of birth categorized as Sweden; Nordics (excl. Sweden); EU28/EEA (incl. UK and Switzerland, excl. Nordics); Europe outside EU28/EEA; Middle East; Africa; Asia; North America; South America. Age grouped (20–29 to 100+). Education (Primary, Secondary, Post-secondary, Unknown). Disposable income quartiles. Broad skill level (ISCO-08): 3–4 (managers/professionals/technicians), 2 (clerical/service/skilled trades/plant & machine operators), 1 (elementary), AF (armed forces), X (not elsewhere classified). Household type (cohabitating, single, other). Accommodation (house, apartment, special/student, elderly care, other). Living area per person (quartiles). Neighborhood DeSO population density (quintiles). Region of residence (East, Mid-West, South, West, South-East, North). Vaccination status as time-varying binary (receipt of first dose). Outcomes: COVID-19-related ICU admission and COVID-19-related mortality. Statistical analysis: Poisson regression to estimate incidence rate ratios (RR) with 95% CI, including log person-time offset. Five models: M1 adjusted for age and sex; M2 additionally for sociodemographics (education, income, occupation, household type); M3 for living conditions (housing type, living area per person, neighborhood density, region); M4 for vaccination status; M5 for all factors (age, sex, SES, living conditions, vaccination). AIC used to compare model fit. Vaccination modeled as time-varying (split follow-up at first dose). Temporal analysis by pandemic waves with breaks: 2020-03-01, 2020-07-07 (Wave 1); 2020-10-25 to 2021-02-01 (Wave 2); 2021-02-01 to 2021-06-01 (Wave 3); 2021-12-01 to 2022-05-01 (Wave 4), with end 2022-06-01. Interaction between region/country of birth and time period to estimate wave-specific RRs. Analyses performed in R 4.2.2.

• Descriptive rates (overall period): Age- and sex-standardized ICU admission rates per 100,000 were highest among individuals born in the Middle East (177.8), Africa (166.7), and Asia (157.2). Age- and sex-standardized mortality rates were highest among those born in the Middle East (265.4) and Africa (220.9). • Overall relative risks (age- and sex-adjusted, M1): Migrants had higher risks than Swedish-born for ICU admission: Africa RR 4.1 (95% CI: 3.7–4.6), Middle East RR 4.3 (4.0–4.6), Asia RR 3.7 (3.3–4.1), South America RR 3.6 (3.1–4.1), Europe outside EU/EEA RR 3.5 (3.2–3.8). For mortality: Africa RR 3.1 (2.7–3.6), Middle East RR 2.7 (2.5–2.9), Europe outside EU/EEA RR 2.3 (2.1–2.5), Asia RR 2.3 (2.0–2.6). • Attenuation after full adjustment (M5: age, sex, SES, living conditions, vaccination): ICU admission RRs attenuated by 54.8% for Africa to RR 2.5 (2.2–2.8); by 43.7% for Middle East to RR 3.0 (2.7–3.2); by 36.6% for Asia to RR 2.7 (2.4–3.0); by 40.3% for South America to RR 2.5 (2.2–3.0). Mortality RRs attenuated by 64.6% for Africa to RR 1.8 (1.5–2.0); by 63.9% for Middle East to RR 1.7 (1.5–1.8); by 51.9% for Europe outside EU/EEA to RR 1.7 (1.5–1.8); by 58.8% for Asia to RR 1.6 (1.3–1.8). • Wave-specific patterns: ICU admission disparities were largest in Wave 1 (e.g., Africa RR 8.2 [6.7–10.2], Middle East RR 6.2 [5.2–7.4], South America RR 5.8 [4.6–7.3]) and declined in subsequent waves; by Wave 4, no disparities in ICU admission remained after full adjustment (M5). Mortality disparities were highest in Waves 1 and 3; although less consistently declining across waves, by Wave 4 there were no disparities after full adjustment. • Role of vaccination in Wave 4: Differences in vaccine uptake accounted for most remaining disparities. Example: ICU RR for Europe outside EU/EEA decreased from 2.8 (1.9–3.9) without vaccination adjustment to 1.5 (1.0–2.2) with vaccination adjustment; several groups’ RRs were no longer significantly different from 1 after including vaccination. Vaccination uptake was lower among migrants (e.g., Swedish-born 90.9% vs. Africa 70.5%, Middle East 74.8%, Europe outside EU/EEA 69.3%). • Study totals: n = 7,870,441; COVID-19 deaths = 18,731; ICU admissions = 8,705.

The study demonstrates pronounced inequalities in severe COVID-19 outcomes by country of birth in Sweden, especially during the initial pandemic wave. Socioeconomic disadvantage and adverse living conditions explain a substantial portion of excess risk among many migrant groups, and vaccination uptake explains much of the remaining disparities in the later (omicron) wave. These findings indicate that migrants were particularly vulnerable early on due to higher exposure and limited mitigation, with potential roles for overcrowded housing, high-density neighborhoods, and frontline occupations. By Wave 4, once vaccines were widely available and included in models, disparities in ICU and mortality largely disappeared, underscoring the central role of equitable vaccine uptake. The results address the research question by showing that disparities fluctuated across waves and that their determinants also shifted, with structural conditions dominating early and vaccination coverage being decisive later.

Migrants from the Middle East, Africa, Asia, South America, and Europe outside the EU/EEA experienced much higher risks of COVID-19 ICU admission and mortality than Swedish-born individuals, particularly in the first wave. Disparities diminished over time and were no longer evident in the fourth wave after accounting for socioeconomic factors, living conditions, and vaccination status. Policy implications include targeting social and living conditions that elevate exposure early in a pandemic and ensuring rapid, equitable vaccine access and uptake once vaccines are available. Future research should integrate data on comorbidities, contact patterns, language proficiency, adherence to public health guidance, and cultural factors to better elucidate mechanisms and inform tailored interventions.

• Lack of direct data on comorbidities and health risk behaviors across groups; SES may only partially proxy these factors. • No data on social interaction patterns, language proficiency, adherence to guidelines, or cultural/religious norms, which may influence exposure and outcomes. • Limited infection ascertainment impedes evaluation of differential exposure. • Potential under-ascertainment of COVID-19 deaths could underestimate mortality risk; unlikely to explain observed disparities. • ICU admission may not perfectly reflect disease severity and can be influenced by treatment availability and practices, introducing time-period confounding and selection bias. • Traditional Poisson regression assumes independence of events, which may be violated in infectious disease transmission; estimates are descriptive and not causal. • Moderating effects of socioeconomic status and living conditions were not modeled.