Hospitalisation rates differed by city district and ethnicity during the first wave of COVID-19 in Amsterdam, The Netherlands

The first cases of coronavirus disease 2019 (COVID-19) were reported at the end of 2019 in Wuhan, China. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus causing COVID-19, has since rapidly spread across the globe. The World Health Organization (WHO) consequently declared COVID-19 a global epidemic on 11 March 2020. By the end of June 2021, over 180 million cases and almost 4 million COVID-19 related deaths had been reported worldwide [1].

In the Netherlands, COVID-19 was declared a Group A notifiable disease under the Public Health legal act on 28th January 2020 [2]. This required clinicians and laboratories to immediately notify the regional public health service (PHS) every suspected and/or confirmed case of COVID-19. The first confirmed case of COVID-19 on 27 February 2020 had an epidemiological link to Northern Italy. Initial sporadic clusters led to a subsequent nationwide spread of the virus, including in the capital city of Amsterdam. From 12 to 15 March, the Dutch government initiated a series of restrictions [3] which were gradually lifted from 11 May onwards. Up to 1 June 2021, the Netherlands had reported approximately 46,000 confirmed cases and over 6000 COVID-19 deaths [4].

The regional PHS of Amsterdam acts upon notifications of infectious diseases including COVID-19 for the wider Amsterdam-Amstelland area, consisting of 6 municipalities including Amsterdam, with a total regional population of about 1,06 million. Anecdotal reports from hospital staff in April 2020 suggested that a disproportionate number of patients of ethnic minority background had been admitted to hospitals and intensive care units (ICU) in Amsterdam. This echoed reports from the United Kingdom [5] (UK) and United States of America [6] (USA) at the time, which suggested that age- and sex-standardised mortality rates were highest among socio-economically deprived groups and ethnic minority groups; findings that were later supported by a systematic review [7]. Early analyses also demonstrated that individuals with a migration background in the Netherlands had higher rates of excess mortality (i.e. higher than would be expected in ‘normal’ conditions) compared to individuals of Dutch origin, in particular in those with a non-Western migration background [8]. Amsterdam is an increasingly ethnically diverse city with more than half the population having a migration background [9]. Disparities in the distribution of communicable and non-communicable diseases have previously been demonstrated in the city between individuals with and without a migration background [10], even when matched for age- and socio-economic status (SES) [11]. In addition, health inequalities have been reported between the peripheral city districts of Amsterdam with lower average income [12] and the central, higher-income city districts of the Centre, West and East districts, with poorer health outcomes being reported in peripheral districts [13]. Investigating whether such disparities also exist for COVID-19 burden is crucial to developing targeted prevention strategies.

We therefore aimed to examine differences in burden of COVID-19 hospitalisations between city districts and ethnic groups in the city of Amsterdam, using routinely collected surveillance data from 29 February to 31 May 2020.

Between 29 February 2020 and 31 May 2020, 2326 COVID-19 cases were notified in Amsterdam, of which 596 (25.6%) hospitalisations and 287 (12.3%) deaths. The number of new cases peaked mid-March, shown by date of symptom onset in Fig. 1 and by notification date in Supplementary Figure S1. Characteristics of COVID-19 cases by hospitalisation status are presented in Table 1. Median age of hospitalised cases was higher (64 [IQR 51–73]) than cases for which hospitalisation status was non-hospitalised or unknown (54 [IQR 34–74]). The majority of hospitalised cases were male (361/596 [60.6%]), while most cases overall were female (1346 [57.9%]). 814 (35.0%) of all notifications were registered as health care workers, of whom 314 (38.6%) worked in a long-term care facility.

Of the 2326 cases, 2002 (86.1%) could be matched to the registration database, including 526/596 hospitalisations (88.2%) and 235/287 (81.8%) deaths (Supplementary Figure S2, Supplementary Table S1). Table 2 shows hospitalisation DSR per 100,000 population by city district and migration background in the matched population. Hospitalisation DSR were almost two-fold higher in individuals living in peripheral city districts compared to central districts (RR = 1.82, 95%CI = 1.65–1.99), with 36.87 additional hospitalisations per 100.000 population (RD = 36.87, 95%CI = 25.79–47.96). The difference between peripheral and central city districts became increasingly apparent over time (Figure S3). Further stratification by individual city district is presented in Supplementary Table S2.

Hospitalisation DSR were more than two-fold higher in individuals with a non-Western migration background compared to ethnic-Dutch individuals (RR = 2.36, 95%CI = 2.17–2.54), with 57 additional hospitalisations per 100.000 population (RD = 57.05, 95%CI = 43.34–70.75). No statistically significant difference was observed between individuals with a Western migration background and ethnic-Dutch individuals. Further stratification by specific large non-Western ethnic groups living in Amsterdam (Dutch Antillean, Moroccan, Surinamese, Turkish and Ghanaian) is presented in Supplementary Table S3. Compared to the ethnic-Dutch population, the hospitalisation RR were highest among individuals of Ghanaian (RR = 4.25, 95%CI = 3.31–5.19; RD = 136.75, 95%CI = -29.83–303.33) and Turkish ethnic origin (RR = 3.08, 95%CI = 2.72–3.43; RD = 87.31, 95%CI = 44.51–130.11). In our additional age-stratified analysis, we found that individuals < 60 years with a non-Western migration background had a three-fold higher DSR compared to ethnic-Dutch individuals < 60 years (RR = 3.18, 95%CI = 2.85–3.52), while the DSR was reduced but remained almost two-fold higher in individuals ≥60 years (RR = 1.83, 95% CI: 1.57–2.09) (Supplementary Tables S4 and S5).

After stratifying by city district, the DSR was highest in non-Western residents of peripheral districts (Supplementary Figure S4): three times greater than in ethnic-Dutch residents of central districts RR = 3.13, 95%CI = 2.88–3.38; RD = 76.64, 95%CI = 57.38–95.89) (Fig. 2).

In multivariable Poisson regression, non-Western migration background, living in a peripheral city district, male sex and older age were associated with hospitalisation (Table 3). No statistically significant interactions were found between migration background and city district (P = 0.08, Supplementary Table S6), migration background and age (P = 0.33, Supplementary Table S7) and migration background and sex (P = 0.45, Supplementary Table S8).

We identified geographical and ethnic disparities in the burden of COVID-19 in Amsterdam during the first wave of COVID-19. Specifically, people living in peripheral, lower-income city districts (New-West, North or South-East) with a non-Western migration background had three-fold higher hospitalisation rates compared to ethnic-Dutch individuals living central, higher-income city districts (Centre, West, South or East).

Previous research has demonstrated similar health disparities between city districts in Amsterdam [13]. A quadrennially repeating comprehensive health survey has consistently shown that residents of peripheral city districts are more likely to suffer from comorbidities and be overweight or obese than residents of central districts [13]. As chronic lifestyle conditions such as type 2 diabetes mellitus and cardio-vascular disease have been shown to increase the risk of developing severe COVID-19 symptoms [20], this suggests that the higher rate of COVID-19 hospitalisations seen in these city districts might be partially explained by a higher prevalence of chronic comorbidities. In addition, the peripheral districts have been shown to have a higher percentage of vulnerable residents [13], based on a vulnerability score that encompasses income, job security and healthcare expenditure. The interplay between SES and risk of both SARS-CoV-2 infection and severe disease has been previously documented [21]. For example, individuals with a lower SES may be more likely to work as essential frontline workers – including healthcare workers and those working in sectors such as cleaning, construction, and transport [22] – with a reduced opportunity to work from home [23]. Lower SES may also lead to a greater reliance on public transport, larger multi-generational households and smaller house size per person [24], all of which may increase the risk of exposure to infection [22, 25]. We therefore postulate that the greater burden of COVID-19 hospitalisation seen in these districts is due to a complex interaction between different factors. Indeed, when restricting to individuals < 60 years of age in our analyses, the hospitalisation RR in those with a non-Western migration background compared to the ethnic-Dutch group remained and was even accentuated. This suggests that the differences in hospitalisation rates cannot be fully explained by an increased tendency of non-Western migrants, compared to ethnic-Dutch elderly, to be treated in hospital instead of receiving palliative care in the community setting or in nursing homes 26. The higher RR in younger age groups might imply that increased exposure secondary to having a public-facing occupation or higher levels of other causes of increased exposure among those with a non-Western migration background may have played an important role. Further elucidating the causal factors underlying SES disparities in COVID-19 burden requires further research.

In addition, our analysis demonstrated that individuals with a non-Western migration background, including groups with a Moroccan, Turkish, Surinamese and Ghanaian background, had a higher COVID-19 burden, even when stratifying by city district. This is in line with previous findings from the UK and USA and, more recently, other high-income countries such as Norway [27]. In the UK, the risk of death among those with a COVID-19 diagnosis was twice as high for people of Bangladeshi ethnicity compared to people of White ethnicity, after adjusting for sex, age, deprivation and region [21]. Additional studies demonstrated that, even when adjusting for age, sex, comorbidities and several SES-related determinants (though notably not public-facing occupation), hospitalisation and ICU admission rates among Black and Asian Minority Ethnic individuals were higher compared to ethnically White individuals [28], suggesting that ethnicity plays an independent role in explaining these disparities. In the US, age-adjusted hospitalisation rates were 5.3 and 4.7 times higher in American Indian or Alaska Native and Black or African American persons respectively, compared to non-Hispanic white persons; although this analysis was not adjusted for social inequalities [6]. Whilst these findings consistently demonstrate that the burden of COVID-19 varies by ethnicity, the contribution of underlying cultural norms, health literacy, and differences in health-seeking behaviour has yet to be revealed.

Considering our findings substantiate reports of increased COVID-19 burden among ethnic minority groups in other countries, it is concerning that information on ethnicity is often not collected in routine surveillance systems in many countries [29]. This results in these important differences in disease burden being concealed within the data, hindering the ability to set up targeted initiatives to reach particularly vulnerable populations. As stated by Tai et al. [30], further research on the interplay between inherent social inequalities and ethnicity is clearly required to ensure optimal surveillance of the impact of COVID-19 on vulnerable groups, to disentangle whether the increased burden is the result of a higher infection rate, the severity of disease manifestation, or both, and to inform and develop appropriate and accessible prevention strategies. Indeed, local prevention teams have now been deployed in Amsterdam to help reduce ethnic inequalities in COVID-19 burden, using strategies such as mobile test buses in neighbourhoods with a high rate of diagnosed COVID-19 infections per population and information provision on preventive measures tailored to specific groups.

An important limitation of our study is that the surveillance data paint an incomplete picture of the first wave of the outbreak, as cases were underreported due to selective testing and data collection was limited. We used the hospitalisation rate per 100,000 population as a marker of outbreak progression, but this limits the distinction that can be made between risk of infection and risk of severe disease requiring hospital admission. In addition, hospitalisations and deaths among already notified cases may also have been underreported, and we were unable to match all notifications to the municipal register. Furthermore, absence of key individual socio-demographic, socio-economic, and clinical characteristics limits the inferences that can be made about causal factors on an individual patient level. For example, we used city district as an imperfect proxy for SES, but SES at the community or individual level within each city district may have been different. By further stratifying groups by migration background and complementing this with qualitative research (for example, through community focus groups) more insight can be gained into which community-specific targeted prevention strategies may help minimise the disproportionate distribution of COVID-19 in Amsterdam.

Our study is the first in the Netherlands to link surveillance data with registration data on migration background to demonstrate the unequal distribution of the burden of COVID-19 within the city of Amsterdam. We show that substantial differences in COVID-19 hospitalisation rates exist between city districts and ethnic groups in Amsterdam. Our findings corroborate reports from other high-income countries to suggest that public health efforts worldwide must be focussed on mitigating further impact of COVID-19 upon communities at highest risk of both infection and serious disease. Action must be taken to strengthen targeted prevention strategies which address the needs of affected communities.