A systematic review and meta-analysis of clinical outcomes following COVID-19 infection in ethnic minority groups in the UK.


                  Objectives
                  This systematic review and meta-analysis evaluated the clinical outcomes of COVID-19 disease in the ethnic minorities of the UK in comparison to the White ethnic group.
               
                  Methods
                  The study included adult residents of the UK with confirmed COVID-19. Outcomes evaluated were mortality, ICU admission, and invasive mechanical ventilation need in ethnic minorities compared to people from a White background. Medline, Embase, Cochrane, MedRxiv, and Prospero were searched for articles published between May 2020 to April 2021. Risk of bias was evaluated using the Newcastle-Ottawa Scale checklist. PROSPERO ID: CRD42021248117.
               
                  Results
                  Fourteen studies (767177 participants) were included in the review. In the adjusted analysis, the pooled Odds Ratio (OR) for the mortality outcome was higher for the Black (1.83, 95% CI: 1.21-2.76, number of studies: k=6), Asian (1.16, 95% CI: 0.85-1.57, k=6), and Mixed and Other (MO) groups (1.12, 95% CI: 1.04-1.20, k=5) in comparison to the White group. The adjusted and unadjusted ORs of intensive care admission were more than double for many of the ethnic minorities (OR Black 2.32, 95% CI: 1.73-3.11, k=5, Asian 2.34, 95% CI: 1.89-2.90, k=5, MO group 2.26, 95% CI: 1.64-3.11, k=4). In the adjusted analysis of mechanical ventilation need, the ORs were similarly significantly raised (Black group 2.03, 95% CI: 1.80-2.29, k=3, Asian group 1.84, 95% CI: 1.20-2.80, k=3, MO 2.09, 95% CI: 1.35-3.22, k=3).
               
                  Conclusion
                  This review found that in the UK, Black, Asian, and MO groups suffered from increased COVID-19 related disease severity and mortality compared to the White majority.
               


Introduction
In the UK, there were 152490 deaths (192 deaths/100000 population), 471045 hospital admissions, and 4717811 confirmed cases (7062 cases per 100000 population) due to the COVID-19 pandemic between March 2020 to June 2021. 1 There are growing concerns in the UK that people belonging to many, if not all ethnic minorities, have been disproportionately impacted by COVID-19. [2][3][4] A systematic review study conducted between December 2019-August 2020 found that Asian people had a higher risk of intensive care admissions and death. 5 However, another review of COVID-19 patients did not find that ethnicity was associated with the worst outcomes. 6 This present study was adapted from the systematic reviews by Sze et al., 2020 5 and Raharja et al., 2020. 6 Hence, it builds on the work of these two earlier reviews. 5,6 Since the demographics, socioeconomic issues, healthcare policies and systems of each country are unique, a UK-centric review was needed to understand if her ethnic minority groups faced a greater risk of adverse outcomes from COVID-19 or not.
In order to reduce the impact of COVID-19 on the population, pandemic related research was prioritized and classified as urgent public health research by the National Institute for Health Research (NIHR). 7 Following the government's call and support for research in this area, numerous studies were conducted. However, even after this call, there were no published systematic review and meta-analysis studies on the impact of COVID-19 on the ethnic minorities in the UK. Therefore, an up-to-date systematic review of UK-based studies was urgently needed to quantify the health inequalities faced by ethnic minority groups concerning COVID-19 This research aims to assess the clinical outcomes of COVID-19 amongst ethnic minorities in the UK.

J o u r n a l P r e -p r o o f 2 Methods
The review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria. 8 The PRISMA checklist is available in the supplementary material (S1 Appendix). The protocol was registered with PROSPERO international prospective register of systematic reviews with the ID.

Information Sources
The reference lists of previous reviews were searched for relevant studies published from January 2020 to August 2020. 5,6 The database searches (Ovid Medline, Ovid Embase, Ovid Cochrane, MedRxiv, and Prospero) were restricted to one year from May 2020 to April 2021.

Search Strategy
The Population, Exposure, Comparator, and Outcomes (PECO) framework was used to formulate the criteria for study selection. The target sample included adult population aged 18  Interventional studies, systematic reviews, observational studies including casecontrol studies, and cohort studies were included, alongwith non-peer-reviewed studies as this is a rapidly evolving field.
Conference abstracts, commentaries, cross-sectional studies, reports, editorials, nonsystematic review articles, case reports, late-breaking abstracts, studies without a comparator group, and papers whose full text was unavailable, were excluded. Risk of infection only studies were excluded. Studies were restricted to those in the English language. Studies from the same population, with similar outcomes, were reviewed and one relevant study was included; whilst the rest were excluded, as this may have created a duplication of data. Studies that grouped all ethnic minorities as one, were excluded. Specialist librarians were asked to review the search strategy with the keywords COVID-19, ethnic minority, and the UK. The search strategy was based on the search originally conducted by Sze et al., 2020 5 and Public Health England (PHE), 2020 and was adapted for this review by the addition of 'UK' as a key term during the searches. 2 The detailed search strategy for each database was provided in the research proposal. 9 The search terms were tailored for each database and the searches were run separately for each database to enhance sensitivity. The search period defined was between May 2020 to April 2021.

Selection Process
J o u r n a l P r e -p r o o f Two reviewers independently screened the titles and the abstracts of the studies; and excluded non-relevant studies. Full texts of the remaining studies were retrieved and reviewed for inclusion in the study against the selection criteria defined earlier. Any disagreements between the researchers were resolved through discussion.

Data Collection Process
One researcher (SS) extracted data from the eligible studies and assessed the risk of bias. Data extraction was checked by a second reviewer (IA). Authors of relevant studies were also contacted for clarity on missing data.

Quality Assessment
The Newcastle-Ottawa Scale (NOS) (S2 Appendix) was employed to assess the risk of bias in the included studies. 12 A NOS score of 7-9 is classed as low risk of bias, a score of less than 5 as high risk of bias; and a score of 5-6 as moderate risk of bias. 13 One researcher (SS) carried out the quality assessment at the study and outcome level using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach. 14 Publication bias was not assessed as there were fewer studies in the adjusted analysis. The funnel plot may not detect publication bias when the number of studies is small. 15

Data Synthesis
The raw counts for the various outcome variables were used to calculate the RR and 95% Confidence Intervals (CIs). Studies that did not use the White group as a comparator group were excluded from the meta-analysis. As all the studies were observational designs, a Der Simonian-Laird Random-Effects Meta-analysis (REM) was conducted for all outcomes. For rare outcomes, OR was assumed to be equal to J o u r n a l P r e -p r o o f RR, and RR was assumed to be equal to HR. 5,6 Mortality in hospital-based studies was not a rare outcome. However, in population-based studies it was rare. The studies which provided Adjusted Odds Ratio (AOR) were pooled together in one group and those which provided Adjusted Hazard Ratio (AHR) were separately pooled together in another group (if the outcomes were not rare). Thus, adjustments were made, whilst extracting, for study characteristics of each included study. In studies where raw data was missing; the authors were contacted to obtain this. In order to include studies with missing data in the meta-analysis, the unadjusted HR/RR were combined using the inverse variance method. Using this method, the pooled risk estimates were calculated separately for each ethnic group and a summary statistic was provided. The results were written in tabulated form and as forest plots. Excel and RevMan were used to analyse and tabulate the data. Origin 2021b was used to convert graphical data into tabulated form. The statistical heterogeneity was explored by calculating the I 2 statistic using RevMan and by looking at the overlap of CIs in the forest plots. A subgroup analysis was conducted to explore differences in risk estimates across the subgroups. 16 Studies based only on ICU patients, general hospital patients, and the general population were analysed separately. A sensitivity analysis was conducted based on only peer-reviewed studies and a separate analysis was conducted based on studies with low risk of bias. 16

Results
The search on Medline, Embase, MedRxiv, Cochrane, and Prospero yielded 939 studies on 14-15 April 2021. Fourteen studies were selected to be included in the systematic review as shown in S1 Fig and S1 Table. J o u r n a l P r e -p r o o f

Study Characteristics
The 14 studies included a total of 767177 participants. The study characteristics are given in Table 1. [17][18][19][20][21][22][23][24][25][26][27][28][29][30] As all the studies were observational, so no adjustment was needed J o u r n a l P r e -p r o o f   Fourteen studies were included in the systematic review and 12 in the meta-analysis.
Thirteen studies provided data on risks of mortality due to COVID-19 in the various ethnic groups, six on ICU admission, and four on IMV need.

Quality of Studies
The risk of bias (ROB) was low, average NOS score was 7.7 (ranging from 4-9). Nine studies had a low risk of bias score on the NOS, four studies had a moderate risk of bias score on the NOS and one study had a high risk of bias as shown in S2 Table. The low scores were largely due to studies that failed to adjust for confounding factors.
Most studies were hospital-based, and such populations are somewhat representative J o u r n a l P r e -p r o o f of the average COVID-19 patient in the community and did not necessarily reflect the wider community. Also, there might be bias introduced in such hospital-based studies as people in some ethnic groups might be reluctant to seek medical advice.

Mortality
The unadjusted OR, RR, AOR, and AHR are shown in S2 Fig, Figs 1

Intensive Care Admission
Six studies provided data about ICU admissions for the various ethnic groups. Five of these studies were suitable for aggregating the raw outcomes and for pooling the unadjusted risk estimates. These studies included a total of 71791 participants who were admitted to critical care units in the UK. Eighty percent of them were White, 5% were Black, 9.5% were Asians and 8% were MO. The unadjusted and adjusted analyses for ICU admission are shown in Fig 3 and S4 Fig. From S4 Fig.,  and patients admitted to ICU with community-acquired pneumonia (non-COVID controls). 27 The study found that the cases with COVID-19 had statistically significantly fewer White (p= 0.012) and more Asian cases (p= 0.002). 27 J o u r n a l P r e -p r o o f   The studies which were included in the IMV need and ICU admission analysis are all low risk of bias studies, published, and hospital-based studies and so further sensitivity analyses were not conducted. The heterogeneity was high for most of the outcomes.

Quality of Evidence Assessment
As best evidence regarding risk factors is usually obtained from observational studies, so the evidence in this review was started with high ratings as advised by Foroutan et al., 2020 (for prognostic studies). 31 The overall GRADE assessment indicated a high level of confidence for all outcomes, except for the mortality outcomes for the Asian group which was moderate and was downrated due to inconsistency, as shown in the summary of findings in S5 Table. 4

Discussion
There was heterogeneity in the populations, settings, methodology, and statistical analysis. The meta-analysis was still conducted since this degree of heterogeneity had been reported by other reviews and conducting the latter was beneficial.
The results indicate greater disease severity in the Black, Asian, and MO groups necessitating ICU admission and IMV provision. Overall, it can be said that ethnicity is a risk factor for worse prognosis in ethnic minorities and they do suffer from increased disease severity. Several studies and data from Office for National Statistics (ONS) and Intensive Care National Audit and Research Centre (ICNARC) also validate this finding of the worst outcomes or ethnic minorities in the UK. Some reviews on this topic have combined data from all over the world. 5,6 The results from these cannot be extrapolated to the UK, as the healthcare systems, health inequalities issues, and ethnic make-up vary from country to country. A more recent review has attempted to address this issue by analysing the COVID data regionally, but they did not include any data on ICU admissions nor conduct any adjusted analysis on mortality in Europe. 38  The odds of mortality for hospitalised COVID cases were raised for the Black, Asian, and MO groups in the adjusted analysis. The result from this review is significant as it shows that after hospital admission, there are additional factors in play, which increase the risk of mortality. So, the difference in mortality cannot be attributed to differential access to healthcare facilities or increased risk of infection alone. The odds of mortality due to COVID for the general population were raised for the Black ethnic group similar to findings by ONS and Mathur. 32, 34 Increased prevalence of comorbidities and raised C-Reactive Protein (CRP) levels, a non-specific marker of inflammation, maybe the cause of the greater severity of COVID-19 in Black and Asian communities. Obesity has been identified as an important risk factor associated with morbidity and mortality due to COVID-19. 25,45,46 It has been proven that Black people with obesity suffer the worst outcomes and the effect of this association is attenuated in them. 25 Another study reported that the association between obesity and mortality was stronger for non-Whites (especially Black and South Asian people) compared to White people (p-value= 0.002). 46 The highest obesity rates have been observed in Black people (68%) compared to the other groups (White 64%, Asian 60%, average 63% adults in the J o u r n a l P r e -p r o o f UK). 47 DM and HTN were three times more common in cases who died due to COVID-19. 45 HTN is three to four times more prevalent in Black people in the UK compared to the rest of the population. 48 18 Lower socioeconomic status is a risk factor for raised CRP levels and it has been hypothesized that stress may be a mediator in this. 51 It has also been hypothesized that chronic inflammation linked to insulin resistance, obesity, cardiovascular disease, stress, and chronic infections in the Black and other ethnic minority groups can trigger a cytokine storm which is associated with the severe form of COVID-19 disease. 44 A cytokine storm is an abnormal dysregulated inflammatory response diagnosed by the presence of respiratory distress, and hyper inflammation (raised CRP levels more than 100mg/L or raised ferritin levels) in COVID-19 cases. 55 This is the first meta-analysis that analyses IMV data, and ICU data from the UK. The

Strengths
This is the first systematic review and meta-analysis conducted to assess the burden of disease faced by ethnic minorities in the UK. The strength of the research lies in its comprehensive analysis of relevant databases for published and pre-print articles.
Although the heterogeneity was quite high, this was to be expected with observational studies that had a very large number of participants. 32 The heterogeneity was explored by conducting a subgroup and sensitivity analysis.

Limitations
A broad ethnic classification was used in this review. The Asian group included very diverse subgroups, each of which has now been shown to have different risk profiles.

J o u r n a l P r e -p r o o f
This review was limited to adults so the results could not be generalised to children.
The PCR test for COVID-19 has a high false-negative rate, which led to some cases being wrongly classified as non-COVID. As this review concentrated on UK-based studies, the results were less generalisable to other countries. It has been noted that many participants' ethnicities have been put down as 'Other', and this may have created erroneous results for the MO group. 22 The representativeness of study populations would be a limitation of the review. The variables used for the adjustment differed greatly from one paper to another. This is a possible limitation as the amount and nature of adjustment could affect the results.

Policy Implications
These results have urgent implications for formulating a COVID-19 response strategy  43 The NICE guidance on possibly excluding cases with cardiovascular disease especially angina patients from critical care; disproportionately influences ethnic minorities as Asian and Black people are more prone to heart disease. 43 The risk assessment of being employed in healthcare does not allow for the redeployment J o u r n a l P r e -p r o o f of staff based on ethnicity alone. 43 As ethnicity has been proven now to be associated with increased mortality, these workers should be given the choice of being redeployed in non-patient-facing roles. 43

Implications of Research
This review needs to be upgraded to a living systematic review so that the changing pandemic risks can be identified in the various ethnicities as the pandemic progresses.
There is a need for more epidemiological population-based studies to assess the true risk experienced by various ethnic groups with regards to the worst clinical outcomes.
The dataset needs to be large enough to appreciate the risk in the various sub-groups.
Ethnic minority focused healthcare research concerning COVID-19 treatments can help to reduce the health inequalities. 44 Research should be conducted to assess why COVID-19 related health inequalities exist for some ethnic minorities.

Conclusion
It can be concluded that the Black, Asian, and MO groups faced the worst outcomes with regards to COVID-19 in the UK. These findings are of immense public health importance and should be used to help formulate policy concerning COVID-19 and reducing socioeconomic disparities.
conflict of interest exists in the submission of this manuscript. The content of the manuscript has previously appeared online, as preprint on MedRivx. Librarians at the University of Manchester and Public Health England (PHE) helped to formulate the search strategy. Librarians from PHE also helped to find the full text of articles.
SS would like to thank her family, especially her sister Hina Siddiq who funded her studies, Sarah Siddiq who edited this article, and her children who put up with her whilst she worked on this paper.

Funding:
This research received no specific grant from any funding agency, commercial or notfor-profit sectors.

Ethical approval:
Not needed.
J o u r n a l P r e -p r o o f