COVID-19 risk strati � cation tools should incorporate multi-ethnic age structures , multimorbidity and deprivation metrics for air pollution , household overcrowding , housing quality and adult skills

Marina A. Soltan (  M.Soltan@bham.ac.uk ) University of Birmingham and University Hospitals Birmingham NHS Foundation Trust, Birmingham, UK Justin Varney Birmingham City Council Benjamin Sutton University Hospitals Birmingham Foundation NHS Trust, Birmingham UK Colin R. Melville School of Medical Sciences, University of Manchester, Manchester, UK Sebastian T. Lugg University of Birmingham and University Hospitals Birmingham NHS Foundation Trust, Birmingham, UK Dhruv Parekh University of Birmingham and University Hospitals Birmingham NHS Foundation Trust, Birmingham, UK Will Carroll University Hospitals North Midlands, Newcastle Road, Stoke on Trent, UK Davinder Dosanjh University Hospitals Birmingham Foundation NHS Trust, Birmingham UK David R. Thickett University Hospitals Birmingham Foundation NHS Trust, Birmingham UK


Introduction
Data published by the Intensive Care National Audit and Research Centre (ICNARC) shows that Black Asian and Minority Ethnicity (BAME) patients account for 34% of critically ill patients with SARS-CoV-2 infection (COVID-19) despite constituting 14% of the UK population. (1) Internationally researchers have called for studies to explain the increased deterioration among BAME subgroups. (2) Understanding the risk factors is essential for informing the development of risk strati cation tools which triage patients to the appropriate level of care.
A recent study of COVID-19 pneumonia patients (n=279) suggests that the CURB65 tool, validated solely in community acquired pneumonia, is a potentially unreliable risk strati cation tool for COVID-19 pneumonia despite its' widespread use in clinical practice. (3) Studies have not thus far explored the resilience of CURB65 for risk stratifying COVID-19 pneumonia among BAME subgroups. Traditional clinical training has reinforced that the unmodi able risk factor of age predisposes to adverse outcomes with little regard to epidemiological multi-ethnic age structures. BAME subgroups have younger age structures (4) which predispose to a lower CURB65 score and are susceptible to different underlying clusters of disease.
Furthermore, BAME patients are more likely than Caucasians to be hospitalised with COVID-19 from the most deprived IMD regions. Understanding these risk factors is invaluable for informing the development of risk strati cation tools which re ect risk factors to which BAME subgroups are potentially disproportionately predisposed.
This multi-centre study aims to test resilience of the CURB65 score in identifying patients at risk of deterioration disaggregated by BAME subgroup and explore the extent to which social determinants of health, including IMD sub-indices with indicators for household overcrowding, air pollution, housing quality and adult skills deprivation, act as risk factors for presentation with multilobar pneumonia, ITU admission and adverse outcomes among hospitalised COVID-19 patients disaggregated by BAME subgroup.

Design and setting
A multi-centre cohort study of hospitalised COVID-19 patients was performed to explore social determinants of health, including IMD sub-indices, as risk factors for presentation with multilobar pneumonia, ITU admission and completed hospitalised episode outcomes. Patient population COVID-19 patients (>16 years old) admitted to 4 hospitals across the West Midlands, University Hospitals of Birmingham, between 1 st December 2019 and 1 st September 2020 were included. Diagnosis was con rmed by PCR analysis of a combined nose and throat swab in accordance with Public Health England guidance. (9) Patient management See online supplement 1.

Data collection and scoring analysis
Data collected from the hospital informatics system included: demographics (ethnicity, age, postcode), admission details, comorbidities, clinical metrics (observations, blood tests), imaging, ITU admission details and completed hospitalised episode outcomes (discharge or death). Chest X-rays were reported by radiologists within 12 hours of being undertaken. For patients presenting with pneumonia, severity was scored using CURB65 which determines 30-day mortality risk based upon: presentation with new onset confusion, urea (>7mmol/L), respiratory rate (³30 breaths/minute), blood pressure (<90mmHg systolic or £60mmHg diastolic) and age (³65 years). (11)

Statistical Analysis
Baseline characteristics were presented as mean and standard deviation (SD) for continuous variables and median and interquartile range (IQR) for nonparametric data. Normality was assessed by Shapiro-Wilk. For categorical and ordinal variables with non-parametric distribution, Fisher's exact test and Mann Whitney U test were used respectively for comparisons between two groups. Multivariate analysis to predict mortality was performed using stepwise logistic regression with conservative criteria for entry or exit from the model of 0.1. Factors found to be signi cant in univariate analysis were included as independent variables. The Hosmer and Lemeshow goodness-of-t test was performed to evaluate the adequacy of the logistic regression model. Matched case-control analyses were implemented to explore underlying multimorbidity among BAME subgroups; controls were Caucasians matched by age, gender and deprivation.
Statistical analyses were carried out using SPSS Statistics V.24 and GraphPad Prism 8.

Ethics statement
Data were entered anonymously in accordance with national and local audit guidance. Health Research Authority (HRA) guidance was followed and ethical approval was not required based on the HRA Decision tool (Online Supplement 2).

Results
Included participants 3671 consecutive patients were eligible for inclusion. 716 patients did not meet inclusion criteria due to: ongoing hospitalisation on 1 st September 2020 (n=55), age <16 (n=22), attendance as an elective admission (n=371) or attendance without admission (n=267). Of these patients (n=2955), those without listed postcodes (n=301) or postcodes not returning deprivation metrics (n=8) could not be included in the analysed group (n=2646). Figure 1 shows the CONSORT diagram.

Study population
The study population is outlined in table 1. The median age of all patients was 76.0 (24.0). Males (54.8%) were hospitalised more than females (45.2%).
Hospitalised admissions by deprivation sub-index are depicted in gure 2 and ITU admissions by deprivation sub-index are depicted in gure 3. The proportion of patients admitted to hospital from the highest (sub-indices 1 and 2) deprivation forms were as follows: Wider BHS (59.0%), Adult Skills (43.6%), Indoor LE (42.3%) and Outdoor LE (56.5%).
Stepwise multiple regression, including the above variables, identi ed 7 variables which were independently associated with mortality: age, sex, cirrhosis, obesity, CCI score, presentation with multilobar pneumonia and ITU admission (table 3). As demonstrated below, BAME patients were more likely than Caucasians to exhibit 5 of 7 variables: (1) male, (2) obesity, (3) higher CCI scores than age, sex and deprivation matched Caucasian controls, (4) presentation with multi-lobar pneumonia and (5) ITU admission To understand the risk factors for presentation with multi-lobar pneumonia, ITU admission and mortality affecting each ethnic group, BAME subgroups were disaggregated.
Ethnicity: taking a closer look Demographics : age structures, sex and ethnicity BAME patients were more likely to be male (OR 1.199(1.009-1.426); p=0.042) and present age<65 (OR 4.846(4.020-5.843); p<0.001) than Caucasians. Caribbean and Caucasian subgroups presented older (median age>65) whilst Indian, Pakistani, African, Chinese and Bangladeshi subgroups presented younger (median age<65); this is consistent with UK population age structures. (4) Comorbidity, multimorbidity and ethnicity Comorbidities by ethnic subgroup are shown in table 4 and Online Supplement 4. BAME patients were more likely to be obese (OR 1.640(1.363-1.975); p=0.001). Caucasians were more likely to have cirrhosis (OR 7.778(1.447-81.10); p=0.014). Although the overall proportion of Caucasians with multimorbidity appears higher than BAME subgroups (table 4), CCI scores in every BAME subgroup were higher than age, sex and deprivation matched Caucasian controls (table 5) and the average number of comorbidities among African, Pakistani and Caribbean patients was higher than age and sex matched Caucasian controls (table 5).

Discussion
Among hospitalised COVID-19 patients presenting with pneumonia and lower CURB65 scores (0-1), mortality was higher among BAME patients than Caucasians; Africans were at highest risk, followed by Caribbean, Indian and Pakistani. BAME subgroups were more likely to be admitted with higher CCI scores than age, sex and deprivation matched controls and from the highest IMD sub-indices of at least one deprivation form. Patients admitted from regions of highest Indoor LE deprivation, Outdoor LE deprivation, Wider BHS deprivation and Adult Skills deprivation were more likely to present with multilobar pneumonia and require ITU admission, which are in themselves independent mortality risk factors. This may explain the higher ITU admissions among BAME patients reported by ICNARC (1) and ONS data reporting higher age standardised mortality rates among patients in the most deprived IMD areas. (6) CCI scores, presentation with multi-lobar pneumonia, ITU admission, age, sex, cirrhosis and obesity were independent risk factors for mortality. BAME patients were more likely than Caucasians to exhibit 5 of these 7 risk factors.
A recent study of COVID-19 pneumonia patients (n=279) found that, as a largely physiological assessment, CURB65 is an unreliable mortality risk tool in COVID-19 pneumonia. (3) CURB65 mortality risk does not take account of hidden risk factors which appear to disproportionately affect BAME subgroups including: male, obesity, multimorbidity, presentation with multilobar pneumonia, household overcrowding, air pollution, housing quality and adult skills deprivation. Furthermore, national epidemiological data shows younger age structures among BAME subgroups which are also re ected in this study and which potentially predispose to an underscored CURB65 among BAME subgroups resulting in possible triage to an inappropriate level of care whilst clinicians are left falsely reassured regarding the severity of presentation and risk of deterioration.
This study nds that air pollution deprivation increases the odds of presentation with radiological multilobar pneumonia and ITU admission among COVID-19 patients. Pollutants compromise the host's immune response against invading pathogens in the respiratory tract. (12) Chronic exposure to nitrogen dioxide and sulphur dioxide concentrations are associated with incidence of pneumonia (13) whilst particulate matter increases the activity of angiotensin-converting enzyme 2 receptors on cell surfaces (14), thus enhancing COVID-19 uptake by the lungs. BAME patients are more likely than Caucasians to be exposed to chronic air pollution on account of residing in regions of highest air pollution deprivation. (15) In this study all BAME subgroups apart from Chinese and Mixed were more likely than Caucasians to be admitted from regions of highest air pollution deprivation. Minimising air pollution deprivation inequalities is essential in reducing the disease burden of community acquired pneumonia (16) including COVID-19.
Furthermore, this study identi es household overcrowding deprivation and poor housing quality as potential risk factors for presentation with radiological multilobar pneumonia and ITU admission among COVID-19 patients. BAME subgroups are more likely than Caucasians to live in the most overcrowded and poorest quality housing. (15) The UK Biobank study has reported that patients with a COVID-19 positive test were more likely to live in crowded households (17) and it is well established that household overcrowding and housing quality failing to meet the Decent Homes Standard are associated with a higher incidence of non-COVID-19 pneumonia and increased risk of disease transmission. (18) Minimising household overcrowding and improving housing quality is essential for limiting the exposure to and spread of toxigenic species including bacteria, fungal and viral pathogens. (19) Moreover, this study nds that adult skills deprivation including limited English language pro ciency and low adult quali cations are potential risk factors for presentation with radiological multilobar pneumonia and ITU admission. Coronaviruses cause pneumonia which gradually progresses to further lung zones between 2 to 14 days. (20) It is possible that the predominantly English language messaging may have contributed to later presentation in patients from regions of highest adult skills deprivation. In this study, Caribbean, Pakistani, African, Bangladeshi, Mixed and any other non-Caucasian ethnic subgroups were more likely than Caucasians to be admitted from regions of highest adult skills deprivation. Minimising education inequalities including by exploring strategies designed to widen access to and boost engagement with health messaging is essential for enhancing compliance with infection, prevention and control measures and ensuring timely presentation.
Patients with obesity, hypertension, IHD, heart failure, CKD, PVD, T2DM, cirrhosis and CKD were at increased risk of mortality. The UK's Chief Medical O cer has highlighted that comorbidities and the proportion of patients with multimorbidity is rising (21) presenting a challenge to the medical profession including within acute and long-term hospital settings (21)(22). This study nds that obesity and multimorbidity are independent mortality risk factors. BAME patients were more likely to present with obesity. CCI scores among every BAME subgroup were higher than Caucasian controls matched by age, sex, Wider BHS deprivation, Outdoor LE deprivation, Indoor LE deprivation and Adult Skills deprivation. Furthermore, the multimorbidity burden among African, Caribbean and Pakistani patients was higher compared with age and sex matched Caucasian controls which may contribute an explanation towards the higher mortality among these subgroups despite low CURB65 scores (0-1).
This study included hospitalised COVID-19 patients within four hospitals across the West Midlands, although these constitute one of the UK's largest NHS Trusts. It did not analyse COVID-19 patients who were not hospitalised or who died in the community. Future studies need to relate these ndings with populations from other urban cities and rural regions with this level of granularity to inform national strategic planning.

Conclusion
Household overcrowding, air pollution, housing quality and adult skills deprivation are potential hidden risk factors for presentation with radiological multilobar pneumonia and ITU admission, which are themselves independent risk factors for mortality. BAME subgroups are more likely to be admitted from the most deprived sub-indices of at least one of these deprivation forms and with higher Charlson Comorbidity (CCI) scores than age, sex and deprivation matched Caucasian controls; multimorbidity is another independent risk factor for mortality. BAME subgroups exhibit younger age structures resulting in potential CURB65 underscoring and disproportionate exposure to unscored risk factors of sex, obesity, multimorbidity and deprivation resulting in potential triage to an inappropriate level of care and clinicians left falsely reassured regarding the severity of presentation and risk of deterioration.
Consideration of multi-ethnic age structures, sex, body mass index, CCI score, chest X-ray imaging and deprivation sub-indices on admission supports clinicians in stratifying high risk patients. COVID-19 clinical risk strati cation tools need to be developed to account for risk factors to which BAME subgroups are predominantly exposed. This will enable the early identi cation of patients at risk of deterioration and ensure triage to an appropriate level of care.  A CONSORT diagram showing participants assessed for eligibility, the inclusion criteria and the nal number of participants included. 3671 consecutive patients were assessed for eligibility for inclusion into this study. 716 patients were excluded on account of having not met the inclusion criteria due to: ongoing hospitalisation on 1st September 2020(n=55), age <18 (n=22), attending hospital as an elective admission(n=371) or attending hospital without admission(n=267). Patients eligible for inclusion in this study(n=2955) were reviewed; patients without listed postcodes(n=301) or postcodes not returning deprivation metrics(n=8) could not be included in the analysed group(n=2646).

Figure 4
Odds ratios of hospitalised COVID-19 patients presenting with multilobar pneumonia, requiring ITU admission and mortality (age and sex adjusted) (a) Odds ratios of presentation with multilobar pneumonia by: gender, ethnicity (BAME, Pakistani, Bangladeshi, Indian, Caribbean, African, Mixed, Chinese, Other ethnic group vs. Caucasian), admission from sub-indices of highest deprivation (Wider BHS deprivation, Indoor LE deprivation, Outdoor LE deprivation, Adult Skills deprivation) vs. admission from all other deprivation sub-indices of the respective deprivation form, admission to ITU vs. not admitted to ITU and mortality (age and sex adjusted) vs. discharge. (b) Odds ratios of ITU admission by: gender, ethnicity (BAME, Pakistani, Bangladeshi, Indian, Caribbean, African, Mixed, Chinese, Other ethnic group vs. Caucasian), admission from sub-indices of highest deprivation (wider BHS deprivation, Indoor LE deprivation, Outdoor LE deprivation, Adult Skills deprivation) vs. admission from all other deprivation sub-indices of the respective deprivation form and presentation with pneumonia (radiological pneumonia vs. radiological multilobar pneumonia) vs. presentation without pneumonia; (c) Odds ratios of age and sex adjusted mortality by: gender, ethnicity (BAME, Pakistani, Bangladeshi, Indian, Caribbean, African, Mixed, Chinese, Other ethnic group vs. Caucasian), admission from sub-indices of highest deprivation (wider BHS deprivation, Indoor LE deprivation, Outdoor LE deprivation, Adult Skills deprivation) vs. admission from all other deprivation sub-indices of the respective deprivation form, presentation with pneumonia (radiological pneumonia, radiological multilobar pneumonia) vs. presentation without pneumonia and ITU admission vs. not admitted to ITU.

Figure 5
Odds ratios of mortality among COVID-19 patients by underlying obesity, hypertension, ischaemic heart disease, heart failure, peripheral vascular disease, COPD, type 2 diabetes mellitus, liver cirrhosis and chronic kidney disease Figure 6