Post by Admin on Oct 12, 2020 6:47:03 GMT
A case-control and cohort study to determine the relationship between ethnic background and severe COVID-19
Rosita Zakeri
Rebecca Bendayan
Mark Ashworth
Daniel M. Bean
Hiten Dodhia
Stevo Durbaba
et al.
Open Access Published:October 09, 2020 DOI:https://doi.org/10.1016/j.eclinm.2020.100574
Abstract
Background
People of minority ethnic backgrounds may be disproportionately affected by severe COVID-19. Whether this relates to increased infection risk, more severe disease progression, or worse in-hospital survival is unknown. The contribution of comorbidities or socioeconomic deprivation to ethnic patterning of outcomes is also unclear.
Methods
We conducted a case-control and a cohort study in an inner city primary and secondary care setting to examine whether ethnic background affects the risk of hospital admission with severe COVID-19 and/or in-hospital mortality. Inner city adult residents admitted to hospital with confirmed COVID-19 (n = 872 cases) were compared with 3,488 matched controls randomly sampled from a primary healthcare database comprising 344,083 people residing in the same region. For the cohort study, we studied 1827 adults consecutively admitted with COVID-19. The primary exposure variable was self-defined ethnicity. Analyses were adjusted for socio-demographic and clinical variables.
Findings
The 872 cases comprised 48.1% Black, 33.7% White, 12.6% Mixed/Other and 5.6% Asian patients. In conditional logistic regression analyses, Black and Mixed/Other ethnicity were associated with higher admission risk than white (OR 3.12 [95% CI 2.63–3.71] and 2.97 [2.30–3.85] respectively). Adjustment for comorbidities and deprivation modestly attenuated the association (OR 2.24 [1.83–2.74] for Black, 2.70 [2.03–3.59] for Mixed/Other). Asian ethnicity was not associated with higher admission risk (adjusted OR 1.01 [0.70–1.46]). In the cohort study of 1827 patients, 455 (28.9%) died over a median (IQR) of 8 (4–16) days. Age and male sex, but not Black (adjusted HR 1.06 [0.82–1.37]) or Mixed/Other ethnicity (adjusted HR 0.72 [0.47–1.10]), were associated with in-hospital mortality. Asian ethnicity was associated with higher in-hospital mortality but with a large confidence interval (adjusted HR 1.71 [1.15–2.56]).
Interpretation
Black and Mixed ethnicity are independently associated with greater admission risk with COVID-19 and may be risk factors for development of severe disease, but do not affect in-hospital mortality risk. Comorbidities and socioeconomic factors only partly account for this and additional ethnicity-related factors may play a large role. The impact of COVID-19 may be different in Asians.
1. Introduction
SARS-CoV2 (Severe acute respiratory syndrome coronavirus 2) is a highly transmissible respiratory pathogen that usually causes minor illness but in a small proportion of individuals leads to severe systemic disease (Coronavirus disease 2019 [COVID-19]), with 20–30% in-hospital mortality [1, 2, 3, 4]. Older people, males, and those with comorbidities such as diabetes and cardiovascular disorders are over-represented among those requiring hospital admission [1, 2, 3, 4]. After COVID-19 spread to multi-ethnic populations in Western Europe and North America, numerous reports suggested a higher disease burden in Black, Asian or other minority ethnic groups [5, 6, 7, 8, 9, 10]. Audit data on patients admitted to UK intensive care units (ICU) with COVID-19 observed a substantially higher proportion of minority ethnic background patients than previous years [7]. US data reported higher per-capita mortality rates for Black and Hispanic compared to White people in several cities or in aggregate analyses across large states, but the underlying reasons are unclear [6,9]. A UK Office of National Statistics (ONS) analysis suggested that individuals of Black and South Asian descent had a higher likelihood of death than White people after adjustment for demographic and socioeconomic factors, but was limited by lack of information on comorbidities and the use of historic (2011) data for the reference population [8]. A recent UK nationwide cohort study [11] also reports increased overall mortality in Black and South Asian compared to White people but, importantly, does not take into account the large variations in the ethnic composition of local populations within different geographical regions.
Higher mortality in minority ethnic groups could simply be because minority populations in most Western countries are typically more concentrated in large cities, and these are the very regions that have been most affected by COVID-19. In the UK's largest city, London, 27 of the 33 Boroughs have an ethnic minority population of at least 25% [12]. Furthermore, nine London Boroughs with a high ethnic minority population are among the ten local authorities with the highest age-standardised COVID-19 mortality rates in the UK [12,13]. Many of these regions are also characterised by a higher community prevalence of comorbidities and greater socioeconomic deprivation. Detailed data on socio-demography and comorbidities at local community level are therefore essential to dissect the complex relationship between ethnicity and COVID-19, ideally in a case-control study design to reduce selection bias. This has not been attempted in the studies reported to date nor have they defined when in the disease trajectory ethnicity-related differences are manifest; i.e. infection, disease progression leading to hospitalisation, or in-hospital survival.
We addressed these questions in a London region with approximately 40% Black and minority ethnic background people among a population of 1.26 million. Our aims were to (1) determine the relationship between ethnicity, local population demography, individual-level comorbidities, socioeconomic profiles, and hospital admission for severe COVID-19; (2) establish whether ethnicity is associated with in-hospital outcome of severe COVID-19.
2. Methods
2.1 Study design and participants
We conducted an observational cohort study at King's College Hospital Foundation Trust (KCHFT), which comprises two separate hospitals in south London. We included consecutive adult patients (age ≥18 years) requiring emergency hospital admission with a primary diagnosis of COVID-19, between 1 March and 2 June 2020. All patients tested positive for viral RNA by quantitative RT-PCR in nasopharyngeal and oropharyngeal swabs.
We performed a case-control (case-population [14]) study with the subset of admitted patients who were inner city residents (>55% of the total hospital cohort). COVID-19 patients (i.e. cases) were matched with population controls sampled from the same inner city region using a primary healthcare database (Lambeth DataNet). This comprises de-identified data on 344,083 (96.8%) community-resident adults registered with 41 practices in inner south-east London [15]. We randomly sampled four controls for each case, individually matched by age (within 5-year age bands) and sex.
2.2 Data sources and processing
Demographic and clinical data for admitted patients were retrieved from the electronic health record (EHR). We used well-validated natural language processing (NLP) informatics tools belonging to the CogStack ecosystem to access both structured fields and unstructured text in the EHR [16, 17, 18]. Additional manual extraction, clinician case review, and mandatory hospital datasets were used for missing variables and validation. For population controls, individual-level anonymised Read-coded data were extracted from the structured fields of the primary care EHR database.
2.3 Exposures
The primary exposure variable was self-reported ethnicity according to the 18 categories recommended by the ONS [19]. These were reduced into four groups: White (British, Irish, Gypsy, any other White), Black (African, Caribbean, any other Black), Asian (Indian, Pakistani, Bangladeshi, Chinese, any other Asian), and Mixed/Other. Patients with missing ethnicity data were excluded. Demographic and clinical variables, identified a priori as potential risk factors for severe COVID-19, included: age, sex, body mass index (BMI), cardiometabolic comorbidities (hypertension, coronary heart disease [CHD], heart failure, previous stroke or transient ischaemic attack [TIA], diabetes, chronic kidney disease [CKD]), asthma and chronic obstructive pulmonary disease (COPD). For BMI, the most recent value within 6 months (median 27 days [LQ-UQ 0–38]) of admission (for hospital cases) or data extraction (primary care) was used. BMI categories were defined as underweight (<18.5 kg/m2), normal (18.5–24.9 kg/m2 [18.5–22.9 kg/m2 for Asians]), overweight (25–29.9 kg/m2 [23–27.4 kg/m2 for Asians]), and obese (≥30 kg/m2 [≥27.5 kg/m2 for Asians]) [20]. Comorbidities were categorised as present if recorded at any time in the EHR up to and including the day of admission (or data extraction in primary care). Socioeconomic status was estimated using the English Indices of Multiple Deprivation (IMD) score, presented in quintiles, as derived from each individual's residential postcode and the relevant 2019 Low Super Output Area (LSOA) code [21]. The IMD score includes data on income, employment, crime, living environment, education and barriers to services. Higher quintiles indicate less deprivation. Disease severity at admission was estimated from the routinely recorded NEWS2 score (National Early Warning Score for degree of illness). Date of self-reported symptom onset was extracted where available. The same variable definitions were used across primary and secondary care.
2.4 Outcomes
Outcomes included hospital admission for COVID-19 and in-hospital mortality. The secondary outcome of ICU admission was evaluated in the hospital cohort. Start of follow-up for all analyses was taken as the admission date. Outcomes were ascertained to 2 June 2020.
2.5 Statistical analysis
Patient characteristics were summarised using frequency (%) and median with interquartile range (IQR). Comparisons across ethnic groups were made using the χ2 or Fisher's exact test for categorical variables and 1-way ANOVA or Kruskal-Wallis test as appropriate for continuous variables. Where individual comparisons were made, e.g. between Black and White ethnicity, the Bonferroni correction was used. To evaluate the association between ethnicity and hospital admission for COVID-19, we fitted conditional logistic regression models. All models were adjusted for the matching variables (age and sex) to eliminate residual confounding [22] and based on their previously reported associations with COVID-19. Successive adjusted models included adjustment for IMD, cardiometabolic comorbidities, all comorbidities (cardiometabolic comorbidities plus asthma and chronic obstructive pulmonary disease), and finally a combination of all variables (fully adjusted model). Variables were selected based on previous literature and clinical relevance. Age was modelled as a categorical variable to allow for potentially non-linear association with outcomes (18–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84, 85+ years). Comorbidities were modelled as binary variables and IMD was fitted as a categorical variable (quintiles). White ethnicity was used as the reference group.
In the hospital cohort, we evaluated the association between ethnicity and risk of in-hospital death using Cox proportional hazards models, with in-hospital mortality as the dependent variable and admission date as the start of the observation window. For the secondary outcome of ICU admission, we used competing risks regression, based on Fine and Gray's proportional sub-hazards model [23], with ICU admission as the dependent variable and death or discharge from hospital assigned as a competing risk. Univariable, age- and sex-adjusted, and fully adjusted multivariable models were performed as for the case-control study with White ethnicity as the reference group. Patients were considered to be right-censored if they were a) discharged from hospital alive, or b) currently in hospital at the study end date. The proportional hazard assumption was examined graphically and using formal tests, using the methods described by Grambsch [24]. No major deviations from this assumption were observed. Analyses were performed using STATA/IC (v16.1; StataCorp LLC, TX). As this study was descriptive, formal power calculations were not performed; however, sample size considerations are highlighted in the Supplemental Methods.
2.6 Sensitivity analyses
Since BMI was missing in >30% of hospitalised patients, the primary analyses were performed without adjustment for BMI. To explore confounding by BMI, all analyses were repeated in the subset of patients with BMI data available on a complete case analysis basis, with BMI modelled as a continuous variable. To investigate potential confounding due to differences in timing of hospital presentation between ethnic groups we performed sensitivity analyses using the date of self-reported symptom-onset as the beginning of the observation window, in lieu of hospital admission date, in the subset of patients where this was reported. Additional sensitivity analyses were performed in patients ≥65 years of age, and with imputation of missing comorbidity and IMD variables (<5%) using multiple imputations by chain equations [25].
Rosita Zakeri
Rebecca Bendayan
Mark Ashworth
Daniel M. Bean
Hiten Dodhia
Stevo Durbaba
et al.
Open Access Published:October 09, 2020 DOI:https://doi.org/10.1016/j.eclinm.2020.100574
Abstract
Background
People of minority ethnic backgrounds may be disproportionately affected by severe COVID-19. Whether this relates to increased infection risk, more severe disease progression, or worse in-hospital survival is unknown. The contribution of comorbidities or socioeconomic deprivation to ethnic patterning of outcomes is also unclear.
Methods
We conducted a case-control and a cohort study in an inner city primary and secondary care setting to examine whether ethnic background affects the risk of hospital admission with severe COVID-19 and/or in-hospital mortality. Inner city adult residents admitted to hospital with confirmed COVID-19 (n = 872 cases) were compared with 3,488 matched controls randomly sampled from a primary healthcare database comprising 344,083 people residing in the same region. For the cohort study, we studied 1827 adults consecutively admitted with COVID-19. The primary exposure variable was self-defined ethnicity. Analyses were adjusted for socio-demographic and clinical variables.
Findings
The 872 cases comprised 48.1% Black, 33.7% White, 12.6% Mixed/Other and 5.6% Asian patients. In conditional logistic regression analyses, Black and Mixed/Other ethnicity were associated with higher admission risk than white (OR 3.12 [95% CI 2.63–3.71] and 2.97 [2.30–3.85] respectively). Adjustment for comorbidities and deprivation modestly attenuated the association (OR 2.24 [1.83–2.74] for Black, 2.70 [2.03–3.59] for Mixed/Other). Asian ethnicity was not associated with higher admission risk (adjusted OR 1.01 [0.70–1.46]). In the cohort study of 1827 patients, 455 (28.9%) died over a median (IQR) of 8 (4–16) days. Age and male sex, but not Black (adjusted HR 1.06 [0.82–1.37]) or Mixed/Other ethnicity (adjusted HR 0.72 [0.47–1.10]), were associated with in-hospital mortality. Asian ethnicity was associated with higher in-hospital mortality but with a large confidence interval (adjusted HR 1.71 [1.15–2.56]).
Interpretation
Black and Mixed ethnicity are independently associated with greater admission risk with COVID-19 and may be risk factors for development of severe disease, but do not affect in-hospital mortality risk. Comorbidities and socioeconomic factors only partly account for this and additional ethnicity-related factors may play a large role. The impact of COVID-19 may be different in Asians.
1. Introduction
SARS-CoV2 (Severe acute respiratory syndrome coronavirus 2) is a highly transmissible respiratory pathogen that usually causes minor illness but in a small proportion of individuals leads to severe systemic disease (Coronavirus disease 2019 [COVID-19]), with 20–30% in-hospital mortality [1, 2, 3, 4]. Older people, males, and those with comorbidities such as diabetes and cardiovascular disorders are over-represented among those requiring hospital admission [1, 2, 3, 4]. After COVID-19 spread to multi-ethnic populations in Western Europe and North America, numerous reports suggested a higher disease burden in Black, Asian or other minority ethnic groups [5, 6, 7, 8, 9, 10]. Audit data on patients admitted to UK intensive care units (ICU) with COVID-19 observed a substantially higher proportion of minority ethnic background patients than previous years [7]. US data reported higher per-capita mortality rates for Black and Hispanic compared to White people in several cities or in aggregate analyses across large states, but the underlying reasons are unclear [6,9]. A UK Office of National Statistics (ONS) analysis suggested that individuals of Black and South Asian descent had a higher likelihood of death than White people after adjustment for demographic and socioeconomic factors, but was limited by lack of information on comorbidities and the use of historic (2011) data for the reference population [8]. A recent UK nationwide cohort study [11] also reports increased overall mortality in Black and South Asian compared to White people but, importantly, does not take into account the large variations in the ethnic composition of local populations within different geographical regions.
Higher mortality in minority ethnic groups could simply be because minority populations in most Western countries are typically more concentrated in large cities, and these are the very regions that have been most affected by COVID-19. In the UK's largest city, London, 27 of the 33 Boroughs have an ethnic minority population of at least 25% [12]. Furthermore, nine London Boroughs with a high ethnic minority population are among the ten local authorities with the highest age-standardised COVID-19 mortality rates in the UK [12,13]. Many of these regions are also characterised by a higher community prevalence of comorbidities and greater socioeconomic deprivation. Detailed data on socio-demography and comorbidities at local community level are therefore essential to dissect the complex relationship between ethnicity and COVID-19, ideally in a case-control study design to reduce selection bias. This has not been attempted in the studies reported to date nor have they defined when in the disease trajectory ethnicity-related differences are manifest; i.e. infection, disease progression leading to hospitalisation, or in-hospital survival.
We addressed these questions in a London region with approximately 40% Black and minority ethnic background people among a population of 1.26 million. Our aims were to (1) determine the relationship between ethnicity, local population demography, individual-level comorbidities, socioeconomic profiles, and hospital admission for severe COVID-19; (2) establish whether ethnicity is associated with in-hospital outcome of severe COVID-19.
2. Methods
2.1 Study design and participants
We conducted an observational cohort study at King's College Hospital Foundation Trust (KCHFT), which comprises two separate hospitals in south London. We included consecutive adult patients (age ≥18 years) requiring emergency hospital admission with a primary diagnosis of COVID-19, between 1 March and 2 June 2020. All patients tested positive for viral RNA by quantitative RT-PCR in nasopharyngeal and oropharyngeal swabs.
We performed a case-control (case-population [14]) study with the subset of admitted patients who were inner city residents (>55% of the total hospital cohort). COVID-19 patients (i.e. cases) were matched with population controls sampled from the same inner city region using a primary healthcare database (Lambeth DataNet). This comprises de-identified data on 344,083 (96.8%) community-resident adults registered with 41 practices in inner south-east London [15]. We randomly sampled four controls for each case, individually matched by age (within 5-year age bands) and sex.
2.2 Data sources and processing
Demographic and clinical data for admitted patients were retrieved from the electronic health record (EHR). We used well-validated natural language processing (NLP) informatics tools belonging to the CogStack ecosystem to access both structured fields and unstructured text in the EHR [16, 17, 18]. Additional manual extraction, clinician case review, and mandatory hospital datasets were used for missing variables and validation. For population controls, individual-level anonymised Read-coded data were extracted from the structured fields of the primary care EHR database.
2.3 Exposures
The primary exposure variable was self-reported ethnicity according to the 18 categories recommended by the ONS [19]. These were reduced into four groups: White (British, Irish, Gypsy, any other White), Black (African, Caribbean, any other Black), Asian (Indian, Pakistani, Bangladeshi, Chinese, any other Asian), and Mixed/Other. Patients with missing ethnicity data were excluded. Demographic and clinical variables, identified a priori as potential risk factors for severe COVID-19, included: age, sex, body mass index (BMI), cardiometabolic comorbidities (hypertension, coronary heart disease [CHD], heart failure, previous stroke or transient ischaemic attack [TIA], diabetes, chronic kidney disease [CKD]), asthma and chronic obstructive pulmonary disease (COPD). For BMI, the most recent value within 6 months (median 27 days [LQ-UQ 0–38]) of admission (for hospital cases) or data extraction (primary care) was used. BMI categories were defined as underweight (<18.5 kg/m2), normal (18.5–24.9 kg/m2 [18.5–22.9 kg/m2 for Asians]), overweight (25–29.9 kg/m2 [23–27.4 kg/m2 for Asians]), and obese (≥30 kg/m2 [≥27.5 kg/m2 for Asians]) [20]. Comorbidities were categorised as present if recorded at any time in the EHR up to and including the day of admission (or data extraction in primary care). Socioeconomic status was estimated using the English Indices of Multiple Deprivation (IMD) score, presented in quintiles, as derived from each individual's residential postcode and the relevant 2019 Low Super Output Area (LSOA) code [21]. The IMD score includes data on income, employment, crime, living environment, education and barriers to services. Higher quintiles indicate less deprivation. Disease severity at admission was estimated from the routinely recorded NEWS2 score (National Early Warning Score for degree of illness). Date of self-reported symptom onset was extracted where available. The same variable definitions were used across primary and secondary care.
2.4 Outcomes
Outcomes included hospital admission for COVID-19 and in-hospital mortality. The secondary outcome of ICU admission was evaluated in the hospital cohort. Start of follow-up for all analyses was taken as the admission date. Outcomes were ascertained to 2 June 2020.
2.5 Statistical analysis
Patient characteristics were summarised using frequency (%) and median with interquartile range (IQR). Comparisons across ethnic groups were made using the χ2 or Fisher's exact test for categorical variables and 1-way ANOVA or Kruskal-Wallis test as appropriate for continuous variables. Where individual comparisons were made, e.g. between Black and White ethnicity, the Bonferroni correction was used. To evaluate the association between ethnicity and hospital admission for COVID-19, we fitted conditional logistic regression models. All models were adjusted for the matching variables (age and sex) to eliminate residual confounding [22] and based on their previously reported associations with COVID-19. Successive adjusted models included adjustment for IMD, cardiometabolic comorbidities, all comorbidities (cardiometabolic comorbidities plus asthma and chronic obstructive pulmonary disease), and finally a combination of all variables (fully adjusted model). Variables were selected based on previous literature and clinical relevance. Age was modelled as a categorical variable to allow for potentially non-linear association with outcomes (18–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84, 85+ years). Comorbidities were modelled as binary variables and IMD was fitted as a categorical variable (quintiles). White ethnicity was used as the reference group.
In the hospital cohort, we evaluated the association between ethnicity and risk of in-hospital death using Cox proportional hazards models, with in-hospital mortality as the dependent variable and admission date as the start of the observation window. For the secondary outcome of ICU admission, we used competing risks regression, based on Fine and Gray's proportional sub-hazards model [23], with ICU admission as the dependent variable and death or discharge from hospital assigned as a competing risk. Univariable, age- and sex-adjusted, and fully adjusted multivariable models were performed as for the case-control study with White ethnicity as the reference group. Patients were considered to be right-censored if they were a) discharged from hospital alive, or b) currently in hospital at the study end date. The proportional hazard assumption was examined graphically and using formal tests, using the methods described by Grambsch [24]. No major deviations from this assumption were observed. Analyses were performed using STATA/IC (v16.1; StataCorp LLC, TX). As this study was descriptive, formal power calculations were not performed; however, sample size considerations are highlighted in the Supplemental Methods.
2.6 Sensitivity analyses
Since BMI was missing in >30% of hospitalised patients, the primary analyses were performed without adjustment for BMI. To explore confounding by BMI, all analyses were repeated in the subset of patients with BMI data available on a complete case analysis basis, with BMI modelled as a continuous variable. To investigate potential confounding due to differences in timing of hospital presentation between ethnic groups we performed sensitivity analyses using the date of self-reported symptom-onset as the beginning of the observation window, in lieu of hospital admission date, in the subset of patients where this was reported. Additional sensitivity analyses were performed in patients ≥65 years of age, and with imputation of missing comorbidity and IMD variables (<5%) using multiple imputations by chain equations [25].