Predicting COVID-19 related death using the OpenSAFELY platform (preprint)

Objectives To compare approaches for obtaining relative and absolute estimates of risk of 28-day COVID-19 mortality for adults in the general population of England in the context of changing levels of circulating infection. Design Three designs were compared. (A) case-cohort which does not explicitly account for the time-changing prevalence of COVID-19 infection, (B) 28-day landmarking, a series of sequential overlapping sub-studies incorporating time-updating proxy measures of the prevalence of infection, and (C) daily landmarking. Regression models were fitted to predict 28-day COVID-19 mortality. Setting Working on behalf of NHS England, we used clinical data from adult patients from all regions of England held in the TPP SystmOne electronic health record system, linked to Office for National Statistics (ONS) mortality data, using the OpenSAFELY platform. Participants Eligible participants were adults aged 18 or over, registered at a general practice using TPP software on 1st March 2020 with recorded sex, postcode and ethnicity. 11,972,947 individuals were included, and 7,999 participants experienced a COVID-19 related death. The study period lasted 100 days, ending 8th June 2020. Predictors A range of demographic characteristics and comorbidities were used as potential predictors. Local infection prevalence was estimated with three proxies: modelled based on local prevalence and other key factors; rate of A&E COVID-19 related attendances; and rate of suspected COVID-19 cases in primary care. Main outcome measures COVID-19 related death. Results All models discriminated well between patients who did and did not experience COVID-19 related death, with C-statistics ranging from 0.92-0.94. Accurate estimates of absolute risk required data on local infection prevalence, with modelled estimates providing the best performance. Conclusions Reliable estimates of absolute risk need to incorporate changing local prevalence of infection. Simple models can provide very good discrimination and may simplify implementation of risk prediction tools in practice.

Short report: Ethnicity and COVID-19 death in the early part of the COVID-19 second wave in England: an analysis of OpenSAFELY data from 1st September to 9th November 2020 (preprint)

Black and minority ethnic groups were at raised risk of dying from COVID-19 during the first few months of the COVID-19 epidemic in England. We aimed to investigate whether ethnic inequalities in COVID-19 deaths were similar in the more recent "second wave" of the epidemic. Working on behalf of NHS England, we used primary care and linked ONS mortality data within the OpenSAFELY platform. All adults in the database at 1st September 2020 and with at least 1 year of prior follow-up and a record of ethnicity were included. The outcome was COVID-19-related death (death with COVID-19 listed as a cause of death on the death certificate). Follow-up was to 9th November 2020. Hazard ratios for ethnicity were calculated using Cox regression models adjusted for age and sex, and then further adjusted for deprivation. 13,223,154 people were included. During the study period, people of South Asian ethnicity were at higher risk of death due to COVID-19 than white people after adjusting for age and sex (HR = 3.47, 95% CI 2.99-4.03); the association attenuated somewhat on further adjustment for index of multiple deprivation (HR = 2.86, 2.46-3.33, Table 2). In contrast with the first wave of the epidemic, we found little evidence of a raised risk in black or other ethnic groups compared to white (HR for black vs white = 1.28, 0.87-1.88 adjusted for age and sex; and 1.01, 0.69-1.49 further adjusted for deprivation). Our findings suggest that ethnic inequalities in the risk of dying COVID-19-related death have changed between the first and early second wave of the epidemic in England. O_TBL View this table: org.highwire.dtl.DTLVardef@987a5org.highwire.dtl.DTLVardef@1a8a141org.highwire.dtl.DTLVardef@1f2de56org.highwire.dtl.DTLVardef@1e2f9b8org.highwire.dtl.DTLVardef@78bfcc_HPS_FORMAT_FIGEXP M_TBL O_FLOATNOTable 2:C_FLOATNO O_TABLECAPTIONAssociation between ethnicity and COVID-19 death 1st Sept - 9th Nov 2020 C_TABLECAPTION C_TBL

Ethnic differences in COVID-19 infection, hospitalisation, and mortality: an OpenSAFELY analysis of 17 million adults in England (preprint)

Background: COVID-19 has had a disproportionate impact on ethnic minority populations, both in the UK and internationally. To date, much of the evidence has been derived from studies within single healthcare settings, mainly those hospitalised with COVID-19. Working on behalf of NHS England, the aim of this study was to identify ethnic differences in the risk of COVID-19 infection, hospitalisation and mortality using a large general population cohort in England. Methods: We conducted an observational cohort study using linked primary care records of 17.5 million adults between 1 February 2020 and 3 August 2020. Exposure was self-reported ethnicity collapsed into the 5 and 16 ethnicity categories of the English Census. Multivariable Cox proportional hazards regression was used to identify ethnic differences in the risk of being tested and testing positive for SARS-CoV-2 infection, COVID-19 related intensive care unit (ICU) admission, and COVID-19 mortality, adjusted for socio-demographic factors, clinical co-morbidities, geographic region, care home residency, and household size. Results: A total of 17,510,002 adults were included in the study; 63% white (n=11,030,673), 6% south Asian (n=1,034,337), 2% black (n=344,889), 2% other (n=324,730), 1% mixed (n=172,551), and 26% unknown (n=4,602,822). After adjusting for measured explanatory factors, south Asian, black, and mixed groups were marginally more likely to be tested (south Asian HR 1.08, 95%CI 1.07-1.09; black HR 1.08; 95%CI 1.06-1.09, mixed HR 1.03, 95%CI 1.01-1.05), and substantially more likely to test positive for SARS-CoV-2 compared with white adults (south Asian HR 2.02. 95% CI 1.97-2.07; black HR 1.68, 95%CI 1.61-1.76; mixed HR 1.46, 95%CI 1.36-1.56). The risk of being admitted to ICU for COVID-19 was substantially increased in all ethnic minority groups compared with white adults (south Asian HR 2.22, 95%CI 1.96-2.52; black HR 3.07, 95%CI 2.61-3.61; mixed HR 2.86, 95%CI 2.19-3.75, other HR 2.86, 95%CI 2.31-3.63). Risk of COVID-19 mortality was increased by 25-56% in ethnic minority groups compared with white adults (south Asian HR 1.27, 95%CI 1.17-1.38; black HR 1.55, 95%CI 1.38-1.75; mixed HR 1.40, 95%CI 1.12-1.76; other HR 1.25, 95%CI 1.05-1.49). We observed heterogeneity of associations after disaggregation into detailed ethnic groupings; Indian and African groups were at higher risk of all outcomes; Pakistani, Bangladeshi and Caribbean groups were less or equally likely to be tested for SARS-CoV-2, but at higher risk of all other outcomes, Chinese groups were less likely to be tested for and test positive for SARS-CoV-2, more likely to be admitted to ICU, and equally likely to die from COVID-19. Conclusions: We found evidence of substantial ethnic inequalities in the risk of testing positive for SARS-CoV-2, ICU admission, and mortality, which persisted after accounting for explanatory factors, including household size. It is likely that some of this excess risk is related to factors not captured in clinical records such as occupation, experiences of structural discrimination, or inequitable access to health and social services. Prioritizing linkage between health, social care, and employment data and engaging with ethnic minority communities to better understand their lived experiences is essential for generating evidence to prevent further widening of inequalities in a timely and actionable manner.

OpenSAFELY: factors associated with COVID-19-related hospital death in the linked electronic health records of 17 million adult NHS patients. (preprint)

Background Establishing who is at risk from a novel rapidly arising cause of death, and why, requires a new approach to epidemiological research with very large datasets and timely data. Working on behalf of NHS England we therefore set out to deliver a secure and pseudonymised analytics platform inside the data centre of a major primary care electronic health records vendor establishing coverage across detailed primary care records for a substantial proportion of all patients in England. The following results are preliminary. Data sources Primary care electronic health records managed by the electronic health record vendor TPP, pseudonymously linked to patient-level data from the COVID-19 Patient Notification System (CPNS) for death of hospital inpatients with confirmed COVID-19, using the new OpenSAFELY platform. Population 17,425,445 adults. Time period 1st Feb 2020 to 25th April 2020. Primary outcome Death in hospital among people with confirmed COVID-19. Methods Cohort study analysed by Cox-regression to generate hazard ratios: age and sex adjusted, and multiply adjusted for co-variates selected prospectively on the basis of clinical interest and prior findings. Results There were 5683 deaths attributed to COVID-19. In summary after full adjustment, death from COVID-19 was strongly associated with: being male (hazard ratio 1.99, 95%CI 1.88-2.10); older age and deprivation (both with a strong gradient); uncontrolled diabetes (HR 2.36 95% CI 2.18-2.56); severe asthma (HR 1.25 CI 1.08-1.44); and various other prior medical conditions. Compared to people with ethnicity recorded as white, black people were at higher risk of death, with only partial attenuation in hazard ratios from the fully adjusted model (age-sex adjusted HR 2.17 95% CI 1.84-2.57; fully adjusted HR 1.71 95% CI 1.44-2.02); with similar findings for Asian people (age-sex adjusted HR 1.95 95% CI 1.73-2.18; fully adjusted HR 1.62 95% CI 1.43-1.82). Conclusions We have quantified a range of clinical risk factors for death from COVID-19, some of which were not previously well characterised, in the largest cohort study conducted by any country to date. People from Asian and black groups are at markedly increased risk of in-hospital death from COVID-19, and contrary to some prior speculation this is only partially attributable to pre-existing clinical risk factors or deprivation; further research into the drivers of this association is therefore urgently required. Deprivation is also a major risk factor with, again, little of the excess risk explained by co-morbidity or other risk factors. The findings for clinical risk factors are concordant with policies in the UK for protecting those at highest risk. Our OpenSAFELY platform is rapidly adding further NHS patients' records; we will update and extend these results regularly. Keywords COVID-19, risk factors, ethnicity, deprivation, death, informatics.