Protection against SARS-CoV-2 Omicron BA.4/5 variant following booster vaccination or breakthrough infection in the UK.

Following primary SARS-CoV-2 vaccination, whether boosters or breakthrough infections provide greater protection against SARS-CoV-2 infection is incompletely understood. Here we investigated SARS-CoV-2 antibody correlates of protection against new Omicron BA.4/5 (re-)infections and anti-spike IgG antibody trajectories after a third/booster vaccination or breakthrough infection following second vaccination in 154,149 adults ≥18 y from the United Kingdom general population. Higher antibody levels were associated with increased protection against Omicron BA.4/5 infection and breakthrough infections were associated with higher levels of protection at any given antibody level than boosters. Breakthrough infections generated similar antibody levels to boosters, and the subsequent antibody declines were slightly slower than after boosters. Together our findings show breakthrough infection provides longer-lasting protection against further infections than booster vaccinations. Our findings, considered alongside the risks of severe infection and long-term consequences of infection, have important implications for vaccine policy.

An adversarial training framework for mitigating algorithmic biases in clinical machine learning.

Machine learning is becoming increasingly prominent in healthcare. Although its benefits are clear, growing attention is being given to how these tools may exacerbate existing biases and disparities. In this study, we introduce an adversarial training framework that is capable of mitigating biases that may have been acquired through data collection. We demonstrate this proposed framework on the real-world task of rapidly predicting COVID-19, and focus on mitigating site-specific (hospital) and demographic (ethnicity) biases. Using the statistical definition of equalized odds, we show that adversarial training improves outcome fairness, while still achieving clinically-effective screening performances (negative predictive values >0.98). We compare our method to previous benchmarks, and perform prospective and external validation across four independent hospital cohorts. Our method can be generalized to any outcomes, models, and definitions of fairness.

Performance of antigen lateral flow devices in the UK during the alpha, delta, and omicron waves of the SARS-CoV-2 pandemic: a diagnostic and observational study.

BACKGROUND: Antigen lateral flow devices (LFDs) have been widely used to control SARS-CoV-2. We aimed to improve understanding of LFD performance with changes in variant infections, vaccination, viral load, and LFD use, and in the detection of infectious individuals. METHODS: In this diagnostic study, paired LFD and RT-PCR test results were prospectively collected from asymptomatic and symptomatic participants in the UK between Nov 4, 2020, and March 21, 2022, to support the National Health Service (NHS) England's Test and Trace programme. The LFDs evaluated were the Innova SARS-CoV-2 Antigen Rapid Qualitative Test, the Orient Gene Rapid Covid-19 (Antigen) Self-Test, and the Acon Flowflex SARS-CoV-2 Antigen Rapid Test (Self-Testing). Test results were collected across various community testing settings, including predeployment testing sites, routine testing centres, homes, schools, universities, workplaces, targeted community testing, and from health-care workers. We used multivariable logistic regression to analyse LFD sensitivity and specificity using RT-PCR as a reference standard, adjusting for viral load, LFD manufacturer, test setting, age, sex, test assistance, symptom status, vaccination status, and SARS-CoV-2 variant. National contact tracing data from NHS Test and Trace (Jan 1, 2021, to Jan 11, 2022) were used to estimate the proportion of transmitting index patients (with ≥1 RT-PCR-positive or LFD-positive contact) potentially detectable by LFDs (specifically Innova, as the most widely used LFD) with time, accounting for index viral load, variant, and symptom status. FINDINGS: We assessed 75 382 pairs of LFD and RT-PCR tests. Of these, 4131 (5·5%) were RT-PCR-positive. LFD sensitivity versus RT-PCR was 63·2% (95% CI 61·7-64·6) and specificity was 99·71% (95% CI 99·66-99·74). Increased viral load was independently associated with being LFD positive (adjusted odds ratio [aOR] 2·85 [95% CI 2·66-3·06] per 1 log10 copies per mL increase; p<0·0001). There was no evidence that LFD sensitivity differed for delta (B.1.617.2) infections versus alpha (B.1.1.7) or pre-alpha (B.1.177) infections (aOR 1·00 [0·69-1·45]; p=0·99), whereas omicron (BA.1 or BA.2) infections appeared more likely to be LFD positive (aOR 1·63 [1·02-2·59]; p=0·042). Sensitivity was higher in symptomatic participants (68·7% [95% CI 66·9-70·4]) than in asymptomatic participants (52·8% [50·1-55·4]). Among 347 374 unique index patients with probable onward transmission, 78·3% (95% CI 75·3-81·2) were estimated to have been detectable with LFDs (Innova), and this proportion was mostly stable with time and for successive variants. Overall, the estimated proportion of infectious index patients detectable by the Innova LFD was lower in asymptomatic patients (57·6% [53·6-61·9]) versus symptomatic patients (79·7% [76·7-82·5]). INTERPRETATION: LFDs remained able to detect most SARS-CoV-2 infections throughout vaccine roll-out and across different viral variants. LFDs can potentially detect most infections that transmit to others and reduce the risk of transmission. However, performance is lower in asymptomatic individuals than in symptomatic individuals. FUNDING: UK Health Security Agency, the UK Government Department of Health and Social Care, National Institute for Health Research (NIHR) Health Protection Research Unit in Healthcare Associated Infections and Antimicrobial Resistance, and the University of Oxford NIHR Biomedical Research Centre.

Self-Aware SGD: Reliable Incremental Adaptation Framework For Clinical AI Models.

Healthcare is dynamic as demographics, diseases, and therapeutics constantly evolve. This dynamic nature induces inevitable distribution shifts in populations targeted by clinical AI models, often rendering them ineffective. Incremental learning provides an effective method of adapting deployed clinical models to accommodate these contemporary distribution shifts. However, since incremental learning involves modifying a deployed or in-use model, it can be considered unreliable as any adverse modification due to maliciously compromised or incorrectly labelled data can make the model unsuitable for the targeted application. This paper introduces self-aware stochastic gradient descent (SGD), an incremental deep learning algorithm that utilises a contextual bandit-like sanity check to only allow reliable modifications to a model. The contextual bandit analyses incremental gradient updates to isolate and filter unreliable gradients. This behaviour allows self-aware SGD to balance incremental training and integrity of a deployed model. Experimental evaluations on the Oxford University Hospital datasets highlight that self-aware SGD can provide reliable incremental updates for overcoming distribution shifts in challenging conditions induced by label noise.

Omicron-associated changes in SARS-CoV-2 symptoms in the United Kingdom.

BACKGROUND: The SARS-CoV-2 Delta variant has been replaced by the highly transmissible Omicron BA.1 variant, and subsequently by Omicron BA.2. It is important to understand how these changes in dominant variants affect reported symptoms, while also accounting for symptoms arising from other co-circulating respiratory viruses. METHODS: In a nationally representative UK community study, the COVID-19 Infection Survey, we investigated symptoms in PCR-positive infection episodes vs. PCR-negative study visits over calendar time, by age and vaccination status, comparing periods when the Delta, Omicron BA.1 and BA.2 variants were dominant. RESULTS: Between October-2020 and April-2022, 120,995 SARS-CoV-2 PCR-positive episodes occurred in 115,886 participants, with 70,683 (58%) reporting symptoms. The comparator comprised 4,766,366 PCR-negative study visits (483,894 participants); 203,422 (4%) reporting symptoms. Symptom reporting in PCR-positives varied over time, with a marked reduction in loss of taste/smell as Omicron BA.1 dominated, maintained with BA.2 (44%/45% 17 October 2021, 16%/13% 2 January 2022, 15%/12% 27 March 2022). Cough, fever, shortness of breath, myalgia, fatigue/weakness and headache also decreased after Omicron BA.1 dominated, but sore throat increased, the latter to a greater degree than concurrent increases in PCR-negatives. Fatigue/weakness increased again after BA.2 dominated, although to a similar degree to concurrent increases in PCR-negatives. Symptoms were consistently more common in adults aged 18-65 years than in children or older adults. CONCLUSIONS: Increases in sore throat (also common in the general community), and a marked reduction in loss of taste/smell, make Omicron harder to detect with symptom-based testing algorithms, with implications for institutional and national testing policies.

CO-CONNECT: A hybrid architecture to facilitate rapid discovery and access to UK wide data in the response to the COVID-19 pandemic.

BACKGROUND: COVID-19 data have been generated across the UK as a by-product of clinical care and public health provision, and numerous bespoke and repurposed research endeavours. Analysis of these data has underpinned the UK's response to the pandemic and informed public health policies and clinical guidelines. However, these data are held by different organisations and this fragmented landscape has presented challenges for public health agencies and researchers as they struggle to find, navigate permissions to access and interrogate the data they need to inform the pandemic response at pace. OBJECTIVE: To transform UK COVID-19 diagnostic datasets to be Findable, Accessible, Interoperable and Reusable (FAIR). METHODS: A federated infrastructure model was rapidly built to enable the automated and reproducible mapping of health Data Partners' pseudonymised data to the OMOP common data model without the need for any data to leave the data controllers' secure environments and to support federated cohort discovery queries and meta-analysis. RESULTS: 56 datasets from 19 organisations are being connected to the federated network. The data includes research cohorts and COVID-19 data collected through routine health care provision linked to longitudinal healthcare records and demographics. The infrastructure is live, supporting aggregate level querying of data across the UK. CONCLUSIONS: CO-CONNECT was developed by a multidisciplinary team enabling rapid COVID-19 data discovery, instantaneous meta-analysis across data sources, and is researching streamlined data extraction for egress into a Trusted Research Environment (TRE) for research and public health analysis. CO-CONNECT has the potential to make UK health data more interconnected and better able to answer national-level research questions whilst maintaining patient confidentiality and local governance procedures.

Chest CT and Hospital Outcomes in Patients with Omicron Compared with Delta Variant SARS-CoV-2 Infection.

Background The SARS-Cov-2 Omicron variant demonstrates rapid spread but with reduced disease severity. Studies evaluating the lung imaging findings of Omicron infection versus non-Omicron variants remain lacking. Purpose To compare Omicron and Delta variants of SARS-CoV-2 by their chest CT radiological pattern, biochemical parameters, clinical severity and hospital outcomes after adjusting for vaccination status. Materials and Methods Retrospective study of hospitalized adult patients rt-PCR positive for SARS-CoV-2 with CT pulmonary angiography performed within 7 days of admission between December 1, 2021 and January 14, 2022. Blinded radiological analysis with multiple readers including RSNA CT classification, chest CT severity score (CT-SS, range 0 least severe to 25 most severe) and CT imaging features including bronchial wall thickening. Results 106 patients (Delta n=66, Omicron n=40) were evaluated (mean age, 58 years ± 18, 58 men). In the Omicron group, 37% (15/40) of CT pulmonary angiograms were categorized as normal compared with 15% (10/66) in the Delta group (p=.016). Using a generalized linear model to control for confounding variables, including vaccination status, Omicron variant infection was associated with a CT-SS that was lower by 7.2 points compared to infection with Delta variant (ß=-7.2, 95%CI: -9.9, -4.5; p <.001). Bronchial wall thickening was more common with Omicron than with the Delta variant (odds ratio [OR] 2.4, 95%CI: 1.01, 5.92, p=.04). Vaccination with a booster shot was associated with a protective effect on chest infection compared with the unvaccinated (CT-SS median 5 (IQR 0-11), CT-SS median 11 (IQR 7.5-14), respectively; p = .03). The Delta variant was associated with a higher odds ratio of severe disease (OR 4.6, 95%CI: 1.2, 26, p=.01) and critical care admission (OR 7.0, 95%CI: 1.5, 66, p=.004) than the Omicron variant. Conclusion The SARS-COV-2 Omicron variant was associated with fewer and less severe changes on chest CT compared with the Delta variant. Patients with Omicron had greater frequency of bronchial wall thickening but lower clinical severity and improved hospital outcomes than those with Delta.

Differential occupational risks to healthcare workers from SARS-CoV-2 observed during a prospective observational study.

We conducted voluntary Covid-19 testing programmes for symptomatic and asymptomatic staff at a UK teaching hospital using naso-/oro-pharyngeal PCR testing and immunoassays for IgG antibodies. 1128/10,034 (11.2%) staff had evidence of Covid-19 at some time. Using questionnaire data provided on potential risk-factors, staff with a confirmed household contact were at greatest risk (adjusted odds ratio [aOR] 4.82 [95%CI 3.45-6.72]). Higher rates of Covid-19 were seen in staff working in Covid-19-facing areas (22.6% vs. 8.6% elsewhere) (aOR 2.47 [1.99-3.08]). Controlling for Covid-19-facing status, risks were heterogenous across the hospital, with higher rates in acute medicine (1.52 [1.07-2.16]) and sporadic outbreaks in areas with few or no Covid-19 patients. Covid-19 intensive care unit staff were relatively protected (0.44 [0.28-0.69]), likely by a bundle of PPE-related measures. Positive results were more likely in Black (1.66 [1.25-2.21]) and Asian (1.51 [1.28-1.77]) staff, independent of role or working location, and in porters and cleaners (2.06 [1.34-3.15]).

SARS-CoV-2 antibody responses post-vaccination in UK healthcare workers with pre-existing medical conditions: a cohort study.

OBJECTIVES: To examine antibody responses after the second vaccination in healthcare workers (HCWs) with underlying health conditions. DESIGN: Cohort study. SETTING: Oxford University Hospitals in the United Kingdom. PARTICIPANTS: Healthcare workers who had SARS-CoV-2 serological data available and received two SARS-CoV- 2 vaccinations. PRIMARY OUTCOME: Peak SARS-CoV-2 anti-spike IgG responses after the second vaccination and associations with underlying health conditions and the estimated risk of severe COVID-19 using an occupational health risk assessment tool. METHODS: We used univariable and multivariable linear regression models to investigate associations between antibody levels and demographics (age, sex, ethnicity), healthcare role, body mass index, underlying health conditions, vaccination status, prior infection and the Association of Local Authority Medical Advisors COVID-age risk score. RESULTS: 1635 HCWs had anti-spike IgG measurements 14-84 days after second vaccination and data on any underlying health conditions. Only five HCWs (0.3%), all on immunosuppressive treatment, (including four organ transplant recipients), did not seroconvert after second vaccination. Antibody levels were independently lower with older age, diabetes, immunosuppression, respiratory disorders other than asthma and markedly so in organ transplant recipients. Levels were independently lower in ChAdOx1 versus BNT162b2 recipients and higher following previous infection. HCWs with 'very high' COVID-age risk scores had lower median antibody levels than those with 'low', 'medium' or 'high' risk scores; 4379 AU/mL, compared with 12 337 AU/mL, 9430 AU/mL and 10 524 AU/mL, respectively. CONCLUSIONS: Two vaccine doses are effective in generating antibody responses among HCWs, including those with a high occupational risk. However, HCWs with underlying health conditions, especially diabetes, immunosuppression and organ transplant, had lower antibody levels, and vaccine response monitoring may be needed.

Increasing test specificity without impairing sensitivity: lessons learned from SARS-CoV-2 serology.

BACKGROUND: Serological tests are widely used in various medical disciplines for diagnostic and monitoring purposes. Unfortunately, the sensitivity and specificity of test systems are often poor, leaving room for false-positive and false-negative results. However, conventional methods were used to increase specificity and decrease sensitivity and vice versa. Using SARS-CoV-2 serology as an example, we propose here a novel testing strategy: the 'sensitivity improved two-test' or 'SIT²' algorithm. METHODS: SIT² involves confirmatory retesting of samples with results falling in a predefined retesting zone of an initial screening test, with adjusted cut-offs to increase sensitivity. We verified and compared the performance of SIT² to single tests and orthogonal testing (OTA) in an Austrian cohort (1117 negative, 64 post-COVID-positive samples) and validated the algorithm in an independent British cohort (976 negatives and 536 positives). RESULTS: The specificity of SIT² was superior to single tests and non-inferior to OTA. The sensitivity was maintained or even improved using SIT² when compared with single tests or OTA. SIT² allowed correct identification of infected individuals even when a live virus neutralisation assay could not detect antibodies. Compared with single testing or OTA, SIT² significantly reduced total test errors to 0.46% (0.24-0.65) or 1.60% (0.94-2.38) at both 5% or 20% seroprevalence. CONCLUSION: For SARS-CoV-2 serology, SIT² proved to be the best diagnostic choice at both 5% and 20% seroprevalence in all tested scenarios. It is an easy to apply algorithm and can potentially be helpful for the serology of other infectious diseases.

Time-of-Day Variation in SARS-CoV-2 RNA Levels during the Second Wave of COVID-19.

Circadian rhythms influence and coordinate an organism's response to its environment and to invading pathogens. We studied the diurnal variation in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) RNA in nasal/throat swabs collected in late 2020 to spring 2021 in a population immunologically naïve to SARS-CoV-2 and prior to widespread vaccination. SARS-CoV-2 diagnostic PCR data from 1698 participants showed a significantly higher viral load in samples obtained in the afternoon, in males, and in hospitalised patients when linear mixed modelling was applied. This study illustrates the importance of recording sample collection times when measuring viral replication parameters in clinical and research studies.

SARS-CoV-2 antibody trajectories after a single COVID-19 vaccination with and without prior infection.

Given high SARS-CoV-2 incidence, coupled with slow and inequitable vaccine roll-out in many settings, there is a need for evidence to underpin optimum vaccine deployment, aiming to maximise global population immunity. We evaluate whether a single vaccination in individuals who have already been infected with SARS-CoV-2 generates similar initial and subsequent antibody responses to two vaccinations in those without prior infection. We compared anti-spike IgG antibody responses after a single vaccination with ChAdOx1, BNT162b2, or mRNA-1273 SARS-CoV-2 vaccines in the COVID-19 Infection Survey in the UK general population. In 100,849 adults median (50 (IQR: 37-63) years) receiving at least one vaccination, 13,404 (13.3%) had serological/PCR evidence of prior infection. Prior infection significantly boosted antibody responses, producing higher peak levels and/or longer half-lives after one dose of all three vaccines than those without prior infection receiving one or two vaccinations. In those with prior infection, the median time above the positivity threshold was >1 year after the first vaccination. Single-dose vaccination targeted to those previously infected may provide at least as good protection to two-dose vaccination among those without previous infection.

Fatal COVID-19 outcomes are associated with an antibody response targeting epitopes shared with endemic coronaviruses.

The role of immune responses to previously seen endemic coronavirus epitopes in severe acute respiratory coronavirus 2 (SARS-CoV-2) infection and disease progression has not yet been determined. Here, we show that a key characteristic of fatal outcomes with coronavirus disease 2019 (COVID-19) is that the immune response to the SARS-CoV-2 spike protein is enriched for antibodies directed against epitopes shared with endemic beta-coronaviruses and has a lower proportion of antibodies targeting the more protective variable regions of the spike. The magnitude of antibody responses to the SARS-CoV-2 full-length spike protein, its domains and subunits, and the SARS-CoV-2 nucleocapsid also correlated strongly with responses to the endemic beta-coronavirus spike proteins in individuals admitted to an intensive care unit (ICU) with fatal COVID-19 outcomes, but not in individuals with nonfatal outcomes. This correlation was found to be due to the antibody response directed at the S2 subunit of the SARS-CoV-2 spike protein, which has the highest degree of conservation between the beta-coronavirus spike proteins. Intriguingly, antibody responses to the less cross-reactive SARS-CoV-2 nucleocapsid were not significantly different in individuals who were admitted to an ICU with fatal and nonfatal outcomes, suggesting an antibody profile in individuals with fatal outcomes consistent with an "original antigenic sin" type response.

Symptoms and Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Positivity in the General Population in the United Kingdom.

BACKGROUND: "Classic" symptoms (cough, fever, loss of taste/smell) prompt severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) polymerase chain reaction (PCR) testing in the United Kingdom. Studies have assessed the ability of different symptoms to identify infection, but few have compared symptoms over time (reflecting variants) and by vaccination status. METHODS: Using the COVID-19 Infection Survey, sampling households across the United Kingdom, we compared symptoms in PCR-positives vs PCR-negatives, evaluating sensitivity of combinations of 12 symptoms (percentage symptomatic PCR-positives reporting specific symptoms) and tests per case (TPC) (PCR-positives or PCR-negatives reporting specific symptoms/ PCR-positives reporting specific symptoms). RESULTS: Between April 2020 and August 2021, 27 869 SARS-CoV-2 PCR-positive episodes occurred in 27 692 participants (median 42 years), of whom 13 427 (48%) self-reported symptoms ("symptomatic PCR-positives"). The comparator comprised 3 806 692 test-negative visits (457 215 participants); 130 612 (3%) self-reported symptoms ("symptomatic PCR-negatives"). Symptom reporting in PCR-positives varied by age, sex, and ethnicity, and over time, reflecting changes in prevalence of viral variants, incidental changes (eg, seasonal pathogens (with sore throat increasing in PCR-positives and PCR-negatives from April 2021), schools reopening) and vaccination rollout. After May 2021 when Delta emerged, headache and fever substantially increased in PCR-positives, but not PCR-negatives. Sensitivity of symptom-based detection increased from 74% using "classic" symptoms, to 81% adding fatigue/weakness, and 90% including all 8 additional symptoms. However, this increased TPC from 4.6 to 5.3 to 8.7. CONCLUSIONS: Expanded symptom combinations may provide modest benefits for sensitivity of PCR-based case detection, but this will vary between settings and over time, and increases tests/case. Large-scale changes to targeted PCR-testing approaches require careful evaluation given substantial resource and infrastructure implications.

Comparison of two T-cell assays to evaluate T-cell responses to SARS-CoV-2 following vaccination in naïve and convalescent healthcare workers.

T-cell responses to SARS-CoV-2 following infection and vaccination are less characterized than antibody responses, due to a more complex experimental pathway. We measured T-cell responses in 108 healthcare workers (HCWs) using the commercialized Oxford Immunotec T-SPOT Discovery SARS-CoV-2 assay service (OI T-SPOT) and the PITCH ELISpot protocol established for academic research settings. Both assays detected T-cell responses to SARS-CoV-2 spike, membrane, and nucleocapsid proteins. Responses were significantly lower when reported by OI T-SPOT than by PITCH ELISpot. Four weeks after two doses of either Pfizer/BioNTech BNT162b or ChAdOx1 nCoV-19 AZD1222 vaccine, the responder rate was 63% for OI T-SPOT Panels 1 + 2 (peptides representing SARS-CoV-2 spike protein excluding regions present in seasonal coronaviruses), 69% for OI T-SPOT Panel 14 (peptides representing the entire SARS-CoV-2 spike), and 94% for the PITCH ELISpot total spike. The two OI T-SPOT panels correlated strongly with each other showing that either readout quantifies spike-specific T-cell responses, although the correlation between the OI T-SPOT panels and the PITCH ELISpot total spike was moderate. The standardization, relative scalability, and longer interval between blood acquisition and processing are advantages of the commercial OI T-SPOT assay. However, the OI T-SPOT assay measures T-cell responses at a significantly lower magnitude compared to the PITCH ELISpot assay, detecting T-cell responses in a lower proportion of vaccinees. This has implications for the reporting of low-level T-cell responses that may be observed in patient populations and for the assessment of T-cell durability after vaccination.

Divergent trajectories of antiviral memory after SARS-CoV-2 infection.

The trajectories of acquired immunity to severe acute respiratory syndrome coronavirus 2 infection are not fully understood. We present a detailed longitudinal cohort study of UK healthcare workers prior to vaccination, presenting April-June 2020 with asymptomatic or symptomatic infection. Here we show a highly variable range of responses, some of which (T cell interferon-gamma ELISpot, N-specific antibody) wane over time, while others (spike-specific antibody, B cell memory ELISpot) are stable. We use integrative analysis and a machine-learning approach (SIMON - Sequential Iterative Modeling OverNight) to explore this heterogeneity. We identify a subgroup of participants with higher antibody responses and interferon-gamma ELISpot T cell responses, and a robust trajectory for longer term immunity associates with higher levels of neutralising antibodies against the infecting (Victoria) strain and also against variants B.1.1.7 (alpha) and B.1.351 (beta). These variable trajectories following early priming may define subsequent protection from severe disease from novel variants.

Real-world evaluation of rapid and laboratory-free COVID-19 triage for emergency care: external validation and pilot deployment of artificial intelligence driven screening.

BACKGROUND: Uncertainty in patients' COVID-19 status contributes to treatment delays, nosocomial transmission, and operational pressures in hospitals. However, the typical turnaround time for laboratory PCR remains 12-24 h and lateral flow devices (LFDs) have limited sensitivity. Previously, we have shown that artificial intelligence-driven triage (CURIAL-1.0) can provide rapid COVID-19 screening using clinical data routinely available within 1 h of arrival to hospital. Here, we aimed to improve the time from arrival to the emergency department to the availability of a result, do external and prospective validation, and deploy a novel laboratory-free screening tool in a UK emergency department. METHODS: We optimised our previous model, removing less informative predictors to improve generalisability and speed, developing the CURIAL-Lab model with vital signs and readily available blood tests (full blood count [FBC]; urea, creatinine, and electrolytes; liver function tests; and C-reactive protein) and the CURIAL-Rapide model with vital signs and FBC alone. Models were validated externally for emergency admissions to University Hospitals Birmingham, Bedfordshire Hospitals, and Portsmouth Hospitals University National Health Service (NHS) trusts, and prospectively at Oxford University Hospitals, by comparison with PCR testing. Next, we compared model performance directly against LFDs and evaluated a combined pathway that triaged patients who had either a positive CURIAL model result or a positive LFD to a COVID-19-suspected clinical area. Lastly, we deployed CURIAL-Rapide alongside an approved point-of-care FBC analyser to provide laboratory-free COVID-19 screening at the John Radcliffe Hospital (Oxford, UK). Our primary improvement outcome was time-to-result, and our performance measures were sensitivity, specificity, positive and negative predictive values, and area under receiver operating characteristic curve (AUROC). FINDINGS: 72 223 patients met eligibility criteria across the four validating hospital groups, in a total validation period spanning Dec 1, 2019, to March 31, 2021. CURIAL-Lab and CURIAL-Rapide performed consistently across trusts (AUROC range 0·858-0·881, 95% CI 0·838-0·912, for CURIAL-Lab and 0·836-0·854, 0·814-0·889, for CURIAL-Rapide), achieving highest sensitivity at Portsmouth Hospitals (84·1%, Wilson's 95% CI 82·5-85·7, for CURIAL-Lab and 83·5%, 81·8-85·1, for CURIAL-Rapide) at specificities of 71·3% (70·9-71·8) for CURIAL-Lab and 63·6% (63·1-64·1) for CURIAL-Rapide. When combined with LFDs, model predictions improved triage sensitivity from 56·9% (51·7-62·0) for LFDs alone to 85·6% with CURIAL-Lab (81·6-88·9; AUROC 0·925) and 88·2% with CURIAL-Rapide (84·4-91·1; AUROC 0·919), thereby reducing missed COVID-19 cases by 65% with CURIAL-Lab and 72% with CURIAL-Rapide. For the prospective deployment of CURIAL-Rapide, 520 patients were enrolled for point-of-care FBC analysis between Feb 18 and May 10, 2021, of whom 436 received confirmatory PCR testing and ten (2·3%) tested positive. Median time from arrival to a CURIAL-Rapide result was 45 min (IQR 32-64), 16 min (26·3%) sooner than with LFDs (61 min, 37-99; log-rank p<0·0001), and 6 h 52 min (90·2%) sooner than with PCR (7 h 37 min, 6 h 5 min to 15 h 39 min; p<0·0001). Classification performance was high, with sensitivity of 87·5% (95% CI 52·9-97·8), specificity of 85·4% (81·3-88·7), and negative predictive value of 99·7% (98·2-99·9). CURIAL-Rapide correctly excluded infection for 31 (58·5%) of 53 patients who were triaged by a physician to a COVID-19-suspected area but went on to test negative by PCR. INTERPRETATION: Our findings show the generalisability, performance, and real-world operational benefits of artificial intelligence-driven screening for COVID-19 over standard-of-care in emergency departments. CURIAL-Rapide provided rapid, laboratory-free screening when used with near-patient FBC analysis, and was able to reduce the number of patients who tested negative for COVID-19 but were triaged to COVID-19-suspected areas. FUNDING: The Wellcome Trust, University of Oxford Medical and Life Sciences Translational Fund.

An Observational Cohort Study on the Incidence of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Infection and B.1.1.7 Variant Infection in Healthcare Workers by Antibody and Vaccination Status.

BACKGROUND: Natural and vaccine-induced immunity will play a key role in controlling the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic. SARS-CoV-2 variants have the potential to evade natural and vaccine-induced immunity. METHODS: In a longitudinal cohort study of healthcare workers (HCWs) in Oxfordshire, United Kingdom, we investigated the protection from symptomatic and asymptomatic polymerase chain reaction (PCR)-confirmed SARS-CoV-2 infection conferred by vaccination (Pfizer-BioNTech BNT162b2, Oxford-AstraZeneca ChAdOx1 nCOV-19) and prior infection (determined using anti-spike antibody status), using Poisson regression adjusted for age, sex, temporal changes in incidence and role. We estimated protection conferred after 1 versus 2 vaccinations and from infections with the B.1.1.7 variant identified using whole genome sequencing. RESULTS: In total, 13 109 HCWs participated; 8285 received the Pfizer-BioNTech vaccine (1407 two doses), and 2738 the Oxford-AstraZeneca vaccine (49 two doses). Compared to unvaccinated seronegative HCWs, natural immunity and 2 vaccination doses provided similar protection against symptomatic infection: no HCW vaccinated twice had symptomatic infection, and incidence was 98% lower in seropositive HCWs (adjusted incidence rate ratio 0.02 [95% confidence interval {CI} < .01-.18]). Two vaccine doses or seropositivity reduced the incidence of any PCR-positive result with or without symptoms by 90% (0.10 [95% CI .02-.38]) and 85% (0.15 [95% CI .08-.26]), respectively. Single-dose vaccination reduced the incidence of symptomatic infection by 67% (0.33 [95% CI .21-.52]) and any PCR-positive result by 64% (0.36 [95% CI .26-.50]). There was no evidence of differences in immunity induced by natural infection and vaccination for infections with S-gene target failure and B.1.1.7. CONCLUSIONS: Natural infection resulting in detectable anti-spike antibodies and 2 vaccine doses both provide robust protection against SARS-CoV-2 infection, including against the B.1.1.7 variant.