Prediction of SARS-CoV-2 PCR positivity in UK Fenland study participants using smartphone-submitted vital sign measurements (preprint)

IntroductionMobile health applications are increasingly being used in health and clinical research. SARS-CoV-2 has proven to have high infectivity, making outbreaks difficult to contain. Early detection can help prevent spread, but there is a need to develop easy-to-use screening tools that can help identify potential infection as early as possible. Here, we describe the development of a machine learning classifier that can predict SARS-CoV-2 PCR positivity using smartphone-submitted vital sign measurements.MethodsThe Fenland App study followed 2,199 UK participants using a smartphone application from August 2020 and for a minimum of six months. Participants completed a baseline questionnaire and then monthly questionnaires about SARS-CoV-2 status and vaccinations. Three times a week, participants provided measurements of their blood oxygen saturation, body temperature, and resting heart rate via a pulse oximeter, digital thermometer, and their smartphone. The participants participated in self initiated SARS-CoV-2 testing as per concurrent public health guidelines.We built predictive models SARS-CoV-2 PCR positivity status as obtained from national surveillance PCR test results.ResultsA total of 77 positive and 6,339 negative SARS-CoV-2 tests were recorded during the study. The final model achieved an ROC AUC of 0.695 ± 0.045. There was no difference in model performance when using 4, 8 or 12 weeks of baseline data before a SARS-CoV-2 test (F(2) = 0.80, p = 0.472). Addition of demographic or symptom information had no impact on model performance.ConclusionsUsing only three smartphone collected vital sign measurements, it is possible to predict SARS-CoV-2 PCR positivity, using a four week baseline period. Smartphone based remote monitoring of patient vital signs could allow for earlier screening for potential infections. This method could be applicable to any infectious disease that causes physiological changes in vital signs.

The removal of airborne SARS-CoV-2 and other microbial bioaerosols by air filtration on COVID-19 surge units (preprint)

Summary Background The COVID-19 pandemic has overwhelmed the respiratory isolation capacity in hospitals; many wards lacking high-frequency air changes have been repurposed for managing patients infected with SARS-CoV-2 requiring either standard or intensive care. Hospital-acquired COVID-19 is a recognised problem amongst both patients and staff, with growing evidence for the relevance of airborne transmission. This study examined the effect of air filtration and ultra-violet (UV) light sterilisation on detectable airborne SARS-CoV-2 and other microbial bioaerosols. Methods We conducted a crossover study of portable air filtration and sterilisation devices in a repurposed ‘surge’ COVID ward and ‘surge’ ICU. National Institute for Occupational Safety and Health (NIOSH) cyclonic aerosol samplers and PCR assays were used to detect the presence of airborne SARS-CoV-2 and other microbial bioaerosol with and without air/UV filtration. Results Airborne SARS-CoV-2 was detected in the ward on all five days before activation of air/UV filtration, but on none of the five days when the air/UV filter was operational; SARS-CoV-2 was again detected on four out of five days when the filter was off. Airborne SARS-CoV-2 was infrequently detected in the ICU. Filtration significantly reduced the burden of other microbial bioaerosols in both the ward (48 pathogens detected before filtration, two after, p =0.05) and the ICU (45 pathogens detected before filtration, five after p =0.05). Conclusions These data demonstrate the feasibility of removing SARS-CoV-2 from the air of repurposed ‘surge’ wards and suggest that air filtration devices may help reduce the risk of hospital-acquired SARS-CoV-2. Funding Wellcome Trust, MRC, NIHR

Development and Validation of a Dynamic 48-Hour In-Hospital Mortality Risk Stratification for COVID-19 in a UK Teaching Hospital: A Retrospective Cohort Study (preprint)

Background: While numerous point-of-admission disease severity models for COVID-19 have been proposed, disease stratification that accounts for changes in a patient’s condition while in hospital is urgently needed to facilitate patient management and resource allocation.Methods: We developed a prognostic model for 48-hour in-hospital mortality using 473 consecutive patients with COVID-19 presenting to a UK hospital between March 1 and September 12, 2020; and temporally validated using data on 405 patients presenting between September 13, 2020 and January 3, 2021.The primary outcome was all-cause in-hospital mortality. We additionally considered the competing risks of discharge from hospital and transfer to a tertiary Intensive Care Unit for extracorporeal membrane oxygenation. We adopted a landmarking approach to dynamic prediction that accounts for competing risks and informative missingness, and selected predictors using penalised regression. The model estimates, at any point during a hospital visit, the probability of in-hospital mortality during the next 48 hours.Results: Our final model includes age, Clinical Frailty Scale score, heart rate, respiratory rate, SpO2/FiO2 ratio, white cell count, presence of acidosis (pH < 7.35) and Interleukin-6. Internal validation achieved an AUROC of 0.90 (95% CI 0.87–0.93) and temporal validation gave an AUROC of 0.91 (95% CI 0.88-0.95). Interpretation: Our model uniquely incorporates both static risk factors (e.g. age) and evolving clinical and laboratory data, to provide a dynamic risk prediction model that adapts to both sudden and gradual changes in an individual patient’s clinical condition. External validation outside the study hospital will be required before adoption.Funding: NIHR Cambridge Biomedical Research Centre, UKRI Medical Research CouncilDeclaration of Interest: None to declare. Ethical Approval: The study was approved by a UK Health Research Authority ethics committee (20/WM/0125). Patient consent was waived because the de-identified data presented here were collected during routine clinical practice; there was no requirement for informed consent.

Neutralizing Anti-interferon-γ Autoantibodies; An Ameliorating Factor in COVID-19 Infection? (preprint)

Background: Elevated levels of interferon-gamma (IFNγ) have been found in COVID-19 infection, however its role in this setting remains poorly understood. Cases of non-tuberculous mycobacterium (NTM) infections due to anti-IFNγ autoantibodies (Ab) were first reported in 2004. NTM and COVID-19 co-infection in a patient with acquired IFNγ deficiency has not previously been described. The impact of anti-IFNγ Ab on the severity of COVID-19 has not been previously explored. Objective: We report a case of COVID-19 infection in a patient hospitalised with NTM infection due to acquired IFNγ deficiency caused by anti-IFNγ Ab. We also explore effects of IFNγ Ab on the severity of COVID-19 infection. Methods: : Detailed immunological investigations were performed. Bio-rad, Bio-Plex methodology was used to detect anti-IFNγ Ab, titration, IFNγ recovery assay and SARS-CoV-2 serology. Anti-IFNγ Ab were tested in patients with severe (COV-Pat) and health care workers with mild/asymptomatic COVID-19 infection (COV-Asx). Results: : Mycobacterium avium intracellulare was diagnosed following bone marrow examination and culture. High titre anti-IFNγ Ab were detected in patient’s serum. The autoantibodies neutralized both endogenously produced and exogenously administered IFNγ. SARS-COV-2 infection was identified during routine inpatient testing. Despite prolonged SARS-COV-2 infection the patient showed only minimal additional symptoms, never developed any significant inflammatory complications and eventually mounted an adequate IgG antibody response to the SARS-COV-2 trimeric S-protein. Elevated titres of anti-IFNγ Ab were detected in COV-Pat and COV-Asx, compared to non-infected healthy controls. The titres were broadly similar between COV-Pat and COV-Asx groups, but much lower compared to patients with acquired IFNγ Ab deficiency. Conclusion: IFN-γ is known to play a central role in hyperinflammatory disease states such as macrophage activation syndrome This study illustrates the potential value of inhibiting IFNγ to prevent pathological inflammatory response to COVID-19.

Development and validation of a dynamic 48-hour in-hospital prognostic risk stratification for COVID-19 in a UK teaching hospital: a retrospective cohort study (preprint)

We propose a prognostic dynamic risk stratification for 48-hour in-hospital mortality in patients with COVID-19, using demographics and routinely-collected observations and laboratory tests: age, Clinical Frailty Scale score, heart rate, respiratory rate, SpO2/FiO2 ratio, white cell count, acidosis (pH < 7.35) and Interleukin-6. We train and validate the model using data from a UK teaching hospital, adopting a landmarking approach that accounts for competing risks and informative missingness. Internal validation of the model on the first wave of patients presenting between March 1 and September 12, 2020 achieves an AUROC of 0.90 (95% CI 0.87-0.93). Temporal validation on patients presenting between September 13, 2020 and January 1, 2021 gives an AUROC of 0.91 (95% CI 0.88-0.95). The resulting mortality stratification tool has the potential to provide physicians with an assessment of a patient's evolving prognosis throughout the course of active hospital treatment.

Neutralising antibodies drive Spike mediated SARS-CoV-2 evasion (preprint)

SARS-CoV-2 Spike protein is critical for virus infection via engagement of ACE2, and amino acid variation in Spike is increasingly appreciated. Given both vaccines and therapeutics are designed around Wuhan-1 Spike, this raises the theoretical possibility of virus escape, particularly in immunocompromised individuals where prolonged viral replication occurs. Here we report chronic SARS-CoV-2 with reduced sensitivity to neutralising antibodies in an immune suppressed individual treated with convalescent plasma, generating whole genome ultradeep sequences by both short and long read technologies over 23 time points spanning 101 days. Although little change was observed in the overall viral population structure following two courses of remdesivir over the first 57 days, N501Y in Spike was transiently detected at day 55 and V157L in RdRp emerged. However, following convalescent plasma we observed large, dynamic virus population shifts, with the emergence of a dominant viral strain bearing D796H in S2 and{Delta} H69/{Delta}V70 in the S1 N-terminal domain NTD of the Spike protein. As passively transferred serum antibodies diminished, viruses with the escape genotype diminished in frequency, before returning during a final, unsuccessful course of convalescent plasma. In vitro, the Spike escape double mutant bearing{Delta} H69/{Delta}V70 and D796H conferred decreased sensitivity to convalescent plasma, whilst maintaining infectivity similar to wild type. D796H appeared to be the main contributor to decreased susceptibility, but incurred an infectivity defect. The{Delta} H69/{Delta}V70 single mutant had two-fold higher infectivity compared to wild type and appeared to compensate for the reduced infectivity of D796H. Consistent with the observed mutations being outside the RBD, monoclonal antibodies targeting the RBD were not impacted by either or both mutations, but a non RBD binding monoclonal antibody was less potent against{Delta} H69/{Delta}V70 and the double mutant. These data reveal strong selection on SARS-CoV-2 during convalescent plasma therapy associated with emergence of viral variants with reduced susceptibility to neutralising antibodies.

A2B-COVID: A method for evaluating potential SARS-CoV-2 transmission events (preprint)

Identifying linked cases of infection is a key part of the public health response to viral infectious disease. Viral genome sequence data is of great value in this task, but requires careful analysis, and may need to be complemented by additional types of data. The Covid-19 pandemic has highlighted the urgent need for analytical methods which bring together sources of data to inform epidemiological investigations. We here describe A2B-COVID, an approach for the rapid identification of linked cases of coronavirus infection. Our method combines knowledge about infection dynamics, data describing the movements of individuals, and novel approaches to genome sequence data to assess whether or not cases of infection are consistent or inconsistent with linkage via transmission. We apply our method to analyse and compare data collected from two wards at Cambridge University Hospitals, showing qualitatively different patterns of linkage between cases on designated Covid-19 and non-Covid-19 wards. Our method is suitable for the rapid analysis of data from clinical or other potential outbreak settings.

Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans (preprint)

Machine learning methods offer great promise for fast and accurate detection and prognostication of COVID-19 from standard-of-care chest radiographs (CXR) and computed tomography (CT) images. Many articles have been published in 2020 describing new machine learning-based models for both of these tasks, but it is unclear which are of potential clinical utility. In this systematic review, we search EMBASE via OVID, MEDLINE via PubMed, bioRxiv, medRxiv and arXiv for published papers and preprints uploaded from January 1, 2020 to October 3, 2020 which describe new machine learning models for the diagnosis or prognosis of COVID-19 from CXR or CT images. Our search identified 2,212 studies, of which 415 were included after initial screening and, after quality screening, 61 studies were included in this systematic review. Our review finds that none of the models identified are of potential clinical use due to methodological flaws and/or underlying biases. This is a major weakness, given the urgency with which validated COVID-19 models are needed. To address this, we give many recommendations which, if followed, will solve these issues and lead to higher quality model development and well documented manuscripts.