Matthew Whitaker

Background: The COVID-19 pandemic induced an unprecedented response from the scientific research community. Previous studies have described disruption of the norms of academic publishing during this time. This study uses an epidemiological statistical toolkit alongside machine-learning methods to investigate the functioning of the scientific information-generation and -consumption ecosystem throughout the pandemic. Methods: A dataset of 17 million scientific research papers that were published between January 2019 and December 2022 was analysed. Data on citations and Altmetrics were harvested, and topic modelling was applied to abstracts. COVID-19-related articles were identified from title text. We investigated publication dynamics, correlations between citation metrics and Altmetrics, rates of publication in preprints, and temporal trends in topics, and compared these metrics in COVID-19 papers vs non-COVID-19 papers. Results: Throughout 2020–2, 3.7% of English-language research output was on the topic of COVID-19. Journal articles on COVID-19 were published at a consistent rate during this period, while preprints peaked in early 2020 and decreased thereafter. COVID-19 preprints had lower publication rates in the peer-reviewed literature than other preprints, particularly those that were preprinted during early 2020. COVID-19 research received significantly more media and social media attention than non-COVID-19 research, and preprints received more attention, on average, than journal articles, with attention peaking during the initial wave and subsequent peaks corresponding to the emergence of novel variants. COVID-19 articles exhibited a higher correlation between Altmetrics and citation metrics compared with non-COVID-19 publications, suggesting a strong alignment between scientific and public attention. Conclusion: This study provides a comprehensive description of the rapid expansion of COVID-19 research, revealing evolving research areas and waxing and waning public interest across different topics. Preprints played an important role in disseminating scientific findings, but the level of coverage of preprinted findings emphasizes the need for guidelines in handling preprint research in media, particularly during a pandemic.

Objective: To determine how people in England changed their infection risk-related behaviours during the COVID-19 pandemic and in response to control measures, 19 June 2020 to 31 March 2022. Design: Over 18 (of 19) rounds, randomly selected participants across England completed a questionnaire about risk-related behaviours, socio-demographics, and symptoms. Participants: Between 85,018 and 154,060 randomly selected participants per round, aged 5+ years, totalling 2,177,657 responses with relevant data. Main outcome measures: Primary outcomes were self-reported shielding and/or taking specific precautions, not leaving home in the prior week, not being in close proximity with anyone outside their household the day before, and wearing face coverings outside the home. Secondary community-level measures of mobility and public health policy stringency were compared to the primary outcomes to provide population-level context to the observed findings. Results: Infection risk-related behaviours varied considerably over the nearly two years under study. Protective behaviours peaked in January 2021, during England’s winter wave, before widespread vaccination. At that time, the estimated proportion of self-reported shielding and/or taking specific precautions reached 21.6% (95% confidence interval 21.4% to 21.8%), of self-reported not leaving home in the week prior to completing the questionnaire reached 7.99% (7.85% to 8.13%) and of self-reported not having contact with anyone outside their household on the day before answering the questionnaire reached 89.2% (89.1% to 89.4%). As self-reported vaccination rates increased and prevalence of infection decreased, protective behaviours decreased, although patterns varied by demographics. Protective behaviours were strongly correlated with community-level mobility data and the stringency of public health measures. Conclusions: Individual-based data showed sizeable proportions of people undertook protective behaviours during the pandemic especially during the second lockdown in January 2021 although there was evidence of “pandemic fatigue” in the study’s later stages. Self-reported behaviours were closely aligned with community mobility data and the stringency of government policies, indicating policy-driven behaviour changes.

Background Pooled Cohort Equations (PCE) are used to predict cardiovascular disease (CVD) risk. Inclusion of other variables may improve risk prediction. Objective Identify variables improving CVD risk prediction beyond recalibrated PCE. Design Prospective cohort study; sex-stratified Cox survival models with LASSO stability selection to predict CVD in non-overlapping subsets: variable selection (40%), model training (30%) and testing (30%). Setting UK population. Participants UK Biobank: 121,724 and 182,632 healthy men and women, respectively, aged 38-73 years at baseline. Measurements Personal/family medical history; lifestyle factors; genetic, biochemical, hematological, and metabolomic blood markers. Outcomes were incident hospitalization or mortality from CVD. Results There were 11,899 (men) and 9,110 (women) incident CVD cases with median 12.1 years follow-up. Variables selected for both men and women were: age, albumin, antihypertensive medication, apolipoprotein B, atrial fibrillation, C-reactive protein, current smoker, cystatin C, family history of coronary artery disease, glycated hemoglobin, polygenic risk score (PRS) for CVD and systolic blood pressure. Also selected: apolipoprotein A1, lipoprotein(a), white blood cell count, deprivation index (men); triglycerides (women). C-statistics for recalibrated PCE were 0.67 [0.66-0.68] and 0.69 [0.68-0.70] in men and women, respectively, improving to 0.71 [0.70-0.72] and 0.72 [0.71-0.73] with LASSO stably selected variables. Categorical net reclassification improvement (7.5% risk threshold) versus PCE was 0.054 [0.038-0.070] (men) and 0.081 [0.063-0.099] (women). Addition of targeted metabolomic data to LASSO stability selection did not improve predictive accuracy. Limitations Analyses were done in a single population study and require external replication. Conclusion Additional personal/family medical history, blood-based markers and genetic information improve CVD risk prediction beyond PCE. Funding source National Institute for Health Research Academic Clinical Fellowship (JE); Medical Research Council studentship (BB); European Union H2020 (MC-H).

Infection with SARS-CoV-2 virus is associated with a wide range of symptoms. The REal-time Assessment of Community Transmission −1 (REACT-1) study monitored the spread and clinical manifestation of SARS-CoV-2 among random samples of the population in England from 1 May 2020 to 31 March 2022. We show changing symptom profiles associated with the different variants over that period, with lower reporting of loss of sense of smell or taste for Omicron compared to previous variants, and higher reporting of cold-like and influenza-like symptoms, controlling for vaccination status. Contrary to the perception that recent variants have become successively milder, Omicron BA.2 was associated with reporting more symptoms, with greater disruption to daily activities, than BA.1. With restrictions lifted and routine testing limited in many countries, monitoring the changing symptom profiles associated with SARS-CoV-2 infection and effects on daily activities will become increasingly important.

Rapid transmission of the SARS-CoV-2 Omicron variant has led to record-breaking case incidence rates around the world. Since May 2020, the REal-time Assessment of Community Transmission-1 (REACT-1) study tracked the spread of SARS-CoV-2 infection in England through RT-PCR of self-administered throat and nose swabs from randomly-selected participants aged 5 years and over. In January 2022, we found an overall weighted prevalence of 4.41% (n = 102,174), three-fold higher than in November to December 2021; we sequenced 2,374 (99.2%) Omicron infections (19 BA.2), and only 19 (0.79%) Delta, with a growth rate advantage for BA.2 compared to BA.1 or BA.1.1. Prevalence was decreasing overall (reproduction number R = 0.95, 95% credible interval [CrI], 0.93, 0.97), but increasing in children aged 5 to 17 years (R = 1.13, 95% CrI, 1.09, 1.18). In England during January 2022, we observed unprecedented levels of SARS-CoV-2 infection, especially among children, driven by almost complete replacement of Delta by Omicron.

Background: Lateral flow immunoassays (LFIAs) are being used worldwide for COVID-19 mass testing and antibody prevalence studies. Relatively simple to use and low cost, these tests can be self-administered at home, but rely on subjective interpretation of a test line by eye, risking false positives and false negatives. Here, we report on the development of ALFA (Automated Lateral Flow Analysis) to improve reported sensitivity and specificity. Methods: Our computational pipeline uses machine learning, computer vision techniques and signal processing algorithms to analyse images of the Fortress LFIA SARS-CoV-2 antibody self-test, and subsequently classify results as invalid, IgG negative and IgG positive. A large image library of 595,339 participant-submitted test photographs was created as part of the REACT-2 community SARS-CoV-2 antibody prevalence study in England, UK. Alongside ALFA, we developed an analysis toolkit which could also detect device blood leakage issues. Results: Automated analysis showed substantial agreement with human experts (Cohen’s kappa 0.90–0.97) and performed consistently better than study participants, particularly for weak positive IgG results. Specificity (98.7–99.4%) and sensitivity (90.1–97.1%) were high compared with visual interpretation by human experts (ranges due to the varying prevalence of weak positive IgG tests in datasets). Conclusions: Given the potential for LFIAs to be used at scale in the COVID-19 response (for both antibody and antigen testing), even a small improvement in the accuracy of the algorithms could impact the lives of millions of people by reducing the risk of false-positive and false-negative result read-outs by members of the public. Our findings support the use of machine learning-enabled automated reading of at-home antibody lateral flow tests as a tool for improved accuracy for population-level community surveillance.

Rapid transmission of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Omicron variant has led to record-breaking incidence rates around the world. The Real-time Assessment of Community Transmission-1 (REACT-1) study has tracked SARS-CoV-2 infection in England using reverse transcription polymerase chain reaction (RT-PCR) results from self-administered throat and nose swabs from randomly selected participants aged 5 years and older approximately monthly from May 2020 to March 2022. Weighted prevalence in March 2022 was the highest recorded in REACT-1 at 6.37% ( N = 109,181), with the Omicron BA.2 variant largely replacing the BA.1 variant. Prevalence was increasing overall, with the greatest increase in those aged 65 to 74 years and 75 years and older. This was associated with increased hospitalizations and deaths, but at much lower levels than in previous waves against a backdrop of high levels of vaccination.

Infection with SARS-CoV-2 virus is associated with a wide range of symptoms. The REal-time Assessment of Community Transmission -1 (REACT-1) study has been monitoring the spread and clinical manifestation of SARS-CoV-2 among random samples of the population in England from 1 May 2020 to 31 March 2022. We show changing symptom profiles associated with the different variants over that period, with lower reporting of loss of sense of smell and taste for Omicron compared to previous variants, and higher reporting of cold-like and influenza-like symptoms, controlling for vaccination status. Contrary to the perception that recent variants have become successively milder, Omicron BA.2 was associated with reporting more symptoms, with greater disruption to daily activities, than BA.1. With restrictions lifted and routine testing limited in many countries, monitoring the changing symptom profiles associated with SARS-CoV-2 infection and induced changes in daily activities will become increasingly important.

Long COVID remains a broadly defined syndrome, with estimates of prevalence and duration varying widely. We use data from rounds 3–5 of the REACT-2 study (n = 508,707; September 2020 – February 2021), a representative community survey of adults in England, and replication data from round 6 (n = 97,717; May 2021) to estimate the prevalence and identify predictors of persistent symptoms lasting 12 weeks or more; and unsupervised learning to cluster individuals by reported symptoms. At 12 weeks in rounds 3–5, 37.7% experienced at least one symptom, falling to 21.6% in round 6. Female sex, increasing age, obesity, smoking, vaping, hospitalisation with COVID-19, deprivation, and being a healthcare worker are associated with higher probability of persistent symptoms in rounds 3–5, and Asian ethnicity with lower probability. Clustering analysis identifies a subset of participants with predominantly respiratory symptoms. Managing the long-term sequelae of COVID-19 will remain a major challenge for affected individuals and their families and for health services.