Bayesian Prediction of Severe Outcomes in the LabMarCS: Laboratory Markers of COVID-19 Severity - Bristol Cohort

ObjectivesTo develop cross-validated prediction models for severe outcomes in COVID-19 using blood biomarker and demographic data; Demonstrate best practices for clinical data curation and statistical modelling decisions, with an emphasis on Bayesian methods. DesignRetrospective observational cohort study. SettingMulticentre across National Health Service (NHS) trusts in Southwest region, England, UK. ParticipantsHospitalised adult patients with a positive SARS-CoV 2 by PCR during the first wave (March - October 2020). 843 COVID-19 patients (mean age 71, 45% female, 32% died or needed ICU stay) split into training (n=590) and validation groups (n=253) along with observations on demographics, co-infections, and 30 laboratory blood biomarkers. Primary outcome measuresICU admission or death within 28-days of admission to hospital for COVID-19 or a positive PCR result if already admitted. ResultsPredictive regression models were fit to predict primary outcomes using demographic data and initial results from biomarker tests collected within 3 days of admission or testing positive if already admitted. Using all variables, a standard logistic regression yielded an internal validation median AUC of 0.7 (95% Interval [0.64,0.81]), and an external validation AUC of 0.67 [0.61, 0.71], a Bayesian logistic regression using a horseshoe prior yielded an internal validation median AUC of 0.78 [0.71, 0.85], and an external validation median AUC of 0.70 [0.68, 0.71]. Variable selection performed using Bayesian predictive projection determined a four variable model using Age, Urea, Prothrombin time and Neutrophil-Lymphocyte ratio, with a median AUC of 0.74 [0.67, 0.82], and external validation AUC of 0.70 [0.69, 0.71]. ConclusionsOur study reiterates the predictive value of previously identified biomarkers for COVID-19 severity assessment. Given the small data set, the full and reduced models have decent performance, but would require improved external validation for clinical application. The study highlights a variety of challenges present in complex medical data sets while maintaining best statistical practices with an emphasis on showcasing recent Bayesian methods.

Contact patterns before and during the UK's Autumn 2020 COVID-19 lockdown among university students and staff

IntroductionUK universities re-opened in September 2020, despite the on-going coronavirus epidemic. During the first term, various national social distancing measures were introduced, including banning groups of >6 people and the second lockdown in November. COVID-19 can spread rapidly in university-settings, and students adherence to social distancing measures is critical for controlling transmission. MethodsWe measured university staff and student contact patterns via an online, longitudinal survey capturing self-reported contacts on the previous day. We investigated the change in contacts associated with COVID-19 guidance periods: post-first lockdown (23/06/2020-03/07/2020), relaxed guidance period (04/07/2020-13/09/2020), "rule-of-six" period (14/09/2020-04/11/2020), and the second lockdown (05/11/2020-25/11/2020). Results722 staff (4199 responses) (mean household size: 2.6) and 738 students (1906 responses) (mean household size: 4.5) were included in the study. Contact number decreased with age. Staff in single-person households reported fewer contacts than individuals in 2-and 3-person households, and individuals in 4-and 5-person households reported more contacts. For staff, daily contacts were higher in the relaxed guidance and "rule-of-six" periods (means: 3.2 and 3.5, respectively; medians: 3) than the post-first lockdown and second lockdown periods (means: 4.5 and 5.4, respectively; medians: 2). Few students responded until 05/10/2020, after which the median student contacts was 2 and the mean was 5.7, until the second lockdown when it dropped to 3.1. DiscussionUniversity staff and students responded to national guidance by altering their social contacts. The response in staff and students was similar, suggesting that students are able to adhere to social distancing guidance while at university.

Contacts and behaviours of university students during the COVID-19 pandemic at the start of the 2020/21 academic year

CONQUEST (COroNavirus QUESTionnaire) is an online survey of contacts, behaviour, and COVID-19 symptoms for University of Bristol (UoB) staff/students. We analysed survey results from the start of the 2020/2021 academic year, prior to the second national lockdown (14/09/2020-01/11/2020), where COVID-19 outbreaks led to lockdown of some student halls of residence. The aim of these analyses was to enhance knowledge of student contact patterns to inform infection disease mathematical modelling approaches. Responses captured information on demographics, contacts on the previous day, symptoms and self-isolation during the prior week, and COVID-19 status. 740 students provided 1261 unique records. Of 42 (3%) students testing positive in the prior fortnight, 99% had been self-isolating. The median number of contacts on the previous day was 2 (interquartile range: 1-5), mode: 1, mean: 6.1; 8% had [≥]20 contacts. 57% of student contacts were other UoB students/staff. Most students reported few daily contacts but there was heterogeneity, and some reported many. Around 40% of student contacts were with individuals not affiliated with UoB, indicating potential for transmission to non-students/staff.

High COVID-19 transmission potential associated with re-opening universities can be mitigated with layered interventions

Background: Re-opening universities while controlling COVID-19 transmission poses unique challenges. UK universities typically host 20,000 to 40,000 undergraduate students, with the majority moving away from home to attend. In the absence of realistic mixing patterns, previous models suggest that outbreaks associated with universities re-opening are an eventuality. Methods: We developed a stochastic transmission model based on realistic mixing patterns between students. We evaluated alternative mitigation interventions for a representative university. Results: Our model predicts, for a set of plausible parameter values, that if asymptomatic cases are half as infectious as symptomatic cases then 5,760 (3,940 - 7,430) out of 28,000 students, 20% (14% - 26%), could be infected during the first term, with 950 (656 - 1,209) cases infectious on the last day of term. If asymptomatic cases are as infectious as symptomatic cases then three times as many cases could occur, with 94% (93% - 94%) of the student population getting infected during the first term. We predict that one third of infected students are likely to be in their first year, and first year students are the main drivers of transmission due to high numbers of contacts in communal residences. We find that reducing face-to-face teaching is likely to be the single most effective intervention, and this conclusion is robust to varying assumptions about asymptomatic transmission. Supplementing reduced face-to-face testing with COVID-secure interactions and reduced living circles could reduce the percentage of infected students by 75%. Mass testing of students would need to occur at least fortnightly, is not the most effective option considered, and comes at a cost of high numbers of students requiring self-isolation. When transmission is controlled in the student population, limiting imported infection from the community is important. Conclusions: Priority should be given to understanding the role of asymptomatic transmission in the spread of COVID-19. Irrespective of assumptions about asymptomatic transmission, our findings suggest that additional outbreak control measures should be considered for the university setting. These might include reduced face-to-face teaching, management of student mixing and enhanced testing. Onward transmission to family members at the end of term is likely without interventions.

COVID-19 transmission in a university setting: a rapid review of modelling studies

Managing COVID-19 within a university setting presents unique challenges. At the start of term, students arrive from geographically diverse locations and potentially have higher numbers of social contacts than the general population, particularly if living in university halls of residence accommodation. Mathematical models are useful tools for understanding the potential spread of infection and are being actively used to inform policy about the management of COVID-19. Our aim was to provide a rapid review and appraisal of the literature on mathematical models investigating COVID-19 infection in a university setting. We searched PubMed, Web of Science, bioRxiv/ medRxiv and sought expert input via social media to identify relevant papers. BioRxiv/ medRxiv and PubMed/Web of Science searches took place on 3 and 6 July 2020, respectively. Papers were restricted to English language. Screening of peer-reviewed and pre-print papers and contact with experts yielded five relevant papers - all of which were pre-prints. All models suggest a significant potential for transmission of COVID-19 in universities. Testing of symptomatic persons and screening of the university community regardless of symptoms, combined with isolation of infected individuals and effective contact tracing were critical for infection control in the absence of other mitigation interventions. When other mitigation interventions were considered (such as moving teaching online, social/physical distancing, and the use of face coverings) the additional value of screening for infection control was limited. Multiple interventions will be needed to control infection spread within the university setting and the interaction with the wider community is an important consideration. Isolation of identified cases and quarantine of contacts is likely to lead to large numbers of students requiring educational, psychological and behavioural support and will likely have a large impact on the attendance of students (and staff), necessitating online options for teaching, even where in-person classes are taking place. Models were highly sensitive to assumptions in the parameters, including the number and type of individuals contacts, number of contacts traced, frequency of screening and delays in testing. Future models could aid policy decisions by considering the incremental benefit of multiple interventions and using empirical data on mixing within the university community and with the wider community where available. Universities will need to be able to adapt quickly to the evolving situation locally to support the health and wellbeing of the university and wider communities.