The Spectre of Berkson's Paradox: Collider Bias in Covid-19 Research.

When non-random sampling collides with our understanding of Covid-19 risk, we must be careful not to draw incorrect conclusions about cause and effect. By Annie Herbert, Gareth Griffith, Gibran Hemani and Luisa Zuccolo.

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

Controlling COVID-19 transmission in universities poses challenges due to the complex social networks and potential for asymptomatic spread. We developed a stochastic transmission model based on realistic mixing patterns and evaluated alternative mitigation strategies. We predict, for plausible model parameters, that if asymptomatic cases are half as infectious as symptomatic cases, then 15% (98% Prediction Interval: 6-35%) of students could be infected during the first term without additional control measures. First year students are the main drivers of transmission with the highest infection rates, largely due to communal residences. In isolation, reducing face-to-face teaching is the most effective intervention considered, however layering multiple interventions could reduce infection rates by 75%. Fortnightly or more frequent mass testing is required to impact transmission and was not the most effective option considered. Our findings suggest that additional outbreak control measures should be considered for university settings.

Contacts and behaviours of university students during the COVID-19 pandemic at the start of the 2020/2021 academic year.

University students have unique living, learning and social arrangements which may have implications for infectious disease transmission. To address this data gap, we created CONQUEST (COroNavirus QUESTionnaire), a longitudinal online survey of contacts, behaviour, and COVID-19 symptoms for University of Bristol (UoB) staff/students. Here, we analyse results from 740 students providing 1261 unique records from the start of the 2020/2021 academic year (14/09/2020-01/11/2020), where COVID-19 outbreaks led to the self-isolation of all students in some halls of residences. Although most students reported lower daily contacts than in pre-COVID-19 studies, there was heterogeneity, with some reporting many (median = 2, mean = 6.1, standard deviation = 15.0; 8% had ≥ 20 contacts). Around 40% of students' contacts were with individuals external to the university, indicating potential for transmission to non-students/staff. Only 61% of those reporting cardinal symptoms in the past week self-isolated, although 99% with a positive COVID-19 test during the 2 weeks before survey completion had self-isolated within the last week. Some students who self-isolated had many contacts (mean = 4.3, standard deviation = 10.6). Our results provide context to the COVID-19 outbreaks seen in universities and are available for modelling future outbreaks and informing policy.

Collider bias undermines our understanding of COVID-19 disease risk and severity.

Numerous observational studies have attempted to identify risk factors for infection with SARS-CoV-2 and COVID-19 disease outcomes. Studies have used datasets sampled from patients admitted to hospital, people tested for active infection, or people who volunteered to participate. Here, we highlight the challenge of interpreting observational evidence from such non-representative samples. Collider bias can induce associations between two or more variables which affect the likelihood of an individual being sampled, distorting associations between these variables in the sample. Analysing UK Biobank data, compared to the wider cohort the participants tested for COVID-19 were highly selected for a range of genetic, behavioural, cardiovascular, demographic, and anthropometric traits. We discuss the mechanisms inducing these problems, and approaches that could help mitigate them. While collider bias should be explored in existing studies, the optimal way to mitigate the problem is to use appropriate sampling strategies at the study design stage.

The spectre of Berkson's paradox Collider bias in Covid-19 research

When non-random sampling collides with our understanding of Covid-19 risk, we must be careful not to draw incorrect conclusions about cause and effect. By Annie Herbert, Gareth Griffith, Gibran Hemani and Luisa Zuccolo.