Abstract
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.
Introduction
As COVID-19 continues to circulate in the population, universities are looking to decide how to keep students, staff and the wider community safe, while providing a high-quality experience for students. Educational settings are associated with a high number of social interactions and may be important drivers of transmission for COVID-19. At the present time a low proportion of the student population are expected to have been infected with SARS-CoV-2, this fact along with the characteristics and behaviours of student populations poses a real risk of rapidly increasing case numbers at the start of the academic term. This in turn presents challenges in preparation for, and ongoing management of, COVID-19 within a university setting. Students travel from all over the UK and abroad to attend university. Most are between 18–24 years of age and many live in large shared multi-occupancy households or university halls of residence (mainly in first year). Although this age group are at relatively low risk severe complications and death from COVID-19, evidence from the Office for National Statistics serosurvey1 shows that they have similar or higher levels of infection compared with other age groups and could, therefore, pass on the infection to university staff and the wider community who may be at higher risk of serious COVID-19 outcomes.
Mathematical models are currently being used to help understand the evolving COVID-19 pandemic and to inform prevention and control strategies. In the absence of a vaccine, a number of other mitigation interventions could be relevant in a university setting, including, but not limited to: face coverings, social/ physical distancing, replacing large group teaching (i.e. in traditional lecture theatres) with online learning, minimising contacts through “bubbles” for small group taught work or in accommodation, reduction or elimination of field work, testing on arrival on campus, regular individual or pooled testing, self-isolation (for those symptomatic, testing positive, or close contacts of a case), education to recognise possible symptoms, handwashing/hand sanitiser, increased ventilation, reduced occupancy of staff and students on campus, and increased cleaning. These interventions could have a varying impact on COVID-19 cases and require different resources to support them. Some UK universities are planning for students to return to campus in autumn 2020 in a blended teaching model, for example with large lectures replaced with online teaching and small group practicals delivered as physically distanced. Other UK universities are considering alternative teaching models, including some planning fully online delivery for at least the first semester (October 2020 – January 2021).
Here we provide a rapid review of both the peer-reviewed and pre-print literature on mathematical models of COVID-19 (SARS-CoV-2) infection in the university or college community. We consider models describing infection spread in the absence of intervention and those comparing different scenarios of control strategies; any intervention for COVID-19 control is considered. We provide a summary of the methodological features and principal outcomes from the models, in terms of number of cases and effectiveness of interventions in reducing spread of COVID-19 in this setting, and recommendations on key interventions for COVID-19 control.
Methods
This paper was prepared in accordance with the PRISMA guidelines2. We searched PubMed, Web of Science, bioRxiv, and medRxiv to identify papers, in English, reporting mathematical models of coronavirus (COVID-19) in a university setting. PubMed and Web of Science searches took place on 6 July 2020; BioRxiv/medRxiv searches took place on 3 July 2020. The search criteria are presented in Table 1. We also sought expert input to locate pre-prints and grey literature via social media (twitter). The results of searches were screened for inclusion by HC and KT initially on the title and abstract, then full text. Papers were included if they described a mathematical model of COVID-19 infection within a university or college setting, in any geographical location. An updated version of one paper was obtained directly from the author3. Information was extracted from the included papers by several members of the team (HC, RDB, AT, EN, GH). This information (detailed in Table 2) included the focus and principal finding of the paper, model type and structure, population modelled, interventions considered, and selected COVID-19 infection related parameters. We did not seek to formally assess risk of bias in the included studies; there is no validated risk of bias assessment for modelling studies. Results were summarised through a descriptive synthesis of each study using text and tables.
Search terms for the rapid review
Summary of reviewed papers included in the review
Results
Summary of study selection
An initial search of PubMed including (COVID or coronavirus or SARS) did not return any hits, so a wider search was included to identify potentially relevant peer-reviewed papers considering other infectious disease models of campus transmission. The search of PubMed yielded 127 hits of which 20 were shortlisted for full text screening and 0 were included in the review. The Web of Science search yielded 7 hits, of which 2 were shortlisted for full text screening and 0 included in the review. The search of bioRxiv and medRxiv returned 205 records; after screening of titles 22 were selected for full text assessment and 3 papers were included in the review. Excluded papers did not report COVID-19 transmission in a university setting. Request to experts (via social media) resulted in 6 papers of which 4 were included in the review. In total, after removing duplicates, we identified 5 papers describing mathematical models in a university context, 4 with a US focus and 1 from UK. A summary of the papers included in the review is presented in Table 2.
How SARS-CoV-2 infection is modelled in a university setting
All of the models included in the review are transmission dynamic, that is, the amount of SARS-CoV-2 transmission is related to the number of infected people over time and responds to changes in infection risk as a result of interventions – this is critical to assess the direct and indirect impact of potential control strategies. Most were compartmental models, which divides up the population into distinct groups; only one5 was agent-based which models individuals within a population. All models used a Susceptible-Exposed (latent)-Infectious-Recovered structure to capture the progression of infection in the population though the time period. Two models allowed for some individuals to be immune to COVID-19 at the start of the academic term5,6, though only Gressman and Peck5 included this in their base case model. The length of time for each infection state (for example, how long people were infectious for) and the assumptions about the transmissibility of the infection varied between models; only two models explicitly included pre-symptomatic transmission. Fundamentally, infection is spread between people through ‘effective’ contacts, so how people come into contact with each other is important. The way this is done across the five included papers is quite different, ranging from random mixing, to random mixing within a small number of different groups, to highly specified mixing patterns in the agent based model based on individuals’ shared activities. Altering the assumptions about these parameters makes a considerable difference to the estimated potential outbreak in a university setting and the impact of any control measures. All the models assume that once infected persons have recovered, they are immune to re-infection. Only three models consider importation of infection from the wider community into the university population; none of the models included in the review considered the exporting of infection from the university community into the wider population. Most models start with very few initial cases, in some instances zero cases with the university epidemic starting due to importation from the local community. For one paper the low starting number is due to testing of students prior to and on return to university, with isolation if positive (see below)4. Two papers present predictions for SARS-CoV-2 infections in the autumn term without any control measures3,5; one estimates 89.4% of the campus population would be infected by the end of the semester5 and the other predicted 19.5% of the university community would be symptomatic (many more would be infected and be asymptomatic). When explicitly considered some of these infections result in fatalities. Infection numbers increased rapidly in the absence of intervention6.
Interventions considered in models
Testing
Testing interventions included: testing students before they travelled to campus; testing students on return to campus; testing symptomatic persons3; and testing asymptomatic individuals3 (here we refer to this regular testing as screening when done on mass, others term this testing, screening, or surveillance). Only one paper considered testing students before they travelled to campus and the same paper also considered testing individuals upon their return to the university setting (pooled testing on arrival)4. Lopman et al. explicitly state they did not include return to campus testing because they felt it would have limited effect given in their model the student prevalence is assumed to be the same as the general population3. Several papers3,4,6,7 include near-immediate testing of symptomatic individuals; Gressman and Peck considered isolation but not testing of symptomatic persons5. Several symptoms associated with COVID-19 are generic to other infections. Particularly as we approach winter many persons may have what appear to be COVID-19 symptoms but not have SARS-CoV-2 – this has implications for the number of tests that may be needed and isolation/quarantine policies. Only one model included a background rate of influenza-like illness, with persons subsequently testing negative for SARS-CoV-23. All papers considered screening with the rate of testing asymptomatic persons per day varying between and within papers. Martin et al. considered varying random daily screening rates so that 25%, 50%, 75%, or 100% of the university community population would be screened per month. Cashore et al.4 investigated testing the entire university community on average once every 5 days, Gressman and Peck5 considered this every month, whilst Lopman et al.3 also looked at varying these screening rates from weekly to once a term. The V-KEMS paper looked at testing parameterised as a rate of infection detection per day (from 0 to 1)7. Gressman and Peck5 also included testing of those identified through contact tracing. The sensitivity and specificity of tests is not 100%. This has implications for the number of true positive cases not identified who can then go on to infect others and also how many individuals you may needlessly isolate and quarantine, discussed further below. Test sensitivity varied across the models from 75%3 to 97%5; only one paper considered variability in testing sensitivity over the course of an infection3.
Contact tracing
Contact tracing was included in three of the models3–5 and the number of traceable contacts modelled per case varied. Lopman et al. assumed that on average 14 contacts of individuals testing positive would be identified by public health authorities and 75% (10.5) of these would be successfully traced and quarantined. In the Gressman and Peck paper each individual testing positive or developing symptoms had their contacts from the last two days traced on the next day with an assumed 19 contacts, of which 58% (11) would be traceable; this is the only model that includes testing for quarantined persons. Cashore et al. included seven contacts meeting the criteria for quarantine for each person testing positive, and investigated via sensitivity analyses the effect of different delay periods for contact tracing and lengths of isolation periods for traced individuals.
Isolation and quarantining
All papers include isolation of symptomatic persons, though some include a delay between symptom onset and testing with isolation only happening once the person has tested positive. All of the reviewed papers except for that by Martin et al. included quarantining of traced contacts in their models, with most using a 14 day quarantine period, whilst V-KEMS have quarantine operating for ‘longer than a reasonable infectious period’7. Where included, models assume quarantining reduces contacts to zero, meaning quarantined persons cannot infect others, and if uninfected, are protected from further exposure. The only paper to explicitly include false positives from testing (specificity) is Gressman and Peck, which is important for considering how many people may be subject to quarantine despite not being in contact with a positive case. None of the papers considered quarantining of individuals traveling from high-risk COVID-19 areas as a stand-alone intervention.
Additional non-pharmaceutical interventions
All but one paper considered some form of other non-pharmaceutical intervention. These included reducing contacts through moving face-to-face teaching online4,5, social distancing3, and reducing contacts through the use of social bubbles7, or reducing transmission in face-to-face interactions by wearing face coverings.3,5 In both Gressman and Peck and Lopman et al. these are modelled by having a term that generally reduces infection transmission by a specified amount, so could encompass a variety of non-pharmaceutical interventions (for example, increased hand hygiene).
Synthesis of key model findings
In the absence of control interventions, all models suggest a significant potential for transmission of COVID-19 in universities with rapid growth in case numbers shortly after the start of the academic year once the students return to campus. Analysis of where in the country, and which countries, students will be coming from may indicate how many infected persons a university might expect on the first day of term. Most models considered only a very low number of infected students at the start of term so infection spread could be quicker than presented in these papers, although this is highly dependent upon the prevalence rates of the home locations of incoming students at the time of departure. The assumptions about the rate of transmission in this context are critical for estimating how large an outbreak any institution might experience and how effective any given control measure may be. Most models considered US institutions of varying sizes, which may only be partially relevant to institutions in other countries, and in most cases there were relatively simple assumptions about how people mix with each other. Lopman et al. state that moving more students to off-campus housing may make little difference to their conclusions because on-campus transmission is only moderately higher than off-campus3. However, this may not be the case in all institutions and modifying how students are placed in first year accommodation may have a large impact in some university settings.
Early outbreak detection is critical to controlling the spread of infection. Given the high proportion of asymptomatic cases likely to be observed in the student population, screening will probably be needed to identify increasing case numbers before there is a very widespread problem. Martin et al. found when there are nine or fewer detectable cases in the population, monthly testing of 100% of the campus community is required to detect an outbreak, unless at least 30% of infected people develop symptoms that lead them to present to a health service and be tested6. Models suggest meaningful control over outbreaks could be achieved through testing of symptomatic people and screening of the asymptomatic university community along with isolation of those who are infected (regardless of symptoms), contact tracing, and quarantining of those traced. However, strategies relying on this approach alone need to be very aggressive, with high levels of ascertainment in both testing and tracing to be effective. The assumed success rate of contact tracing in the models that considered this (58%, 75% and 100%) was generally higher than currently being achieved with the track and trace system in the UK (an average of 73% for the 3 weeks reported to 12 August8) and compliance with quarantine was 100%, which may well be unrealistic.
Cashore et al. included testing of students before travel to university, testing on arrival and testing of the university community every five days4. This would be logistically and financially challenging for many institutions. Models indicate that testing for symptomatic persons needs to be timely (less than a week between symptoms and positive test) and accompanied by effective isolation of the affected individual and subsequent contact tracing and quarantining of contacts in order to be really effective; much of the reduction in case numbers through more timely testing was due to the greater number of contacts reached in these scenarios3. The number of students in isolation and quarantine at any one time was considerable in some models, for example, Cashore et al. suggest planning for a peak quarantine capacity in excess of 700 in a population of 343104 (cumulative estimates over the term were not reported). We consider it quite likely that there will be more delays in the system than modelled in many of the papers included in this review, that is, symptomatic people may not get tested and isolate themselves immediately, and contact tracing with quarantine may not be as successful, so control may be more challenging than presented without additional mitigation. Combinations of interventions are likely to be required to effectively control infection5 and several options are available (see appendix). When evaluated, the number of cases was sensitive to the frequency of screening3,4 even with other mitigation measures in place; screening was required at least monthly for the whole university community to make a discernible impact. In models where other non-pharmaceutical interventions were not considered, screening was essential for COVID-19 control and findings were sensitive to the frequency of screening4. Screening was relatively less important when a package of interventions including partial online teaching to reduce contacts, use of face-coverings, and social/ physical distancing was used3,5. Gressman and Peck found moving classes with 30+ students online was a big driver in keeping infections low (moving the largest classes with 100+ students was particularly critical) and that face-masks were also moderately important (universal mask-wearing was assumed to reduce transmission by 50%). However, this model assumed an average of only one residential contact per student, and as residential and broad social contacts (dining halls/ restaurants, social events etc) were increased, these became more important in driving case numbers5. Limiting face-to-face contacts reduces infection spread. If students are placed into bubbles where only a small number of individuals mix with each other, low levels of interaction between bubbles are required to avoid large outbreaks and bubbles are of less benefit with increasing chance of infection from the wider community7 and may not be acceptable to students. Additionally, Gressman and Peck conclude large classes need to be held online to reduce the risk of a significant outbreak and it is not enough to break large face-to-face classes into smaller ones.5 Even small numbers of contacts between students in large group settings (dining halls/parties) may be sufficient to sustain an outbreak, even with other interventions in place5.
Discussion
All models suggest a significant potential for transmission of COVID-19 in universities. Early identification of cases is critical to inform effective infection control. Smaller outbreaks are easier to control, thus symptom monitoring and testing will be an important feature of outbreak control and management. Seasonal effects may be important; SARS-CoV-2 may be more transmissible in winter and an increase in background influenza-like illness may lead to increased testing and self-isolation, increasing the burden on university support systems. Modelling findings suggest that regular testing strategies (including testing asymptomatic individuals), isolation of infected individuals and contact tracing, together with use of face coverings and social distancing measures to reduce class sizes and face-to-face teaching could be effective in controlling transmission of COVID-19 in universities. One study found moving teaching online and wearing masks5 was more important than screening (monthly) to mitigate cases, however, the screening may be more valuable with higher frequencies of testing. In models much of the importance of testing was due to the impact of effective contact tracing and isolation/quarantining reducing onward transmission, but high numbers of contacts must be reached to be effective. Specificity of testing is important because even a low false positive rate could lead to a considerable number of uninfected students being quarantined5, especially when contact tracing is taken into consideration. A policy of reconfirming positives could avoid unnecessary quarantining, which could be important for managing the student population’s adherence to any guidance. Most of the studies identified that even with strong control measures in place, isolation of symptomatic cases and quarantine of contacts for 14 days may have a large impact on attendance of students (and staff) requiring online options for quarantining students even where in-person classes are taking place. Multiple interventions are likely to be needed to control infection spread within this setting and efforts to reduce transmission may be limited given the connectivity of students, staff, and the wider community, the varying capacity of institutions to employ such aggressive mitigation interventions, and likely varying compliance with any interventions imposed.
To our knowledge this is the first review of modelling studies considering COVID-19 infection in a university setting. The results are timely given universities are currently making decisions about if and how to re-open for the autumn term – there is no universal guidance for universities about how to do this and indeed university responses may need to be tailored to the institution given the variations in geography (campus or city based, in areas of high or low community infection) and student body.
The general messages here about the importance of quarantining and the need for multiple interventions for COVID-19 control in a university setting are consistent with work considering all settings9. We only found pre-prints relevant for inclusion, thus the papers here have not been through the critique of peer-review, which often results in authors being asked to provide additional information required for accurate interpretation of results or to amend analysis approaches. This presents a challenge for the review, however, bringing together the available evidence is critical now to help understand how universities may be affected by COVID-19 when they re-open in the next few months. We also only considered papers in English and from a select number of potential sources, therefore it is possible we have missed some models that could be relevant. Most papers had a US focus; the size and set-up of universities in other countries may be quite different to these institutions, influencing contact patterns and thus the spread of infection so policy makers should be mindful that mitigation strategies may perform differently in different settings. This rapid review has focused on mathematical models of SARS-CoV-2 infection and disease in a university setting, however, we are aware of other work that may be useful for informing policy decisions around how best to manage COVID-19 in this sector. Research assessing real contact patterns between college students in different countries has shown differences in the number and network of close proximity interactions, and that this influenced the spread of influenza10, thus considering local contact patterns is important for understanding potential risk and mitigation strategies. An increased risk of infectious diseases has been observed with mass gatherings11 and epidemiological studies have shown the capacity for infectious diseases to spread rapidly in a university context. For example, in a study of 2507 students in their first year at University of Nottingham in 1997, carriage of meningococci (a bacterium transmitted through droplets of respiratory and throat secretions from carriers) increased rapidly over the first week of term from 6.9% on day 1 to 23.1% on day 412. For students living in catered halls the average carriage rate increased from 13.9% during the first week of term in October, to 31.0% in November and 34.2% in December12. Mathematical models for influenza have captured the spread of this illness in students and outbreaks with high attack rates have been observed in universities13. Indeed ‘Freshers (freshman) flu’ - caused by a number of infections (not necessarily influenza) - is well-known for causing illness amongst students during the first few weeks at university.
Students come to university from across the country and world – given the rapidly changing epidemiological situation is it unclear how many infected students each institution may have at the start of term. Individuals are at risk of contracting COVID-19 in the community, so even institutions considering wholly online teaching will still need to consider how best to support students and staff who may become unwell. There are also challenges when students return home, both within and at the end of term, in terms of potentially spreading infection. International students are likely to face particular challenges with uncertainties around travel and their ability to return home at the end of term. Interventions to minimise contacts for a period at the start and end of term, including quarantining all students, could be effective at avoiding outbreaks. While the university campus environment can be controlled to some extent, the students live within a wider local community accessing public transport, shops, bars, and restaurants, and many have part-time jobs, thus the relationship between the university community and the general population is important for driving infection. Universities are considering intervention measures that go beyond government advice to augment national outbreak control systems to meet the particular challenges faced in this sector. All universities should be working on strategies that reduce face-to-face interactions in the university community and mitigate potential SARS-CoV-2 transmission where interactions must take place, through use of physical distancing, hand sanitising and with use of face coverings. Structural changes such as moving teaching online and influencing the dynamics of students mixing, particularly those living in first year university accommodation, are difficult for people to avoid. Testing was found to be important in some models to limit outbreaks, and multiple interventions, including regular university population testing has been recommended by Independent SAGE14. However, not all universities have access to in-house testing and even where they do this may be logistically challenging and expensive if deployed widely and regularly. Any testing provision will require close liaison with existing national systems (Local Authority and Public Health England). For those with capacity, batch or pooled testing may facilitate testing, but this raises issues about the sensitivity and specificity of such protocols. Individuals must also be willing to undergo testing – differential compliance between groups of lower and higher risk for transmission could have important implications for how effective testing may be6. Asymptomatic screening is assumed to happen at regular intervals across term when included in the models4, however, testing frequency may need to be adjusted in response to changing incidence/prevalence in order to be most effective in informing control strategies4. The national contact tracing system has been criticised for not being adequate, prompting some areas to set up their own15 - universities may need to consider whether additional tracing support may be helpful for their context. Antibody responses may not be long lasting16,17 (although other elements of the immune system may be longer lasting, though this is unknown) - so models assuming long term immunity may be over optimistic. Many potential interventions strategies are being considered and even if theoretically effective for controlling COVID-19 may not be feasible on the ground, thus clear information should be given to students, staff, and the wider community about what is realistic and the implications of this. There may be important negative consequences of some COVID-19 mitigation strategies in terms of the wellbeing of individuals
The situation with COVID-19 is rapidly evolving and there are a number of unanswered questions around the epidemiology. The infectiousness of asymptomatic persons is critical; it is believed to be lower than that of symptomatic individuals but largely unknown. This is particularly important in a university setting where it is currently assumed the vast majority of students are likely to be asymptomatic. Universities have a duty of care to students and to staff and the models identified in this review primarily consider the typical student and staff population. We know individuals from particular ethnic backgrounds and with underlying health conditions are at increased risk of poor outcomes from COVID-19 and this should be considered in university planning. Modelling and epidemiological evidence from the UK suggests that staff, such as catering and cleaning personnel, may be at particular risk. Other groups such as library or student administrative staff also have high rates of contact with students and may therefore be at elevated risk compared to other staff. The models have considered accommodation and teaching facilities, but not specifically risks of transmission in other settings such as public transport (including student buses), the use of other facilities such as sports centres or canteens, or social settings such as clubs and societies. Additionally, mathematical models could usefully focus on the incremental benefit of each additional intervention, particularly screening, to aid decision making.
Most of the models have considered campus universities in the US. In the UK there may be additional considerations. Differences for campus or city integrated universities in terms of mixing with general population are important for importation of infection to the university community, but also spill over of any university outbreaks into the community. However, limited work has been done in this area to date. There is much uncertainty in the model parameters and the structure of the models, particularly related to how people mix with each other. Student behaviour and mental health under quarantine is important, particularly as many students have small living quarters. Quarantining for students new to university may find isolation particularly lonely and significantly adversely affect mental health – if parents choose to pick up their students for quarantining at home this may drive an increase in community transmission leading to a worse outcome nationally. Additional risk to the broader population is due to education and accommodation breaks during reading weeks, Christmas and Easter holidays. Requiring students to remain at university during those times is unlikely, and should a substantial number of cases be present in the student population, a large-scale return to home is likely to disperse the infection across the country.
In summary, the modelling COVID-19 studies in a university setting that have been reported to date point to the essential requirement for mitigation measures tailored to the student context. Enhanced testing, including rapid testing of symptomatic individuals as well as screening of asymptomatic individuals should be given close consideration. Rapid contact tracing, effective student support for adherence to isolation and quarantine alongside a suite of other established mitigation measures all have a place in effective outbreak control. The situation is dynamic and although the large scale migration of over 2 million students to arrive at or return to their university locations in the Autumn is the immediate focus of attention, universities will need to closely monitor the developing situation and be able to adapt and rapidly respond to the likely occurrence of future outbreaks.
Data Availability
All data available in the manuscript.
Funding statement
There was no project specific funding for this work; the funders of the co-authors did not have a role in the conception, design or writing of the manuscript.
Author contributions statement
EBP conceived the study, HC and KT designed and ran the searches, and reviewed the papers for inclusion. HC, RDB, AT, EN, GH extracted the data from relevant papers. All authors wrote and critically reviewed the manuscript.
Competing interests
The authors declare no competing interests.
Acknowledgements
HC, KT and EBP acknowledge support from the NIHR Health Protection Research Unit in Behavioural Science and Evaluation at University of Bristol. The Health Protection Research Unit (HPRU) in Behavioural Science and Evaluation at University of Bristol is part of the National Institute for Health Research (NIHR) and a partnership between University of Bristol and Public Health England (PHE), in collaboration with the MRC Biostatistics Unit at University of Cambridge and University of the West of England. We are a multidisciplinary team undertaking applied research on the development and evaluation of interventions to protect the public’s health. Our aim is to support PHE in delivering its objectives and functions. Follow us on Twitter: @HPRU_BSE
RDB acknowledges support from the Elizabeth Blackwell Institute for Health Research, University of Bristol.
LD is supported by Medical Research Council (MRC) (MC/PC/19067) The Alan Turing Institute EPSRC EP/N510129/1.
Appendix
Potential interventions – highlighted rows show interventions considered in some form in mathematical models included in this review
