Controlling the spread of Covid-19 in vulnerable settings

This research used data made available via the Office for National Statistics (ONS) Secure Research Service, which is being expanded and improved with ADR UK funding.

Author: Dr Becky Arnold (University of Keele)

Date: March 2022

Research summary

A researcher from the University of Keele has developed a flexible simulation code to model the spread of Covid-19 within vulnerable settings. This data-led model has been used to inform the decision making of policy departments by providing evidence and advice about the likely outcomes of potential testing strategies within these settings.

This project was conducted for the UK Health Security Agency. The spread of Covid-19 is of particular concern in vulnerable settings, such as care homes, hospitals, prisons, and schools. “What is the most effective testing strategy to utilise?” is therefore a matter of significant concern. Within the context of public institutions with limited resources, the cost-effectiveness of different strategies is also highly pertinent.

These are difficult questions to answer due to the complex, varied nature of these settings and the number of different components that make up a testing strategy. These components include lateral flow device & PCR testing frequency, symptomatic testing policy, outbreak policy, and isolation or release policy for infected individuals. 

Hear Dr Becky Arnold discuss this research as part of the ONS Research Excellence series on Thursday 15 June, 11:00-12:00. Register now.

Data used

A large number of data sources were used, such as:

• Test and Trace data
• Vaccination data (for the general population, care home residents and workers, and prison residents and workers)
• Mortality data (for the general population, care home residents, and people in prison)
• Publicly available data from government reports, including:
  • this report on the demographics of care workers
  • information from University of Oxford and UK Government on the reliability of lateral flow and PCR tests
  • hospital and intensive care occupancy due to Covid-19 data
  • Census data about residents in communal settings such as care homes and prisons
  • healthcare data for prisons
  • demographics of the prison population.

Data from June 2020 onwards was used for some parts of the project. The majority of the data used was collected between December 2021 and April 2022.

Methods used

Settings such as care homes, schools, prisons, and hospitals are very different environments, but also share commonalities. They all function on the paradigm of staff and “customers” (individuals that receive care or are serviced by the staff such as care home residents, prisoners, patients and students), and every day a testing regime is applied, test results are received, people are moved in or out of isolation, people interact, and infection is spread. With this in mind, the researcher designed a flexible, data-driven, agent-based model which can be tailored to simulate different vulnerable settings.

The creation of this model can be broken down into five steps:

Step 1: Created simulated people, with properties such as: role (staff, prisoner, student etc.) age, sex and vaccine doses received. These properties were drawn from distributions based on real data which included a mix of secured and public datasets. The probability of general admission hospitalisation, intensive care hospitalisation, and death was then assessed for each simulated person based on their properties. These probabilities were based on careful analysis of both secured datasets such as Test and Trace and publicly available data.

Step 2: Defined how people interact within the setting. An N by N matrix (a type of mathematical table where N is the number of people being simulated) was created which defines how likely each individual is to interact with every other individual. Staff have some daily probability of interacting with each other and with customers, and customers can also interact with each other. These probabilities are drawn from a combination of data, literature reviews, conversations with departmental officials (such as at the Ministry of Justice and Department of Health and Social Care), and conversations with individuals who worked in these institutions. The matrix also encapsulates the specifics of different settings. For example, in prison, pairs of customers will be cellmates and have a 100% probability of interaction every day. This innovative design means that by modifying this matrix, the model can be transformed to simulate wildly different settings.

Step 3: Defined the testing strategy to be simulated. This included the frequency of lateral flow & PCR testing for customers and staff, how testing of symptomatic and asymptomatic individuals is applied, outbreak testing policy, and isolation or release policies for confirmed cases.

Step 4: Simulated the spread of infection in the institution for the desired length of time (usually one year). Infections were seeded in and spread based on the interaction matrix, then the testing regime was applied. The researcher then kept a record of which individuals were infected in the simulation and recorded the costs incurred by the testing regime.

Step 5: Based on who was infected, the researcher estimated the number of general or intensive care hospitalisations and deaths. By using these estimations and the logged costs of the testing regimes, the researcher was able to perform a health economic assessment of the testing strategy in different vulnerable settings.

Research findings

This research project found that when resources are highly limited, asymptomatic testing is more beneficial than outbreak testing. Symptomatic testing is critical in all circumstances and requires a relatively low fraction of the overall budget.

• Asymptomatic testing is the normal testing regime individuals conduct even if they are not displaying symptoms. Its purpose is to identify cases before infection can spread.
• Symptomatic testing is the testing of individuals who have developed symptoms. Its purpose is to confirm cases so they can be isolated.
• Outbreak testing is a period (typically seven days) of enhanced testing in an institution undergoing a confirmed outbreak. Its purpose is to identify all cases and contain the outbreak.

Outbreak testing requires high frequency (in some cases daily) testing of all individuals in the institution. It also requires PCR tests, which are 10 times as expensive as lateral flow devices. Because of the commonly high prevalence of Covid-19 in the UK, regular outbreaks are inevitable. As a result, outbreak testing is extremely expensive and demands a large fraction of the overall vulnerable setting’s testing budget. This project found that placing financial resources towards asymptomatic testing was significantly more effective in reducing loss of life and hospitalisations than putting them towards outbreak testing. This is because asymptomatic testing can catch “patient zero” and therefore identify potential outbreaks before they have a chance to spread, greatly reducing the number of overall infections.

The project also found that when resources are highly limited, testing staff is more beneficial than testing customers in communal settings such as care homes and prisons. This is because staff return home every day, but customers’ contact with the outside world is greatly reduced (limited to visitors and hospital or court appointments). The testing regime for staff and customers does not have to be the same (and in fact is almost always different). For example, the testing regime for care home staff may be daily lateral flow tests, and for customers it may be once monthly PCRs. This project compared different intensities of testing for staff and customers, finding that high staff testing and low customer testing produced by far the best outcomes. This is because staff, due to their greater contact with the outside world, are more likely to be “patient zero”.

Research impact

The development of this adaptable, comprehensively researched, and peer-reviewed model enabled the UK Health Security Agency (UKHSA) to provide robust advice about testing strategies in vulnerable settings. It has also enhanced the overall understanding of the UKHSA of the differences between settings, and the unique considerations for managing the spread of Covid-19 within them.

This model enables effective deployment of limited resources, ensuring their benefits are maximised. The model also allows the UKHSA to plan responses to different scenarios. The optimum testing regimes for potential future variants that are more/less deadly or transmissible can now be studied in advance, with minor modifications to the model’s input parameters. Optimal testing regimes under conditions of low, medium or high Covid-19 prevalence in the UK can also now be evaluated.

The data-driven health economic assessment of the outcomes of different testing strategies is a powerful tool for UKHSA. It enables them to provide evidence to HM Treasury supporting funding requests for lifesaving testing regimes.

The public policy on testing in vulnerable settings was based in part on evidence provided by this project. 

Research outputs

Publications and reports

Due to the sensitive nature of this project, the model and its precise outputs cannot be made fully public.

Presentations and awards

• Cross-government Analysis, ONS Research Excellence Awards 2022
• Data Science Seminar, June 2022

About the ONS Secure Research Service

The ONS Secure Research Service is an accredited trusted research environment, using the Five Safes Framework to provide secure access to de-identified, unpublished data. If you use ONS SRS data and would like to discuss writing a future case study with us, please ensure you have reported your outputs here: Outputs Reporting Form