Skip to main content

COVID-19 Population Risk Assessment

In February 2021, we used the University of Oxford’s QCovid® risk prediction model to identify additional people to be added to the Shielded Patient List (SPL) as a precautionary measure. 

Shielding notice

Following the announcement regarding the end of national shielding in England in September 2021, NHS Digital has managed the closure of the Shielded Patient List. This closed on 30 June 2022. 


We used QCovid® to develop the COVID-19 Population Risk Assessment, which combined a number of factors such as age, sex registered at birth, ethnicity, body mass index (BMI) and specific health conditions and treatments to estimate the risk of a person catching coronavirus and becoming seriously unwell. 

The COVID-19 Population Risk Assessment took place in February 2021 using the original version of QCovid®.

We used patient data held centrally to identify people who might be at high risk and generated risk assessment results for these people. People whose results are above the agreed threshold for high risk (clinically extremely vulnerable) of severe illness from coronavirus were added to the Shielded Patient List (SPL) in England.

Our use of QCovid® in the COVID-19 Population Risk Assessment using patient data held centrally has been registered by NHS Digital as a Class 1 medical device with the Medicines and Healthcare products Regulatory Agency (MHRA).

Read more about QCovid®, how it works and how the NHS is using it.


How we identified people as being potentially at high risk

We worked with the University of Oxford to identify the pieces of information about people and their health that are needed by the risk assessment model to generate as accurate risk assessment results as possible.

We identified where this information could be found in existing datasets held by NHS Digital. We then gathered the specific pieces of information from the records of people who could potentially be considered high risk (clinically extremely vulnerable). As the national information and technology partner to the health and social care system in England, we have legal permission to securely collect and analyse this data.

The relevant coded data was found to be recorded in 7 national datasets:

We then ran this data securely through QCovid® to generate risk assessment results. The people whose risk assessment results were above the agreed threshold for high risk were then added to the SPL as a precautionary measure. 

Read more about how QCovid® works and the risk factors used to estimate risk assessment results.

More detailed information about the datasets used, the information we used from those datasets and processing rules is available. 


Health conditions and treatments

We used data about the following conditions and treatments to generate population risk assessments.

Cardiovascular diseases
Respiratory diseases and treatment

Asthma

Taking anti-leukotriene or long acting beta2-agonists (LABA)

Chronic obstructive pulmonary disease (COPD)

Cystic fibrosisbronchiectasis or alveolitis

Pulmonary hypertension or pulmonary fibrosis

Metabolic (diabetes), renal and liver conditions

Gestational diabetes is high blood sugar (glucose) that develops during pregnancy and can resolve after giving birth. People who have had gestational diabetes are at increased risk of developing type 2 diabetes or having undiagnosed diabetes.  

Some people with past gestational diabetes have been identified in combination with other factors by the QCovid® model as being potentially at high risk from coronavirus.

People with a history of gestational diabetes who have been advised to shield, but feel this is no longer relevant, are able to contact their GP for further advice. GPs can evaluate a patient’s risk with the most up to date information using the clinical tool, and remove them from the SPL if necessary.

More detailed information about gestational diabetes.

Neurological and psychiatric conditions
Immune and haematological conditions

Rheumatoid arthritis, SLE (systemic lupus erythematosus, more commonly known as lupus) or a seronegative arthritis such as ankylosing spondylitis

Sickle cell disease, HIV or severe combined immunodeficiency

Thrombosis or pulmonary embolus

Immunosuppressants, cancer conditions and treatments

Lung or oral cancer (such as laryngeal (larynx)nasopharyngeal or mouth cancer)

Solid organ transplant (such as lungliverstomachpancreasspleenheart or thymus)

Cancer of the blood or bone marrow (such as leukaemia, myelodysplastic syndromes, lymphoma or myeloma)

Chemotherapy in the last 12 months

Oral prednisolone (prescribed by a clinician 4 or more times in the last 6 months)

Immunosuppressants (prescribed by a clinician 4 or more times in the last 6 months)


Gestational diabetes

Some people with previous gestational diabetes were identified by the QCovid® model as being at high risk. This will be appropriate for many as the model performs an individual assessment based on a wide range of risk factors, and also considers a person's risk in comparison to others of the same age and sex.  

However, because the risk assessment is based on routinely coded data from multiple systems, some people may have been identified as having diabetes when in fact they had gestational diabetes. Others may have incomplete data, in which case the risk assessment may have defaulted, on a precautionary basis, to a higher level of risk for that category and this may influence the overall assessment results. 

If a person was identified as being at high risk due to previous gestational diabetes only (no other significant conditions) and both:

  • had a Body Mass Index (BMI) between 16 and 41
  • had a HbA1c (a test to check average blood glucose/sugar levels) since delivery and within the previous 12 months which is normal or in the pre-diabetes or non-diabetic hyperglycaemia range

they did not need to be included in the SPL and could have been removed if they requested this. They would have still been called for an earlier vaccine. Shielding was advisory and people who received a letter could choose whether or not to follow shielding advice. Clinicians were able to use their clinical judgement and discuss any decision to remove a patient from the SPL.

Where a person with previous gestational diabetes, and a normal HbA1c in the previous 12 months, has a BMI below 16 or above 41, the COVID-19 Clinical Risk Assessment Tool was available to generate the patient’s risk assessment results. These results were used to inform an individual assessment of risk to determine, in consultation with the patient, whether they should remain on the SPL or be removed.

Importantly, people with previous gestational diabetes have an increased risk of developing Type 2 diabetes (which would put them at higher risk from coronavirus) and they should continue to undergo the recommended annual checks for this.

The Royal College of General Practitioners (RCGP) developed a flow chart to support GPs when considering risk for patients with a history of gestational diabetes.


Who was risk assessed

Risk assessment results were only generated for:

  • people who could potentially meet the threshold for being considered at high risk (clinically extremely vulnerable)
  • people who had not previously been identified by existing SPL processes

This means we did not generate risk assessment results for some people, including:

  • people who were already on the SPL
  • people who were previously on the SPL, but had been removed by a clinician involved in their care
  • people whose combined risk factors could not possibly meet the agreed threshold
  • people who had already been assessed individually by a clinician and identified as low or medium risk
  • people under the age of 19, or over 100, because QCovid® is not designed for use on these groups
  • people coded with Down’s syndrome in their record who had not already been identified through the national SPL. This is due to Down’s syndrome already being included in the national SPL ruleset and a known coding issue.

Agreed threshold for adding people to the Shielded Patient List

England’s Chief Medical Officer (CMO), in consultation with senior clinicians, looked at the results of the QCovid® research conducted by the University of Oxford to work out what risk threshold should be used to decide if someone may be high risk and added to the SPL as a precautionary measure. 

The research showed that most people included in the study who died from coronavirus would have had risk assessment results that placed them in approximately the top 2% of the population in England that are at the highest risk. For the combined risk of catching and dying of coronavirus, most results were higher than or equal to:

  • an absolute risk of 0.5% (or 5 in 1,000)
  • a relative risk of 10 (or 10 times the baseline risk)

The CMO and senior clinicians decided that these thresholds should be used to help protect people who may be high risk.

Absolute risk is the overall risk, based on what happened to other people with the same characteristics and risk factors who caught coronavirus and went to hospital or died as a result.

 

Relative risk is the level of risk compared to a person who is the same age and sex registered at birth, but without any other risk factors.


Missing or unknown data

If certain information was missing from a record held by NHS Digital or had been recorded as unknown, then default values were used by the COVID-19 Population Risk Assessment as a substitute.

We took a precautionary approach to how we handled information which might be missing from records. This was to make sure that we did not underestimate people’s risk and exclude them from the high risk (clinically extremely vulnerable) group who were added to the SPL.

Where data was missing, we used some default values that have higher than average risk associated with them, according to QCovid®. This means that we were likely to overestimate the level of risk for some people with information missing from their records and therefore identified more people as potentially high risk (clinically extremely vulnerable). 

As a result, some people may have been advised to shield based on these default values. Based on advice from clinical advisors, the Department of Health and Social Care determined that this precautionary approach was clinically the most appropriate. 

The patient’s GP record indicated that they had been added to the SPL using the COVID-19 Population Risk Assessment, and whether a default value was applied. Read more about how this is shown in GP IT systems.

Default values

Body Mass Index (BMI)

If a person’s BMI was not available, a BMI of 31 was used by default. This was the midpoint BMI in the population data that was available when we analysed it.

If BMI was more than 47, then a BMI of 47 was used to generate the results. If BMI was less than 15, then a BMI of 15 was used to generate the results. This was because almost all (99.9%) BMIs in the UK are between 15 and 47. BMIs calculated as being lower than 15 or higher than 47 were likely to be a mistake which would affect the accuracy of the result.

Sex registered at birth

If a person’s sex registered at birth was not available from GP data, and gender was not available from the Patient Demographic Service (PDS), a value equal to the sex with the highest risk was used by default. This is in line with the precautionary approach to overestimate rather than underestimate risk.

Ethnicity

If a person’s ethnicity was not available, a value equal to the ethnicity with the highest risk was used by default. This is in line with the precautionary approach to overestimate rather than underestimate risk.

Postcode

If a person’s postcode (used to identify a Townsend deprivation score) was not available, the average UK score for the Townsend deprivation index was used. This is in line with the defaults used in QCovid®.


Validating additions to the SPL generated by the population risk assessment

To act as quickly as possible and reduce GP workload, patients identified as potentially high risk were added directly to the SPL as a precautionary measure. Clinicians were able to follow existing processes to review, add and remove patients from the SPL where this was appropriate at any time.

Clinicians were advised to continue to review individual patients on an ongoing basis, according to clinical judgement and patient request. Guidance about how to identify this cohort of patients in IT systems was made available to general practice and hospital trusts.

Clinicians were able to use the COVID-19 Clinical Risk Assessment Tool to review additions to the SPL from the COVID-19 Population Risk Assessment.

At the time of patients being added to the SPL, the tool used the same QCovid® model used in the COVID-19 Population Risk Assessment and could be used to generate results for individual patients. The tool has since been updated to use the most up-to-date versions of the QCovid® model. Because the population risk assessment used data from a number of national datasets and may have included different default values, there may have been some variation between results. Detailed guidance for clinicians is available.


Patients not added to the SPL

GP practices were able to access details of their patients who were risk assessed, but whose risk assessment results did not meet the agreed threshold to be added to the SPL. Some of these patients may have been close to the threshold for addition to the SPL and GPs may have wished to review these individually. The information reflected whether default values were used.  

Clinicians were able to use the COVID-19 Clinical Risk Assessment Tool to review these patients, alongside their clinical judgement, as they may have had access to more up-to-date information.

This list did not include every patient registered at the practice that may have been close to the threshold because we did not risk assess all patients. GPs were therefore advised to still consider using the risk assessment tool on an individual basis for patients they believed may have been at risk.


Transparency notice

This explains how personal information may be used by NHS Digital in relation to the COVID-19 Population Risk Assessment.

View the COVID-19 Population Risk Assessment transparency notice


Summary statistics - high risk population

Last edited: 30 June 2022 2:53 pm