We have previously described how OpenSAFELY has been used to generate evidence for research and NHS Service Analytics. In this blog post Millie and Laurie share how some of the outputs from the work on COVID-19 therapeutics have been used to inform policy and guidelines in the NHS and beyond. We hope this will be an interesting read for everyone. We think it will be of particular interest to health policy makers and guideline developers, such as NICE, as we describe practical examples of how OpenSAFELY can generate “real world evidence” in near real time to inform the roll-out of new treatments.

Coverage and uptake of COVID-19 treatments

On 11th December 2021, new COVID-19 Medicine Delivery Units (CMDUs) were launched across England, offering COVID-19 therapeutics (antiviral drugs and neutralising monoclonal antibodies (nMABs)) as treatment to patients with COVID-19 in community settings.

To assess the coverage of these new treatments, using the OpenSAFELY platform we identified eligible patients and investigated how coverage varied over time between key clinical, regional, and population subgroups, and whether any treatments given could have been inconsistent with guidance. This allowed us to identify detailed characteristics of those who were (and were not) getting treatments and revealed apparent lower treatment coverage among certain groups in the beginning of the roll out, in particular: different NHS regions, ethnic groups, people aged ≥80 years, those living in socioeconomically deprived areas, and care home residents. Full methods and results are described in our paper published in BMJ Medicine and all of the underpinning analytic code can be found in the corresponding OpenSAFELY repository.

To assist with ongoing monitoring and targeted initiatives at the outset of the roll-out, using our vaccine coverage framework, we rapidly developed a near real-time COVID-19 therapeutics dashboard. Since the first release of this report in March 2022, we have updated the results seven times and added additional breakdowns of results i.e. coverage by Sustainability and Transformation Plan (used as a proxy for CMDU as it has almost a 1:1 mapping).

This coverage work is a great demonstration of how OpenSAFELY can deliver detailed coverage reports in fine-grained clinical and demographic risk groups, with publicly auditable methods, using linked but pseudonymised patient-level NHS data in a highly secure Trusted Research Environment. It has not only helped to monitor gaps in coverage, but has also been useful during the development of the study designs for the efficacy and safety analyses of these new treatments. In addition it has been used in the front line to make operational decisions around treatment delivery (more about this below).

Comparative effectiveness work

With the frequent change of predominant circulating variants of COVID-19, the effectiveness of COVID-19 treatments soon became uncertain, given that randomised controlled trials were conducted against previous variants. Determining the current clinical and cost effectiveness of these treatments therefore became critical when NICE needed to determine which drugs to recommend for routine clinical care.

The design of the OpenSAFELY platform allowed for the analytic code underpinning the COVID-19 therapeutics coverage work to be quickly re-used to facilitate the delivery of a number of papers assessing the effectiveness of COVID-19 medicines using “real-world” data.

The first of these papers (published in BMJ) compared the effectiveness of sotrovimab (a nMAb) versus molnupiravir (an antiviral) for preventing severe COVID-19 outcomes in non-hospitalised high-risk adult patients. It showed that those who received sotrovimab were at lower risk of severe COVID-19 outcomes than those receiving molnupiravir.

Given the approval of new medications in early 2022 and the subsequent changes in treatment guidelines, we went on to compare the effectiveness of Paxlovid and sotrovimab in preventing severe outcomes from COVID-19, as well as comparing Paxlovid and molnupiravir. We showed that treatment with Paxlovid was associated with a similar rate of severe outcomes from COVID-19 infection as with sotrovimab. Full methods and results are described in our preprint paper.

Given that patients on kidney replacement therapy (KRT) are at very high risk of severe outcomes from COVID-19 (and the fact that little is known on the effectiveness of COVID-19 treatments in this population due to the limited inclusion of these patients in clinical trials), we conducted a study comparing the effectiveness of sotrovimab against molnupiravir in preventing severe COVID-19 outcomes specifically within this vulnerable population. In our study we showed that in routine care of non-hospitalised patients with COVID-19 on kidney replacement therapy, identified through our unique linkage with the UK renal registry, those who received sotrovimab had substantially lower rates of severe COVID-19 outcomes than those receiving molnupiravir.

As part of our work examining real-world effectiveness of treatments for COVID-19, we have also looked at risk of severe outcomes comparing people who receive treatment to those who were untreated. Given the significant biases that these types of studies are prone to, this has not been an easy piece of work! To do this we have used the OpenSAFELY platform to emulate target trials to estimate the effectiveness of sotrovimab or molnupiravir, versus no treatment. In our study we found that compared with no treatment, sotrovimab was associated with lower rates of adverse outcomes after COVID-19 in the BA.1 period, but there was weaker evidence of benefit in the BA.2 period. Molnupiravir was not associated with reductions in event rates in either period.


Both streams (coverage and effectiveness) of our COVID-19 therapeutics work have been used to make operational decisions around treatment delivery. For example, the NHS England commissioning framework has a whole section on health inequalities based on our findings and advises local NHS areas to take specific actions to address these disparities. The table below shows where our evidence on COVID-19 therapeutics is influencing guidance and demonstrates clearly how OpenSAFELY can be used to develop robust real world evidence to influence decision making both in England and globally.

Table 1 Impact of OpenSAFELY COVID-19 therapeutics work. Note, this is not an exhaustive list and it is possible that our work has been used elsewhere.

Therapeutics for people with COVID-19 guidance (NICE)
Commissioning Framework: COVID-19 Therapeutics for Non-Hospitalised Patients (NHS England]
British Columbia COVID Therapeutics Committee (CTC) (Canada)
STAKOB (Germany)
Japanese Association for Infectious Diseases (Japan)
COVARS (France)

COVID-19 therapeutics data

Since the start of the roll-out, all data on COVID-19 therapeutics has been collated centrally by NHS England. But as the way treatments are delivered is reorganised within each Integrated care board (ICB), this will no longer happen. Going forward, two changes to data availability are expected. Firstly oral treatments will be available on GP “FP10” prescription. This will mean we will be able to query these prescriptions via the standard data available in OpenSAFELY. However, we have been told that in some areas treatment decisions, including oral treatments, will continue to be made by the CMDUs. Secondly, sotrovimab is an infusion and, in most areas, will continue to be administered in more specialist units. It is currently unclear what will happen in terms of the data regarding which treatment has been prescribed but it is most likely that specialist units will have to send a data return to the ICB via their “high cost drug” system.

Despite these changes, data regarding drug prescribing still exists, but will not be available for analysis, despite all the benefits we have demonstrated. Information will most likely be captured in local “high cost drug” systems (sometimes colloquially known as “PbR excluded” or “Blueteq data”), which the NHS collates in every ICB in the country for payment purposes. At the height of the COVID-19 pandemic we worked with the “Commissioning Support Units” who do this work on behalf of ICBs and arranged for a one-off ‘stitching together’ of all high cost drugs datasets into a single comprehensive dataset. We used this dataset to study outcomes from treatment with biologic drugs and their relationship to COVID-19 outcomes. If this ‘high-cost drug’ dataset was routinely made available, researchers could continue to use this data for important health service analyses, including projects such as our innovative work on COVID-19 therapeutics. We have already written a detailed description of the high cost drugs dataset and outlined the rationale for why we think the NHS should make available the data it already collects to accredited researchers via Data Access Request Services.

Where next?

Due to these anticipated changes to the data flow as a result of NHS reorganising how COVID-19 services are delivered for patients, the shape and scale of the real world evidence we can generate using COVID-19 therapeutics data in the future will be impacted. For example, if these changes are implemented as expected, for future studies involving oral treatments we may not be able to identify all patients receiving such treatments, leading to reduced sample size and potential selection bias. It would also mean that we would not be able to conduct any studies on sotrovimab. However, we still have ideas on how we would be able to continue working on the same sorts of studies but ultimately we hope that the ‘high-cost drug’ dataset will be routinely made available.