National Statistical

How common is long COVID? That depends on how you measure it

The ONS has worked at speed to help inform the UK’s response to the coronavirus pandemic. But producing reliable estimates as the pandemic unfolds is challenging, particularly for emerging phenomena such as “long COVID”, which is not yet fully understood or well defined. Daniel Ayoubkhani writes about the challenges of defining the condition as well as estimating the numbers of people suffering from it.

In today’s publication we used three different approaches to estimate the number of people experiencing long COVID symptoms. These show that between 3% and 12% of people infected with coronavirus have symptoms 12 weeks after the initial infection; or between 7% and 18% when considering only people who were symptomatic at the acute phase of infection. These results are based on self-reported data provided by a sample of over 20,000 Coronavirus Infection Survey (CIS) participants.

In one sense, the results are reassuring: the majority of people infected with coronavirus (88% to 97%) don’t experience symptoms beyond the first 12 weeks, and some of those who do will start to feel better over time.  But for the minority of people who do go on to experience long-term symptoms, the effects can be debilitating, and we should remember that the absolute numbers could be considerable: our most recent population-level estimates suggest that 643,000 people in private households in the UK could be experiencing activity-limiting long COVID symptoms.

Aside from the results themselves, one of the main conclusions from today’s analysis is that the estimated prevalence of long COVID is sensitive to the way that it’s measured and who is included in the study sample. This is also reflected in the wide range of prevalence estimates coming out of a rapidly increasing number of long COVID studies internationally, each with its own study population, data collection and methodology.

This diversity in approaches and estimates may stem from the lack of a consensus in defining long COVID. Clinical case definitions of ongoing symptomatic COVID-19 and post-COVID-19 syndrome do exist in the UK, and while we’ve tried to align with this guideline as far as possible for our own analysis, it wouldn’t be feasible or desirable to precisely use a clinical definition for statistical research.

 

A better understanding

Today’s publication is the first time we’ve published a range of long COVID prevalence estimates using different approaches. However, in April 2021 we published results using one of the three approaches, based on tracking continuous symptoms. One of the more striking findings from the latest release is the revision to the 12-week prevalence estimate using this approach, from 14% back in April to 3% now. This fall in the prevalence estimate is largely because we know more about participants’ long-term symptoms today than we did then.

In order to get accurate estimates, we need a sufficiently long time period after someone has been infected to be able to monitor their symptoms and observe when their symptoms came to an end. When we published our estimates in April, the cut-off date for assessing if people still had symptoms was 6 March 2021, just as the UK was emerging from the “second wave” of infections. This meant that lots of people in our sample had recently been infected and had only a limited period over which to assess average symptom duration, with most study participants having less than 12 weeks of follow-up data. Over 50% of participants were thought to have ongoing symptoms at that point. However, now that we have data that extends well beyond 12 weeks after infection for most participants in the sample, we can see that symptoms have since stopped for the majority of those symptomatic on 6 March.

We’ve also taken on feedback from clinical and statistical experts and made some methodological updates since we published our results in April, most notably how we measure when a participant’s symptoms came to an end. One consequence of these changes is that some participants who we classified as having ongoing symptoms in our April analysis have been reclassified as having experienced the end of their symptoms before 6 March, which has further reduced the prevalence estimate.

 

Experimental statistics

Throughout the COVID-19 pandemic, the ONS has contributed to the public health response by publishing new data, including on long COVID, in a timely manner to inform urgent policies and interventions. We’ve been clear throughout all of our releases on long COVID prevalence that the statistics are experimental. This means that although the estimates are produced to the same high standards and undergo the same rigorous quality assurance processes as other official statistics, they remain under development and are not yet fully refined.

In fact, one of the main reasons we publish experimental statistics is to enable our methods to be scrutinised and improved. It’s therefore inevitable that the estimates will continue to evolve over time as we iteratively develop our knowledge and methodology. We would thus advise users to interpret today’s release, and others about long COVID, with caution.

We will continue to refine our estimates to take account of the scientific evidence as it evolves, collaborating with partners across academia, clinical practice, patient representatives and government to develop and implement our research plans. For more information or to share your views on our plans and methods, please email health.data@ons.gov.uk quoting “long COVID” in the subject line.

Daniel Ayoubkhani is a Principal Statistician in the Health Analysis and Life Events Division at the ONS.

Exit mobile version