The Ongoing Risk Of Long COVID

What are the rates of ongoing COVID-19 symptoms in 2023 and beyond?

Most stock photos for “Long COVID” are just people in masks, which I suppose is apt. Photo by Adrian Balasoiu on Unsplash

It’s 2023, and the individual risk from COVID-19 has never been lower. We’ve developed new treatments, brilliant vaccines, and have very strong immunity to the disease built up over the years since it first emerged. While COVID-19 is undoubtedly a public health issue - it probably always will be - the risks of bad things like hospitalization and death are to the individual the lowest that they’ve ever been.

That’s not to say COVID-19 has gone away. COVID-19, as with influenza, tuberculosis, measles, and most other public health threats, is still a serious problem. The current yearly death rate is at least double that of a bad influenza year, which is likely because while COVID-19 is similarly lethal to the flu in 2023 with vaccines, hybrid immunity, and improved treatments, it is still far more infectious. However, in general, it’s fair to say that COVID-19 has moved from a public health emergency to a public health problem.

But there’s still a really important issue that people raise whenever you discuss the long-term view of COVID-19 - what about Long COVID? We know that many people suffer issues after their acute infections, and there are still people having life-shattering problems even years after their brush with the coronavirus. People online have been arguing for some time that Long COVID represents an ongoing public health crisis that will eventually impact everyone in the world.

This sounds scary, but also seems quite unlikely. Long COVID is undoubtedly a serious problem, but the magnitude of the issue is, if anything, diminishing over time. Let’s look at what the evidence says about of Long COVID now, in 2023.

Prevalence vs Incidence

For this discussion, there are two important epidemiological terms you’ll have to understand. Simple stuff, but useful.

Pictured: Simple but useful, just like epidemiology. Photo by Christin Hume on Unsplash

The first term is incidence. The incidence of a disease is the rate of new cases of that disease within the set period of time.

The second term is prevalence. Prevalence is the rate of existing cases of a disease within a period of time.

Both of these terms are often described as the rate of a disease colloquially, but they are of course very different. For example, the prevalence of diabetes has been rising for decades across the world, but surprisingly the incidence has been stable in many places for some time. This is because, while diabetes is not necessarily being diagnosed more often, it is far more survivable than it was in the 90s. More people living longer with diabetes means that even if the incidence is the same - the number of people getting diagnosed each year doesn’t differ - the prevalence rises as those who are diagnosed stay alive for longer.

Prevalence Of Long COVID

So. How many people across large populations are experiencing long-term issues that were caused by COVID-19?

Many of the scariest headlines come from studies that estimate incredibly high rates of Long COVID prevalence. But if you look more closely than the headlines, the question of Long COVID prevalence turns out to be difficult to answer, for several reasons. I’ve written about these issues before. The main problems are:

1. Definitions

2. Causality

Long COVID is a very poorly-defined problem, which has caused most clinical definitions to be extremely broad. For example, here is the CDC’s question-based definition of Long COVID, from their Household Pulse survey:

Did you have any symptoms lasting 3 months or longer that you did not have prior to having coronavirus or COVID-19?

Long term symptoms may include: Tiredness or fatigue, difficulty thinking, concentrating, forgetfulness, or memory problems (sometimes referred to as “brain fog”), difficulty breathing or shortness of breath, joint or muscle pain, fast-beating or pounding heart (also known as heart palpitations), chest pain, dizziness on standing, menstrual changes, changes to taste/smell, or inability to exercise.

This includes pretty much any health problem imaginable. There’s no marker for severity, or even for specific symptoms, and it’s based entirely on self-report which widens the criteria further. If you ask people whether they’ve had any of a wide range of symptoms, many will almost by definition answer yes. Fatigue is one of the main symptoms of Long COVID, but it’s also a very common symptom that people were reporting at a population level long before 2020.

The issue of causality is also challenging. We know that some portion of the population will have various symptoms throughout the year. Some of these have easy to identify reasons behind them, but some don’t. This is why we need control groups to understand the impact of Long COVID - it’s not that people are lying about having symptoms, it’s that we don’t necessarily know whether COVID-19 is the cause of those long-term problems.

Did x cause y, or was it just an association? Photo by Tamara Gak on Unsplash

If you look at studies with tight definitions of Long COVID that use control groups, the prevalence is almost always drastically lower than the proportions suggested by the scarier headlines. For example, here’s a recent study from Australia which suggested that there were no differences in long-term health problems between people who tested positive for influenza and those who tested positive for COVID-19. Of course, this study has a number of biases that make it hard to apply more broadly, but it does at least indicate that people who were sick enough to get a test for flu were not dissimilar in terms of their long-lasting symptoms to those who were tested for COVID-19 in 2022.

Unfortunately, there are currently no ongoing population-wide estimates of Long COVID that include specific definitions and control groups. That means that any estimates that we do have are biased, because we can’t be sure how many of the symptoms that these people report were caused by COVID-19. This is not to say that people aren’t sick, just that we may not know why they are sick.

That being said, there are two fairly good estimates of the rate of people reporting Long COVID over time - one in the UK, the other in the US. The UK study, conducted by the Office for National Statistics (ONS), is probably the most robust estimate of the prevalence of Long COVID over time. The US CDC study, called the Household Pulse, is quite good, but as the CDC acknowledges in the technical notes has a number of flaws that may lead to the estimate being a bit overinflated.

If we look at the most recent estimate from these two studies, we get a population rate of 5.5% from the CDC, and 2.9% from the ONS of people reporting long-term symptoms that may be associated with a past COVID-19 infection. The two surveys estimate that 1.4% and 0.6% of people, respectively, have long-term symptoms that limit their daily activities significantly, while the remaining 75-80% of people with long-term symptoms report “little” to “no” impact on their daily life.

These figures are dropping over time. At the peak prevalence, the Household Pulse survey estimated that 7.5% of the country reported Long COVID, while the ONS estimated that 3.1% of the UK reported the same. The decreases may not seem huge - 7.5% to 5.5%, and 3.1% to 2.9% - but together they represent millions people fewer who report having long-term symptoms associated with a past infection.

Long COVID Incidence

The prevalence of reported Long COVID is relatively low, and falling over time. Even the prevalence of people reporting highly activity-limiting symptoms has fallen, although by a slightly smaller proportion.

But that’s not the question that people are most interested in at this point, in 2023. What people tend to be most concerned about is what the risk is for them if they catch COVID-19. In other words, the incidence.

There are a number of different answers to this question. We could, for example, look at another recent Australian study that has gone viral. This preprint looked at people in Western Australia who tested positive to COVID-19, and found that in this population 18.2% of the people who took part in a follow-up survey reported some rate of long-lasting symptoms 3 months later. People have seized on this number as a reasonable representation of how many of those infected in a highly-vaccinated population are likely to get Long COVID after their acute period is over.

Unfortunately, this paper doesn’t really give us a population estimate of Long COVID. By relying on only people who tested positive to COVID-19, rather than estimating the total number of infections, the survey will by definition be biased towards more severe cases. Similarly, the fact that only about 1 in 7 of the people who tested positive took part in the survey limits what we can say about Long COVID at a population level based on this paper.

Pictured: Not necessarily representative of the entire population. Photo by Scott Evans on Unsplash

Fortunately, we can again look at the ONS and CDC estimates of Long COVID to get some idea of the 2023 incidence of the condition. The Household Pulse survey estimated, in August 2022, that 14.8% of people in the US had ever experienced Long COVID. That estimate rose to 15.4% by August 2023, an increase of 0.6% over the course of 12 months.

Now, we don’t know precisely how many people caught COVID-19 in that 12-month period. However, the ONS survey again has a very useful estimate which suggests that at least half of the UK was infected in this time period. If we take that 50% as the lower bound of our estimate, that would mean that a maximum of 1.2% of those infected between August 2022 and August 2023 reported new long-term symptoms that may be associated with the infection afterwards.

Looking at the UK data, and using a similar methodology, the incidence of Long COVID comes out to around to a maximum of 1.08%. To contrast these figures to earlier in the pandemic, the ONS estimates that 15.1% of the population had been infected with COVID-19 by May 2021. Three months later, 1.5% of the UK was estimated to be reporting long-term symptoms. This translates to an incidence of at least 10% for Long COVID during the initial wave and early 2021.

In other words, using ONS data it looks like the rate of new cases of Long COVID has decreased by about an order of magnitude between 2020 and 2023. That’s huge.

It’s also interesting to look at the incidence of Long COVID by age. Assuming a flat incidence of new COVID-19 infections (which we know is wrong, but gives us a useful rough estimate), I looked at the incidence of Long COVID in different age groups based on the ONS data:

Note: THIS GRAPH IS WRONG. As I said, this is *assuming a flat incidence of infection*. In practice, this graph will underestimate the risk to older people (particularly those 70+), and overestimate the risk to younger people. Still, it’s reasonably indicative of the differences.

You can see the stark differences here. People over 50 remain at fairly high risk of getting Long COVID, even in 2023, but people under 25 have extremely small risks.

It’s also useful to consider that 80% of the people in this survey responded “not at all” or “a little bit” when asked the question “Does [Long COVID] reduce your ability to carry-out day-to-day activities compared with the time before you had COVID-19?”. In other words, only 20% of those above estimates relate to people who reported that Long COVID impacts their daily life by “a lot”. While the overall estimate is that 100 in 10,000 people will report some symptoms associated with COVID-19 long-term, that falls to just 20 in 10,000 reporting serious symptoms. For people under 25, fewer than 1 in 10,000 will report such issues.

Similarly, we’ve got no control group here. It’s entirely possible that the estimates are still off by quite a bit. It’s hard to know precisely how much this impacts the results, but as an example both of the Australian estimates I’ve noted above come from the same broad population at the same point in 2022 - one estimated that 18.2% of people experienced Long COVID, the other suggested that there was no difference between COVID-19 and a control group in terms of long-term symptoms.

Depending on which study you prefer, you could argue that Long COVID impacts nearly 1 in 5 people, or no one at all. Personally, I don’t think either of these arguments is likely to be true, but it shows how important definitions and methodology can be.

Bottom Line

The bottom line here is both reassuring and a bit depressing. There is no doubt that Long COVID remains a serious problem, particularly for people infected in 2020. In the ONS data, there are over a million people who report that they have long-lasting symptoms from COVID-19 that have already lasted a year or more. Those people may never recover. Regardless of your perspective on the pandemic, that’s a really big problem.

Conversely, your risk of getting Long COVID today is extremely low, particularly if you’ve been recently vaccinated. We know that Long COVID is closely related to the severity of your COVID-19 infection, and therefore things that reduce severity - such as medications and vaccination - make you less likely to experience long-term symptoms as well.

Altogether, the population-wide evidence suggests that the incidence of Long COVID is now very low. We must continue to fund research to help people who became unwell with Long COVID earlier in the pandemic, and those unlucky enough to get it today - a small percentage of a big number is still a big number, and even 0.01% adds up to quite a few people across the entirety of the US - because Long COVID can be incredibly debilitating for those who get it.

However, the risk of Long COVID has dropped dramatically since the onset of the pandemic, and is likely to continue to fall over time. If you catch COVID-19 today, the risk of experiencing long-term problems has dropped substantially since 2020.

Long COVID will likely always be a problem, but it has now become a public health issue rather than a public health emergency.