A COVID infection leads to an average of nine extra visits per year.

When I wrote this, February of 2024, 1.2 million people in the United States had died of COVID – a loss of life that exceeds what we saw in World War 2, the so-called “Spanish” flu of 1918, the Civil War, and the entire AIDS epidemic. This story is not about them.

This story is about those who survived. And, given that virtually everyone currently living in the US, with the exception of some of us born very recently, has been infected, this is a story about all of us, and the lingering impact the virus has had on the health of the nation.

I’m not really talking about long covid – which is a distinct syndrome however ambiguously defined. I’ve complained about the lack of a strong case definition before. No, I’m talking about how much health care COVID has caused us to consume – and how much it continues to cause us to consume.

Here’s the thing – COVID infections lead people to engage in healthcare activities – doctor’s visits, ER visits, hospitalizations – at least in the acute phase. But what about when that phase is over? Well, new data suggests COVID sort of resets the healthcare thermostat – an infection leads to more healthcare consumption not only during the acute phase but for weeks, months, and even years after.

Without a unified healthcare delivery system, it’s hard to figure out just how much healthcare people consume before or after a COVID infection in the US. There is one exception, of course – one bastion of truly socialized medicine in the United States: The Veterans Health Administration which runs the VA healthcare system.

And it is from VA data that we get this really telling study from Paul Hebert and Colleagues, at VA Puget Sound, appearing in JAMA Network Open, that quantifies how much healthcare people consume before and after a COVID infection.

OK – before I give you the results, let’s appreciate that this is a tricky study.  Clearly, people who are sicker at baseline – who consume more healthcare at baseline – are at higher risk of COVID and are more likely to require more medical care post-COVID. You need a good control group.

So – you want to find people who are very similar to someone who got COVID, but who didn’t get COVID. At least, who didn’t get it while under observation from the study. But methodologically, this is really hard. And it’s hard because, as I teach my trainees, in a study like this you need to know when to “start the clock”.

What I mean by that is that you need to define the precise transition time between the pre-covid and post-covid state for each individual.

For someone who is infected with COVID, that’s easy. You can define the transition as the first positive test. That’s not exactly right of course, but it’s close.

But what about the control group?  How do we compare their healthcare usage before the got COVID vs. after they got COVID if they never got COVID?

This is actually a pretty fundamental difficulty with studies in the field of pharmacoepidemiology – where a drug is given at a certain point in time – except of course here the drug is a virus.

The authors adapt some well-worn pharmacoepidemiology techniques to solve this dilemma. First, they identified 202,803 Veterans who had been infected with COVID. They then matched those 200,000 to 200,000 Veterans who hadn’t been infected (the VA is pretty big) ensuring exact matches based on sex, state of residence, immunosuppressive medication use, and vaccination status. Then, they further matched based on a propensity score for COVID infection – this is a single number that reflects an individual’s likelihood of being infected, and contained data from 37 covariates which included other demographics, comorbidities, and even healthcare utilization variables.

In the end, what they had was something of a “digital twin” study – 200,000 who were infected, and 200,000 who, based on all their covariates, were incredibly similar, but who did NOT get infected. Now you can assign that transition time in the control patient easily – it’s just the date their digital twin got infected. Neat, right?

You can see here just how similar these groups were. No, it’s not a randomized trial. But it’s not bad. Age, BMI, Smoking status – all were well matched between infected and controls.

Adding to the validity of the approach, prior to the infection, both groups had similar healthcare utilization. For example, let’s look at the use of Emergency Care.

Prior to infection, rates were very low, but similar across both groups. And then, as you might expect, rates simply soar in the infected group. This is because, and I don’t mean to be pedantic, that when people get infected with COVID they are much more likely to go to the ER than if they are not infected with COVID. Note that the ER visit rate drops quite quickly, but remains slightly above the control group, after the infection.

Ok this is where the study gets interesting. Let’s look at primary care utilization. Pre-infection – similar rates. Around the time of infection – big increase in the COVID group – no surprise there. But then look at the tail. Even 52 weeks out from infection, the COVID group still has higher utilization. This amounts to around 2 extra primary care visits per year per COVID infection.

Look at visits for mental health. Again, a spike at the time of infection – which is a bit more surprising than seeing a spike for ED or PCP visits, but regardless the long tail tells us that something significant is going on here.

The authors combine primary care, specialty care, surgical care, emergency room care, and mental health care and determine that the average COVID-19 patient has around 9 more visits in the 52-weeks after infection than the very well-matched controls.  That is simply a huge amount of healthcare utilization.

OK – this is a problem – healthcare is a limited resource after all. But what can we do about it? Well – subgroup analyses in the paper identify groups for whom infection is NOT so bad in terms of increasing healthcare utilization. Utilization always went up after infection. But the increase in utilization was less profound in younger individuals, and in later waves of the pandemic which is encouraging. And I should point out that the difference in utilization was also less marked among those who were vaccinated – which is, you know, a factor we can actually address.

This study telegraphs the untold sequela of the pandemic – the stuff that is hard to quantify in our day to day practices, but that we nevertheless feel. Doctors like me are busier than ever, more stressed than ever, dealing with patients who are sicker than ever. Part of this is the long tail of the pandemic – the flotsam and jetsam of a world-changing event. We survivors are, of course, the lucky ones. But we are also the ones who must live in a world with challenges we are only beginning to discover.

A version of this commentary first appeared on Medscape.com