Estimating risk of long COVID using a Bayesian network-based decision support tool 2025 Lau et al

[Andy]

Abstract​

Importance​

Long COVID causes substantial health burden globally, affecting over 30 % of adults who have ever had symptomatic COVID-19. Individuals at continued risk of long COVID need better and more accessible information to make choices about vaccines and treatments.

Objective​

To quantify modifiable risk factors for having long COVID six months post-infection, and develop a decision support tool for managing the risk factors.

Design, setting, and participants​

A Bayesian network (BN) model was developed to estimate the probability of long COVID depending on demographics (sex, age), comorbidities, and modifiable factors (vaccination history, number of previous SARS-CoV-2 infections, and drug treatments during acute infection). Data were sourced from published studies and government reports.

Main outcome(s) and measure(s)​

Outcome measures include probability of hospitalisation, ICU admission, and dying from COVID-19 during the acute infection under different scenarios of demographics, comorbidities, vaccine coverage and effectiveness. The BN also estimates the risk of developing long COVID depending on modifiable risk factors, and persistent symptoms related to specific systems (cardiovascular, gastrointestinal, musculoskeletal, pulmonary, neurological, renal, metabolic, coagulation, fatigue, and mental health).

Results​

Vaccination, receiving drug treatment within three days of acute infection, and avoiding repeated infections are the greatest modifiable influences of long COVID development, decreasing risk by up to 63 % under modelled scenarios. The interactive user-friendly web-based decision support tool (https://corical.immunisationcoalition.org.au/longcovid) enables easy access to model outputs, and allows individuals to calculate their personalised probability of long COVID under different scenarios of modifiable risk factors.

Conclusions and relevance​

The decision-support tool can be used by individuals or in conjunction with clinicians for shared decision-making on vaccination, pursuing early drug treatment during acute infection, and continuing protective behaviors such as masking and social distancing. The model can also generate population-level estimates of outcomes to assist public health decision-makers to design better-informed public health policies.

Open access

 

[Sean]

Vaccination, receiving drug treatment within three days of acute infection, and avoiding repeated infections are the greatest modifiable influences of long COVID development, decreasing risk by up to 63 % under modelled scenarios.​

 

[Utsikt]

Vaccination, receiving drug treatment within three days of acute infection, and avoiding repeated infections are the greatest modifiable influences of long COVID development, decreasing risk by up to 63 % under modelled scenarios.

So learned immunity is good, but the cost of experiencing an infection doesn’t outweigh the benefit of the immunity the infection provides.

In other words: maximise vaccination and minimise infection. The complete opposite of what most countries are doing right now.

 

[Yann04]

decision support tool (https://corical.immunisationcoalition.org.au/longcovid

this web tool is kind of cool. Almost worth sending about as a soft advocacy thing.

 

[rvallee]

So learned immunity is good, but the cost of experiencing an infection doesn’t outweigh the benefit of the immunity the infection provides.

In other words: maximise vaccination and minimise infection. The complete opposite of what most countries are doing right now.

The strategy of leveraging learned immunity was meant to reduce acute illnesses leading to hospitalizations, but its proponents did not model the long-term consequences of doing so because they simply did not believe in them. So by their own definition of success, they were right, as long as no one actually looks at the full set of consequences.

It was about yanking the bandage quickly on the misguided idea that it would hurt less, which is obviously how the experts put it because this is how Boris Johnson, who has very few filters for things like this, said it out loud, no doubt this is what he was told by medical advisers to the UK government. They assumed it would be a one-time thing, based on nothing at all. Mostly based on movie logic: "it has to work because this is all we got". Except it was never all they had, they just chose this because it was easier for them and they wouldn't face the consequences. They had rich people breathing down their necks telling them to get it done, and many of them were in that same group.

This would have worked with a neutralizing vaccine, or if COVID had mutation rates closer to the flu. But both those conditions were always false, and the push for a strategy of mass infections was promoted before vaccines were available, and simply ignored the fact that the mutation rates were too high for this to ever work out.

After years of terribly flawed pseudoresearch trying desperately to blame Long Covid on any bit of nonsense that makes it a personal responsibility, the reality that it's the infections that are the only relevant factor cannot be avoided. It will still be avoided, because every government and public health and medical institutions embraced the strategy, even promoted it to the public, and so they are all guilty of causing this. They will never own to it, this is not how humans work.

So now we not only have to deal with the long-term consequences of COVID, themselves on top of the long-term consequences of other infectious diseases, but the long-term consequences of those other diseases were also made worse by endlessly promoting the idea that natural infections are good for health because they (are said to) avoid severe acute illness leading to hospitalizations. This is three wrongs and no right, an impressive level of failure for anyone, coming from experts this is nothing short of the death of expertise.

Many of those same experts will whine and whine about the things happening in the US under RFK Jr, utterly oblivious to the fact that they are just as wrong about some things, in slightly different ways.

 

[EndME]

Regarding the Long-Covid aspact, I struggle to see how this is not just another useless "EHR study on Long-Covid".

From what I can see for "Long-Covid" they looked at 2 variables in the EHR records. Positive Covid test result recorded in EHR-records and symptom presence of at least one symptom from a list of certain symptoms recorded at least 12 weeks after this positive test result. They refer to this as "persistent symptom" but as far as I can tell do so without examing this.

This does not match the already existing very inclusive and weak Long-Covid definition of the WHO which requires "Post COVID-19 condition occurs in individuals with a history of probable or confirmed SARS CoV-2 infection, usually 3 months from the onset of COVID-19 with symptoms and that last for at least 2 months and cannot be explained by an alternative diagnosis."

So learned immunity is good, but the cost of experiencing an infection doesn’t outweigh the benefit of the immunity the infection provides.

It seems to me you are right and for example there seems to be very strong evidence for vaccines reducing severity of acute infection ect, but as far as I can tell this study doesn't allow any causal interpretations. The risk analysis showed that roughly 60% of people would have a fitting Long-Covid symptom in their EHR record 6 months after infection. If I'm reading things correctly if you have more infections your risk of another symptom event falling into that 6 month timeframe naturally increases, completely independently of Covid. For example if you're a perfectly healthy football player who's only problem is that he tears a muscle every 3 years, your risk of a torn hamstring (possibly recorded as muscle pain in your EHR record) fitting the Long-Covid bill in this study increases if you have more infections because the chances of a six month overlap increase without there possibly being any connection whatsoever. I wonder whether the authors tried to control for such factors for example by looking at random timeframes of 6 months as well, but that initself may become a bit tricky if there are spurious sources of bias related to seeking healthcare services (which might well exist due to comorbid conditions). Cumulative risk always increases in such a definition independently of what the actual relationship is and it seems to me that is what was found.

 

[rvallee]

this web tool is kind of cool. Almost worth sending about as a soft advocacy thing.

It's neat but I don't see it as having much real-life value. There is no way to know which infection will lead to long-term illness. This factor is entirely probabilistic and hidden from view. In some very healthy fit people it was the first infection that did it, in people living with multiple health problems it could be the 3rd, or 5th. It is rather much like artillery warfare. Most soldiers in the field will not have a single shell drop close to them, while some will get their legs blown off or mostly vaporize from a direct hit.

The only way to avoid getting hit is not to be on the field, and that's no longer an option. Because of the strategy of mass infections, the only way out of this is by developing treatments, it's far too late to even consider any other option, they're not even available anymore thanks to the promotion of a scientifically invalid strategy.

The harm this will do to the credibility of medicine is many times worse than the antivaccine movement. But they will blame TikTok anyway.

 

[Utsikt]

It seems to me you are right and for example there seems to be very strong evidence for vaccines reducing severity of acute infection ect, but as far as I can tell this study doesn't allow any causal interpretations.

Vaccine doses don’t follow the same mechanisms as your example, because they don’t lead to increased rates of sampling like infections do.

If confounders are sufficiently controlled, and fewer vaccines doses are associated with higher rates of symptoms after infections, is that not proof of casuality?

I agree that you need to get around the sampling rates to determine if more infections is actually worse than fewer infections, even though I can’t think of any logical way for that to not be the case.

 

[EndME]

Vaccine doses don’t follow the same mechanisms as your example, because they don’t lead to increased rates of sampling like infections do.

If confounders are sufficiently controlled, and fewer vaccines doses are associated with higher rates of symptoms after infections, is that not proof of casuality?

Yes, in my head I was not thinking about the vaccines data (and rather about the whole statement I quoted "So learned immunity is good, but the cost of experiencing an infection doesn’t outweigh the benefit of the immunity the infection provides.") because I think for vaccines alone there's anyways already enough studies showing benefits as outlined here, even if it's possibly mostly relevant to reducing risks of severe acute Covid or just acute infection which should be convincing enough already. I had the impression that data on vaccine effectiveness against hospitalisation and ICU admission at the beginning of the pandemic was pretty clear.

I haven't had a closer look but it's my impression that much of the data here is the same data of those Al-Aly studies which have been critiqued much on S4ME, much due to insufficient control of confounders, and some Australian data. At least Al-Aly's data always looked meaningless to a layman like me and at least doesn't seem to have advanced the field, but I think there should anyways be enough other good data on vaccinations that proves causality.

I agree that you need to get around the sampling rates to determine if more infections is actually worse than fewer infections, even though I can’t think of any logical way for that to not be the case.

I assume it to be the case and I think most BPS enthusiasts would as well, but my argument is more just that these studies don't add anything. It's just the repetition of the same things that RECOVER has been critiqued for for 5 years.