COVID-19 and QALYs

|

How bad is COVID-19? That is far too complex a question for a blog post, so let's focus on a much simpler question: What would an answer to the previous question look like? COVID-19 will cause deaths and illness, and attempts to reduce the deaths and illness will themselves impose economic costs. We could use a metric like QALYs (quality-adjusted life years) to aggregate the burden of COVID-19. Then, in assessing public policy interventions, such as closing schools and limiting travel, we could at least informally perform some sort of cost-benefit analysis, assessing the economic cost of an intervention and comparing it with the benefit in reduced deaths and illness.

Yet the QALY does not seem to enter into discussions of COVID-19. We see a great deal of discussion of mortality rates among those infected by the novel coronavirus, including reports of how those rates vary by age band. The question, though, is not just the age of those who have died, but how much longer they would have lived. I have not seen any systematic attempt to convert these mortality rates into a QALY number, let alone any formal cost-benefit analyses assessing how different interventions might save QALYs. Perhaps the government or some private citizens have created such analyses, but if so, they don't seem to show up on Google or Twitter.

Why? Some possible explanations:

(1) Too many unknowns. Cost-benefit analysis, the argument goes, is a process that we can undertake only when we have sufficient time to accomplish it with care. There are many unknowns concerning COVID-19, including how many people carry the virus with mild symptoms or asymptomatically. It is difficult to construct a baseline case of how many people might become infected and die absent government intervention, so any cost-benefit analysis would largely be guesswork.

(2) Real option theory. A more sophisticated version of the previous hypothesis is that a cost-benefit analysis might be too simplistic and in particular might understate the value of extraordinary efforts to contain the virus while we develop better information about its dangers. For example, a cost-benefit analysis based on current assumptions of mortality might suggest that many interventions to reduce the spread of the virus are with high probability not cost-benefit justified, but that with low probability the pandemic will be so bad that such interventions are critical. In theory, we might assume some distribution of how deadly the virus is and simulate the effects of interventions now, taking into account that our information will improve in the future and thus allow better decisions then. But this means that the cost-benefit analysis must account for many more unknowns, plus it is difficult to aggregate the results into a single punchy conclusion.

(3) Lost QALYs. Discussion of QALYs almost always focuses on QALYs saved by interventions. The first step in a COVID-19 analysis would focus on the QALYs that would be lost in baseline. We frequently see reports of how many people die from cancer, but few reports of how many QALYs are lost to cancer. There are exceptions, such as this one measuring the global burden of disease with the related measure of DALYs (disability-adjusted life years). An explanation for the general lack of discussion is that the baseline burden of disease doesn't matter for policy as much as lives saved from particular interventions. Because we are not used to thinking about disease burdens in QALYs or DALYs, we don't seek to measure the burden of some new disease in those terms, even though we need to do that before we can assess interventions.

(4) Not medical treatments. Typically, QALYs are used to measure life-years saved by new medical treatments. Here, the question is the expected QALY savings from quarantines, canceling sporting events, and the like. We have standard methodologies for conducting randomized trials of medical treatments, and these methodologies can be used to generate QALY numbers. We may not have standard methodologies for measuring QALYs for government actions that are not medical treatments. In principle, however, we could use epidemiological models to estimate morbidity and mortality and translate that to QALYs.

(5) Prospect theory. The value of a QALY may differ based on whether we focus on a QALY saved or a QALY lost. Because our baseline is the pre-COVID-19 world, whether we are focusing on the burden of disease or the effectiveness of interventions, we frame the new suffering as losses rather than gains. Thus, arguably we should assign a higher value to a QALY with COVID-19 than when focusing on treatments of more familiar diseases. But questions of what a QALY is worth are second-order questions, worth asking when we want to compare QALYs to economic costs of particular interventions. It would be nice to have estimates of QALYs saved by interventions even if we might reasonably disagree about the trade-off between QALYs and economic costs.

I have no idea which steps to slow the spread of COVID-19 are cost-justified and which steps are not. My own ignorance doesn't matter, but I worry that health organizations and governments may not even be trying to compare costs and benefits in any systematic fashion. Overreaction in the form of an availability cascade is especially likely at the beginning of a crisis, yet there is also a danger of complacency. There may be some good reasons not to base policy decisions solely on comparisons of QALYs and costs, but production of at least back-of-the-envelope estimates could be useful in anchoring serious policy discussion, even in or maybe especially in times of crisis.

NEXT: McGinnis: The Empire Strikes Back Against Originalism

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. Excellent question – one that I feel embarrassed hadn’t crossed my mind (given I’m a big believer in using QALY in public health policy in general).

  2. You can’t do a cost benefit analysis because there is no way to make a baseline because the costs are so varied…what is the metric by which you compare a school being closed until summer break (missing 2 months of school is lower test scores) compared to cancelling a concert where large crowds would gather ($ of ticket holders), compared to churches suspending Easter services (missing connection with God) to the potential of killing 2-3% of those only potentially infected?

    1. I agree that it’s very hard to do. But I think even back-of-the-envelope estimates would be better than nothing.

      1. Thanks for responding prof.!

        A bad measurement is potentially worse than no measurement at all, and measurement and quantification have become somewhat of a fetish among ecomomists. An example of a bad measurement leading to bad decision-making is the risk indexes prior to the 2008 crash.

        On a more practical note, the decisions like this are going to be politically based for optics rather than on quantifiable metrics. And with different criteria for different actors. For schools, they will cancel quickly, as they will treat the virus like a bomb threat, which virtually 100% of the time are fake, but they don’t want to take any chances and can be held liable. While event promoters, though I am no expert on the subject, won’t be liable for a virus spreading at a concert (unlike broken fire doors) and the risk of attendance and getting sick is borne by attendees and the cost for cancelling is borne by the artist/promoters (who might have insurance).

  3. Why isn’t anyone doing this kind of analysis (or if they, admitting to it)?

    That’s simple.

    In your “some sort of cost-benefit analysis, assessing the economic cost of an intervention and comparing it with the benefit in reduced deaths and illness.”, you’re putting dollars up against dead bodies.

    While such analysis is common, it’s universally frowned upon to publicly admit to such, because it will always be bad optics to say “well, we could have saved 500 kids at that school, but we didn’t think 500 kids were worth the $500,000 in lost productivity.” Even if that’s the actual decision-making process, nobody wants to admit it.

    1. I agree with you (though of course everyone would think 500 kids worth $500,000 in lost productivity). But it still doesn’t explain why we don’t estimate costs and/or benefits separately.

      1. But it still doesn’t explain why we don’t estimate costs and/or benefits separately.

        You’re confusing people not using that language with people not doing it.

        From the individual person making the decision “stay home sick and lose wages, or go to work and spread it if I’m actually sick?” to the city/state that shuts down events, people are making that analysis.

        They just talk about it that way.

        1. Probably a better comparison would be “the healthy person making a decision to stay home from work to avoid catching the virus and losing wages as a result”. That scenario has a cost to the individual in exchange for reducing risk, the scenario you mention always has a net gain to the individual (they still get paid; other people may get sick but they bear the cost of that).

      2. Depends on which kids we’re talking about, actually.

    2. He’s doing more than putting dollars against dead bodies. He’s rating the bodies. He’s saying these bodies are worth a lot, these other people not so much, and those others, worthless. Quality people are worth paying attention to. Non-quality people, not so much. Under his approach, something that helped high-quality people more but actually killed off low-quality people could come out preferable to something that helped everyone.

  4. #5 on ‘prospect theory’ is an intriguing problem. When looking back at SARS or MERS, we pretty much think of them as one-time epidemics. Each one escaped from wherever it was hatched, ran its course, and is done, and we return to baseline. And so, as you say, we think of QALY’s as lost with respect to the baseline. Likewise for historical specific bad flu epidemics such as ‘swine flu’ or ‘bird flu’.

    But in contrast, if ‘the flu’ is the generic term for whatever sets of viruses go around in any cold season, then they are in the generic baseline, and having vaccination programs and other preparations adds QALY’s to the baseline.

    AIDS is now endemic to the human species worldwide, and will be for the foreseeable future. It is the new baseline. And so we estimate QALY’s added thru treatment.

    And the point is, we don’t know enough about COVID-19 to assume that SARS and MERS are the right analogies, running a lice cycle and expending themselves, or is COVID-19 something which will keep going round and round the world in cycles, as was the case with bubonic plague for many centuries.

  5. Could we be honest and call it productivity-adjusted survival? The OP appears to be referring to the economic loss resulting from deaths. This is a reference people’s economic value; less valuable people contribute less to society, so their lives are worth less.

    Could we stop the doublespeak and stop pretending we are talking about people’s interest in their own lives? Could we at least be honest that the issue here is which lives are of interest to others?

    The US has long frowned on doing the sort of adjusted survival European countries routinely do. There is a reason.

    Obviously control of symptoms and freedom from pain is an independent benefit that has independent value. But the idea that disabled lives get severely discounted because we think their lives just aren’t worth much is not an idea that has been accepted in the US the way it has in Europe. It shouldn’t just be taken for granted as a starting point.

    1. Reader Y….You clarified the moral issue quite well: This is a reference people’s economic value; less valuable people contribute less to society, so their lives are worth less. Could we stop the doublespeak and stop pretending we are talking about people’s interest in their own lives? Could we at least be honest that the issue here is which lives are of interest to others?

      This proposal by Professor Abramowicz is ghoulish. Imagine some anonymous bureaucrat (with their biases) deciding whether you get treatment based on your perceived value to society. The communists did that.

      1. I don’t follow. I don’t see where it has much to do with your value to society. From my brief research, QALY (Quality Adjusted Life Years) simply calculates the expected length of a person’s life discounted by their projected health problems during that time. 1 is a year of perfect health, and 0 is dead.

  6. Good article. I did a rough analysis using WALYs (well-being adjusted life-years, aka WELBYs), being similar to QALYs but broader as it measures overall well-being (specifically, life satisfaction) rather than specifically health. This lets us compare pandemic deaths with the harm of economic crashes such as the 2008 financial crisis.

    From this it turns out that a financial crisis like the last one would be far worse than the deaths. Calculation in footnote 7:
    https://medium.com/me/stats/post/854570873153

    1. Though I’m not suggesting that implies the deaths should not be averted, since that would no doubt still lead to some kind of financial crisis. I’m not proposing any particular policy. But the fact that the loss to billions through hardship will probably be far greater than the loss of life implies that protecting the economy is the overriding consideration, though the optimal solution may well include saving both.

Please to post comments