Cost Benefit Analysis of Flattening the Curve
Eline van den Broek-Altenburg, an assistant professor at the Larner College of Medicine, University of Vermont, and Adam Atherly, a professor and director of the Center for Health Services Research at the Larner College of Medicine, University of Vermont have produced a short analysis titled “Economic Cost of Flattening the Curve.”
It’s interesting and valuable, but it’s mis-titled. The best part is their analysis of the benefit of saving lives. Even that, though, is misleading because they write as if their implicit assumption is that implementing a plan would save all the lives that would have been lost to the COVID-19 disease. NO ONE believes that.
Still, let’s go on. They point out some very interesting facts about the data from Italy, facts that have, in most of what I’ve seen, not been given their due. The most important one is this:
The Italian National Health Institute pegged the median age of death from COVID-19 in Italy at 80.5. This is consistent with early data from the United States.
They also say that there’s good reason to think that the mean age of death for the above is close to the median.
If you’ve ever seen one of these studies, you know where they’ll go next: calculate the number of extra life years saved by a death averted. They write:
The average 80-year old in the United States has a life expectancy of about 9 years, suggesting that on average, a death averted will “buy” 9 extra years of life. In QALY-estimations, this number needs to be adjusted for the “quality of the years”. In Italy, 99% of deaths had an underlying pathology that needs to be incorporated in QALY adjustments. If we use diabetes as a reasonable proxy for the many chronic diseases, we would adjust the 9 years down to 7.8 years or QALYs. In other words: the average loss per person of quality-adjusted life years is 7.8.
QALY stands for Quality Adjusted Life Year. It’s what a lot of health economists use in estimating gains from various drugs, surgeries, changes in safety, etc.
Then they take an expected range of deaths of 200,000 to 1.7 million. Although only 2 weeks ago, I told my friend Charley Hooper that I expected at least 200,000 deaths, my own view now is that 200,000 is too high.
Then they multiply these deaths by 7.8 years to get 1.56 million to 13.26 million QALYs.
They next want to consider the cost of achieving the benefit of not having these deaths. Let me remind you, though, that many of the deaths have already happened and you can’t prevent all future deaths. So their measure necessarily overstates the benefits (assuming you buy the idea of QALYs) of any government policy.
Now let’s go to the even bigger mistake. They don’t measure the cost of the measures to reduce (or, they would have to say, end) deaths. No. What they do instead is look at the cost of the stimulus bill that Congress is considering. But wait. The stimulus bill isn’t aimed at preventing deaths. It’s aimed at compensating people for losses (along with some pork). So I found the rest of their analysis not that useful.
Here’s how to do it right.
Estimate the number of deaths averted, and the related number of QALYs saved, by implementing some government strategy. Then estimate the cost of this strategy. A rough measure would be, say, one month of average wages lost (minus 20% of wages lost, assuming people value the leisure somewhat) times the number of people losing these wages, plus one lost month of profits on businesses. That would be a much better measure, if the measure(s) cost about one month of output. I think it would be the same order of magnitude as their measure.
That means that their conclusion is understated. Here’s their conclusion:
In theory, if decision rules like cost effectiveness represent sensible approaches to making policy choices, health economists should be out front helping guide policy. Our calculations suggest that current strategies will be cost effective only if the predicted mortality rates are at the top of the predicted range and costs are at the bottom of the range – a combination of worst care / best case that is unlikely. Consideration of targeted mitigation strategies that minimize the economic cost while protecting the most vulnerable are warranted.
Why do I say their conclusion is understated? They’ve probably got the order of magnitude right on the cost side even though theirs is not an estimate of the cost. But by assuming all lives are saved, even those that are already lost, they may well have at least doubled the number of quality-adjusted life years saved.
HT2 Jerrod F. Anderson