I recently came across these two articles that both asked for a cost-benefit analysis of the massive interventions that are underway in Italy: https://www.theatlantic.com/international/archive/2020/03/coronavirus-covid19-politics-crisis-boris-johnson-britain/607456
I decided to give it a shot.
TL;DR: Italy's interventions may well be cost-effective with an estimated cost of $2,200/life-year saved, but this hinges on multiple, large assumptions and is prone to change based on further information/refinement.
Using this article referencing an economist estimating a 10-15% decrease in Italy's GDP and this article from the Lancet showing average age of non-survivors of COVID-19 to be 6930566-3), I went about roughly calculating expected costs per life-year saved. I'm treating this more as a Fermi problem to get a back of the envelope estimate, so lots of assumptions are being made along the way.
Per a WHO actuarial table on China, a 65-69 yo male has 14.9 years life expectancy, and a female has 16.9 years life expectancy; for simplicity, I will use 16 years as a combined expectancy. With a case-fatality rate of 2%, this would mean about 0.32 unadjusted life-years lost per case.
Italy has reduced its GDP by 10-15% for maybe 2 months, resulting in what may be about a 2% annual decline. If applied globally, this would translate to 2% of $17,300 per capita (PPP), or $350.*
Given the above, we can see what benefit would be derived from various assumptions of the benefit of Italy's policies. If their policies reduce mortality by 50% (from 2% to 1%), then they are gaining 0.16 life-years, and at a cost of $350 per person, that comes to $2,200/life-year gained. A 25% reduction (from 2% to 1.5%) would mean $4,400/life-year gained, and so on. A general equation would be Cost/life-year= $350/(0.16*ARR), where ARR is the absolute risk reduction in percentage of deaths.
This doesn't directly translate to QALYs since I did no quality adjustment, but this still falls well under the UK NICE guidelines deeming $20-30K as a reasonable threshold for cost effective treatments. The problem is that the ARR is very difficult to estimate, but we can get a lower limit for a cost effective ARR, which would be 0.07%.
Obviously there are lots of assumptions and rough estimates in this analysis, but I think that back of the napkin calculations can still be useful and important, helping appropriately frame the discussion and allowing updates as more information becomes available. Most of these assumptions are very favorable to the analysis, such as removing quality adjustment, assuming that the patients who died have a baseline risk of dying on the actuarial table, and not including opportunity or testing costs to the evaluation, but it at least can serve as a starting place.
Would love any feedback or other thoughts/refinements!
*This does NOT take into account costs associated with treating those with the disease, which I consider to be dead weight in the sense that we will be treating them regardless of the policies being in place. This also does not look into the costs/benefits associated with reducing resources that could be spent towards other diseases (opportunity cost), like MIs, strokes, etc. that continue to occur during the pandemic.
Source: Original link