Counting the costs of COVID-19: why future treatment option values matter

I critique a recent analysis (Miles, Stedman & Heald, 2020) of COVID-19 lockdown costs and benefits, focussing on the United Kingdom (UK). Miles et al. (2020) argue that the March-June UK lockdown was more costly than the benefit of lives saved, evaluated using the NICE threshold of {\pounds}30000 for a quality-adjusted life year (QALY) and that the costs of a lockdown for 13 weeks from mid-June would be vastly greater than any plausible estimate of the benefits, even if easing produced a second infection wave causing over 7000 deaths weekly by mid-September. I note here two key problems that significantly affect their estimates and cast doubt on their conclusions. Firstly, their calculations arbitrarily cut off after 13 weeks, without costing the epidemic end state. That is, they assume indifference between mid-September states of 13 or 7500 weekly deaths and corresponding infection rates. This seems indefensible unless one assumes that (a) there is little chance of any effective vaccine or improved medical or social interventions for the foreseeable future, (b) notwithstanding temporary lockdowns, COVID-19 will very likely propagate until herd immunity. Even under these assumptions it is very questionable. Secondly, they ignore the costs of serious illness, possible long-term lowering of life quality and expectancy for survivors. These are uncertain, but plausibly at least as large as the costs in deaths. In summary, policy on tackling COVID-19 cannot be rationally made without estimating probabilities of future medical interventions and long-term illness costs. More work on modelling these uncertainties is urgently needed.