The instantaneous risk (i.e. at any time point) for a vulnerable person to get infected with #COVID19 in a random mixing population is simply the prevalence of the virus in the population.
(1/9) https://twitter.com/reidatcheson/status/1307349171595288577

Though, if we assumed that a % of a population will get infected eventually before the epidemic recedes, the entire population social-distancing increases the proportion of 'at-risk' people getting infected, relative to a scenario where only at-risk people sheltered.
(2/9)

The risk of #COVID19 to different demographies is highly variable, with elderly being at >1,000x higher risk than young people. As such, a reduction in the number of infections, but leading to a higher proportion of elderly being infected can lead to a higher death toll.
(3/9)

Let us assume that 50% of the population could get infected by #COVID19 before a vaccine is available. We divide the population into a high-risk (10% >75 years old; IFR~10%), and low-risk group (90% <75 years; IFR~0.2%). Now consider two countries both with 10M inhabitants.
(4/9)

In country A, everyone social-distances, with considerable success. As a result only 10% of the population gets infected. Though, as everyone social distances, the risk of infection is similar for everyone. This leads to ~10,200 deaths.
(5/9)

In country B, no one outside those in the high-risk group takes precautions. As a result 50% of the population gets infected. Though, those in the low-risk group are 10x more likely to get infected. This leads to ~6,000 deaths, about 60% of the death toll of country A.
(6/9)

In this hypothetical toy example, country B is likely in a better situation not only because it experienced fewer deaths but also because a far larger proportion of the population has been immunised (50% vs. 10%).
(7/9)

This toy example is not particularly realistic nor carefully parametrised and does ignore several important considerations. There are also challenges to implement an effective strategy to protect the 'at-risk' population in a humane and effective way.
(8/9)

That said, the toy example illustrates that population-wide 'social distancing' and lockdowns are not necessarily the most effective strategies to minimise the death toll of #COVID19, even when ignoring the collateral damage of invasive public health interventions.
(9/9)