How Decision Theory Can Help Guide Our Actions
With lots at stake in the Covid-19 pandemic, Professors Sylvia Richardson and Sir David Spiegelhalter look at just what we could possibly get out of the decision-making theory when deciding our next moves.
The opinions were mixed on the wisdom of the wisdom of the decision to loosen all social distancing regulations in England from July 19. Without taking an opinion on the plan, as statisticians, we can ask precisely how this choice was made – and also, in particular, whether a systemic view was being taken.
This implies recognizing the dependency between the various metrics we use to measure the health impact of disease as well as its social impact, particularly how these change over time: as an example, how the impact of relaxed restrictions will undoubtedly change as the occurrence of disease increases.
What can we learn from decision theory? It helps us select among actions whose consequences are uncertain. In its ideal form, we name probabilities for occasions, associate numerical utility with all the impacts of activities, and then choose the change with the maximum anticipated utility. Of course, in the real world, we cannot confidently do this full technical evaluation, however, it still provides us with a logical thought process to initiate a 360-degree view of the consequences of plans.
Let’s draw on the example of the current wave of infection. We have seen various projects predict a multitude of situations and therefore an increased burden on the NHS. The federal government estimated 100,000 cases a day later in the summer. Recent patterns have been encouraging. However, in the coming weeks, multiple ups and downs in the daily number of favorable examinations are expected.
The atmosphere of transmission is highly heterogeneous and affected by numerous time-sensitive factors, from progression of resistance to seasonal vacation mobility, from changing work patterns to sporting events. Since most epidemic models rely on typical patterns of contact between groups, they also cannot fully explain the heterogeneity of transmission, since the predictive circumstances obtained have significant uncertainties. In light of this, it is necessary to communicate clearly how thought has been given to the economic and social effects of possible caseload, specifically with regard to the substantial disruption caused by the existing contact tracing and isolation plan, which is discussed daily in the press.
Recall that the screening and isolation plan of the Test and Trace system was established when the goal was to manage the infection rate at a relatively low level by using social distancing as efficiently as possible: a very different context from today. The number of individuals currently isolated is an important action of the consequence of a combination of relaxation and transmission control policies. If we are to properly adjust our transmission control plan to a situation of increased contacts, we need evidence of a number of alternatives. It seems extraordinary that there is still such significant unpredictability about the relative effectiveness of release testing compared to ten-day self-isolation orders, when release testing guidelines were discussed in the context of the institutions as early as December 2020.
Regrettably, much of the discussion does not seem to have taken a 360-degree view of the systems; instead, we fear that the situations have focused directly on a few metrics. The focus on cases, hospitalizations and fatalities, the influence of ripple effects, for example, of those who need to self-insulate in what has been called “the relic,” seems to have been largely neglected in presenting the economic benefit of increasing constraints. Similarly, it would certainly be important for there to be a comprehensive discussion of the changing impact of loosening constraints if example rates rise considerably in the autumn, either globally in the UK or in your area.
We don’t seem to have learned any important lessons from the financial crisis, where previous financial modeling mostly ignored trust between unusual occasions and instead made dubious independent presumptions, with huge repercussions that we all experienced. For example, we would expect a view of the existing system to include consideration of a metric that evaluates the chance of a vaccine-resistant homegrown variant emerging for the offered levels of population infection and vaccine coverage. To apply rational decision-making to the Covid pandemic, we need metrics that cover the full range of consequences.
Read the original article on: Royal Statistical Society.
