With coronavirus infection rates falling – and the predicted economic and welfare costs of lockdown so high – governments across the world have tentatively started lifting restrictions and reopening their economies.
But to do this safely and prevent a second wave of infections, policymakers need to understand the effects that lockdown measures such as closing schools or businesses – known as non-pharmaceutical interventions (NPIs) – are having on containing the virus. This will help predict the effects of rules being eased, and also show which interventions are most effective.
At the moment, though, we don’t know which lockdown measures work, and testing their effectiveness is difficult. So we need to get inventive. One overlooked way of assessing them is to use randomised controlled trials (RCTs).
The unique power of randomisation
When it comes to testing new medicines, there’s a broad consensus that RCTs are the gold standard. These trials compare the outcomes of randomly selected groups of people that do and do not get a treatment. Researchers can use RCTs to be confident that a new medical procedure works. The same rules should apply when testing the effectiveness of NPIs.
By introducing or loosening an individual control measure in some randomly selected regions but not in others, the RCT method would offer the unique opportunity to show whether or not that intervention (and not some other common factor) causes a reduction in the disease transmission. What’s more, this method gives policymakers and researchers full control over implementing changes and directly measuring outcomes, reducing the scope for errors. These are advantages not shared with many other methods.
And although RCTs have not gathered much attention for studying NPIs during the current pandemic, economists and other social scientists have already shown that they can be used outside of clinical medicine. For example, over the past two decades RCTs have become a popular method among behavioural and developmental economists testing which interventions are most effective in alleviating poverty.
How an RCT would work in practice
As an example, consider the question of reopening schools.
In many countries, schools are all scheduled to reopen at the same time (and at the same time as other lockdown measures are being lifted). Changing several policies all at once will make it hard to attribute any subsequent rise in infections to the removal of a specific measure.
Researchers in Norway have floated an alternative plan. Why not randomly split districts where opening schools is generally considered as safe into two groups? Schools in the “treatment” group would be allowed to open two or three weeks earlier than those in the “control” group (what’s known as a “phase-in” approach). By closely monitoring and comparing infection numbers across districts, policymakers would gain a much clearer picture of whether opening schools causes cases to rise more rapidly, and by how much.
This example illustrates a broader principle. RCTs could be used to answer a large spectrum of policy questions about easing or introducing new lockdown restrictions. These could range from the number of students per classroom, to the requirement to wear masks in public, to the opening of parks and beaches.
However, as discussed in a recent study, there are several things needed to make RCTs a successful tool for evaluating NPIs. Foremost is the ability to accurately measure and model the effects of an NPI on virus transmission, as well as the economic and psychological costs associated with either lifting or maintaining the intervention.
RCTs can fix current errors in thinking
Now that several countries are beginning to lift some restrictions, political commentators are quickly giving in to the temptation of comparing infection or fatality rates between places with and without certain rules in place. However, such comparisons are easily misleading because they lack the random assignment that would show whether or not a particular measure has caused a reduction in disease transmission.
We would not trust a medical study that selectively treats patients based on their prior health status. And similarly, we should not place our trust in comparing fatality rates between states that have chosen to lift rules early and those that still have rules in place. Using the RCT method and lifting regulations randomly where it is safe would avoid falling into this trap.
Understanding which lockdown measures work is essential if we want to deal with a potential second wave of COVID-19 with more fine-grained policies that do not grind whole economies to a standstill. RCTs are the best way to identify such measures, and can help us optimally balance costs and benefits.
So why are policymakers reluctant to implement them more widely during this pandemic?
First, it would require them to admit that they do not yet know which policies work. Second, policymakers may worry that the public could perceive lifting rules randomly as unethical.
But in the current situation, where the optimal timing and extent of lifting or introducing different measures is unknown, such concerns may be largely unwarranted. Policymakers wouldn’t be knowingly withholding a beneficial intervention or introducing a harmful one if there are no existing reasons to believe that one policy is better than another.
Furthermore, if they follow a “phase-in” strategy, an intervention considered as safe would ultimately be applied to the whole population, first to the treatment group and then, after a short delay, to the control group too.
Johannes Lohse receives funding from the University of Birmingham, the British Academy and UKRI.