Randomized controlled trials are the cornerstone of medical research, but they have been less commonly used to study how behavioral changes can improve health. The pandemic has shown that such large-scale trials can be conducted — for example, to show the efficacy of mask wearing.
One such study tested interventions to promote mask wearing across 600 villages in rural Bangladesh. As well as measuring efficacy, the team wanted to test interventions that could be easily adapted and scaled in other countries.
A range of interventions were tested, each of which was supported by previous research and can work in certain circumstances, says behavioral scientist Neela Saldanha at the Yale Research Initiative on Innovation and Scale, who was involved with the research. Mask wearing tripled in the treatment villages, compared with its use in untreated villages, but some interventions had no effect, including text reminders and financial and non-financial incentives.
This research showed that you could conduct a highly ambitious, large-scale study within a short time frame. However, it requires a different way of working, says Saldanha.
“I think the challenge is to do it at scale in a community setting, because unfortunately in the pandemic it’s not just an individual behavior change. It’s a total community behavior change,” says Saldanha. The worldwide team had to conduct the study in parallel, implementing interventions in some villages in Bangladesh while simultaneously using different interventions at other study sites in the country.
There are benefits to putting in the resources to test multiple interventions in parallel, either in the same country or across multiple countries, instead of testing interventions in succession. But this requires coordination and financial support. Funding to fight SARS-CoV-2 has made large-scale behavioral studies such as this possible.
“If you’re in the fortunate position to have resources to examine multiple alternatives, I think two things: one is we know about sunk costs, or at what point do we abandon an alternative,” says Saldanha. A ‘sunk cost’ represents the time and resources lost in trying an intervention that proves to be unsuccessful. “And the second one is, how do we make sure we are not overestimating or cherry picking at the end of the analysis.”
Randomized trials of behavioral interventions require parallel processing, taking risks, working in teams, and using the time pressure in a positive way. However, the results may not always be transferable to other contexts. There is no guarantee that interventions that work in one country will work in others, but the results can inform decision-making about which interventions could receive the most focus and what others could be worth trying in a new context, says Saldanha.
Although it can be a boon to be able to conduct an ambitious study in a short time, it is not without caveats. There is a danger in trying every possible intervention when you have a large sample size and the resources. Saldanha notes that for each behavioral intervention, there should be a clear reason for its inclusion and an explanation of what results are expected. And if an intervention does not seem to work, researchers should think about why that might be.
Trials of behavioral interventions are being carried out in other fields of health and medicine — and bigger seems to be better. A recent ‘megastudy’ of interventions to promote physical exercise involved more than 60,000 participants and was run by 30 scientists at 15 institutions. Of 54 different 4-week programs, 45% of the interventions increased visits to the gym, but only 8% led to significant and measurable behavior change after the 4-week intervention ended. Notably, impartial experts failed to predict which interventions would work, which shows the importance of randomized trials.