When do we know enough to recommend action? The need to be bold but not reckless

Researchers’ scientific training instils the need to recognise and explicitly acknowledge the limitations of their findings as the basis for policy recommendations. It is a matter of ethics (being truthful) and our reputations as scientists. When we present our results, we therefore take pains to state the caveats, such as potential biases, lack of statistical significance and uncertain generalisability, which could alter conclusions. “On the one hand this, but on the other hand that” rarely provides guidance for practical decisions, however, and policymakers generally tune this out.

My colleagues and I were recently commissioned to make recommendations regarding a large public programme targeting particular health inequalities, which were not narrowing despite years of programme efforts. Our task was to …