New advances for decision-making in intervention optimization using MOST

What is it about?

Intervention optimization requires complex decision-making based on the results of an optimization trial. We propose a Bayesian method for this decision-making, and we test the new method in comparison to alternative strategies across very many (80,000) simulated optimization trials. We find that the new method, a posterior expected value approach, outperforms the alternatives on average.

Featured Image

Why is it important?

Improved decision-making performance in intervention optimization means that intervention scientists are more successful in identifying the interventions that have potential for high value. This can mean that better interventions are selected and advanced to subsequent stages of MOST. Furthermore, the results of this simulation study are exciting because the proposed method, a posterior expected value approach, opens new possibilities for intervention optimization (for example, by facilitating optimization based on multiple outcomes).