Most health decisions involve trade-offs between harms (such as false positives or overdiagnosis) and benefits (such as cancer deaths prevented). Individuals making decisions, or policy makers developing guidelines, are often faced with uncertainty about the magnitude of both harms and benefits for a variety of reasons including wide confidence intervals, biased study designs, and lack of data applicable to a specific setting. In economic analysis, value of information (VOI) is an explicit approach used to estimate the degree to which reducing uncertainty about a decision reduces the chances of making a "wrong" decision, and whether obtaining additional information through research is worthwhile. This approach can be extended to clinical decisions using harm-benefit trade-offs. One of the key insights of this approach is that the chances of making the "wrong" decision, or the "wrong" recommendation, depend on how much one is willing to "pay" for a given benefit. In economic analysis, this is depicted using cost-effectiveness acceptability curves, which show how the likelihood that a given choice is optimal varies as the willingness to pay (for example, in euros per quality-adjusted life year) varies. In a harm-benefit analysis, the curve depicts how the optimal decision changes with the number of harms one is willling to accept in order to gain one benefit.
Simple models of screening incorporating esetimates of sensitivity, specificity, and disease prevalence illustrate how imprecision in estimates of these characteristics affects the harm-benefit ratio in terms of false-positives per true disease detected, and how the optimal test strategy changes as willingness to "pay" for disease detection with false-positives increases. Similar results can be shown with more complex models. By isolating individual components of the decision, the relative reduction in uncertainty by obtaining more precise or less biased estimates can be determined, and used to prioritize additional research.
Arguments about screening recommendations frequently focus on the estimates of harm or benefit, rather than on the inherent trade-off between the two. Using harm-benefit curves has the potential to facilitate evidence-based guidelines by focusing discussion on how the uncertainty in current estimates of harm and benefit affects optimal decisions at different thresholds of acceptable harm. Frank discussions between patients, clinicians, and other stakeholders about what that threshold should be can help make guidelines more transparent, and help identify priorities for future research.