4 Integrating Evidence into Decisions
Learn how to assess evidence outcome by outcome, weigh benefits and harms alongside values and context, and communicate decisions that reflect uncertainty.
Assess evidence outcome by outcome
Start by specifying who is affected, which options are being compared, and which outcomes matter. Consider benefits, harms, burdens, and, when relevant, costs, equity, and practical feasibility.
Assess the separately for each important outcome. The considers risk of bias, inconsistency across studies, indirectness, imprecision, and publication bias. Its ratings—high, moderate, low, and very low—describe confidence in an effect estimate; they do not label an option as good or bad.
A large, well-run study may give precise evidence about a common outcome while providing little information about a rare harm. Evidence from a different population may also be indirect for the people facing the decision. Avoid assigning one undifferentiated quality label to the entire evidence base.
Compare effects and what matters
Report estimated effects in terms decision-makers can interpret. Absolute frequencies can make the expected change clearer—for example, about 15 fewer events per 1,000 people over one year—especially when reported alongside the comparison-group risk and the time period. Relative effects alone may obscure how much an outcome changes in practice.
Consider how the outcomes matter to those affected. A modest reduction in symptoms may be worthwhile to one person but not worth a substantial treatment burden to another. When preferences vary, there may be no single option that is best for everyone.
An considers the balance of desirable and undesirable effects together with values and preferences. It may also address resources, equity, acceptability, and feasibility.
Let uncertainty shape the choice
Uncertainty may arise from limited studies, imprecise estimates, differing findings, or evidence that does not closely match the decision context. Ask whether the plausible range of effects could change the preferred option. In particular, consider whether that range includes outcomes on both sides of a meaningful decision threshold, such as a benefit large enough to justify a serious adverse effect or a substantial cost.
Distinguish from . Research may estimate a treatment’s effects reasonably well while people disagree about whether those effects are worth the burden. Alternatively, people may agree on what matters but lack reliable estimates of the outcomes. Better evidence may reduce ; discussion and can help clarify individual priorities.
Uncertainty does not automatically mean doing nothing. A decision may still be needed, and uncertainty may support a , a choice tailored to individual preferences, monitoring, or further evidence gathering. GRADE guidance links stronger recommendations to clearer net benefits and greater confidence. or about how people value outcomes can support a .
Apply the reasoning to a conditional choice
Suppose, for illustration, an intervention is estimated to prevent 15 hospitalizations per 1,000 people over a year, while possibly causing a rare serious harm. The hospitalization estimate is uncertain, and evidence about the harm is sparse. Some people may prefer the intervention if their risk of hospitalization is high; others may decline it if they place greater weight on avoiding the potential harm.
A well-calibrated conclusion would not claim that the intervention is proven beneficial for everyone. It would state the estimated benefits and harms, explain which estimates are uncertain, identify factors that could change the balance, and make clear that individual preferences may reasonably lead to different choices. For a population-level decision, also explain relevant resource, equity, and implementation considerations.
Communicate the reasoning clearly
A concise explanation of a decision should state:
The decision and options being considered.
The main benefits and harms, preferably with absolute numbers and a time frame.
Confidence in the evidence, including important limitations for critical outcomes.
The trade-off or threshold that determines which option is favored.
What could change the conclusion, including different risks, preferences, or new evidence.
The action and its strength: a firm recommendation when the balance is clear, or a conditional choice when reasonable people may decide differently.
Use plain language, distinguish estimates from judgments, and avoid implying more certainty than the evidence supports. When communicating with individuals, personalize risk information where possible and invite them to consider what matters most. Numerical formats and visual aids can help make benefits and harms easier to compare.