Evidence and decisions

Applying research evidence to an individual decision

Connect research findings with baseline risk, practical circumstances and informed preferences without overstating certainty.

A research result describes what happened in a study population under particular conditions. Applying that result requires another set of questions. How similar is the intended setting? What are the available options? Which outcomes matter to the person making the decision?

Good interpretation begins by separating what the research estimates from the judgement needed to use it. An average effect can inform a discussion without determining the same answer for everyone. This guide concerns that interpretation process, rather than recommending a particular test or treatment.

Check whether the evidence fits

Identify the study population, the intervention or exposure, the comparison, the outcomes and the follow-up period. Compare these with the question in front of you. Differences in age, baseline risk, coexisting conditions, delivery or available support may affect applicability.

A specialist service with frequent follow-up may produce results that are difficult to reproduce in a setting with less support. A short-term symptom outcome may not answer a question about long-term function. Such differences should be described specifically, rather than hidden behind a general statement that a study is relevant.

First check that the study is sufficiently trustworthy. Applicability cannot rescue an estimate distorted by serious bias. Equally, a well-conducted study can still be only indirectly relevant to a different decision.

Translate relative effects into absolute numbers

Suppose a fictional study estimates a relative risk of 0.80 for an undesirable event. If a comparable group would otherwise have 100 events per 1,000 people over a year, applying that relative effect gives 80 events per 1,000: an estimated reduction of 20.

If the baseline risk were instead 10 events per 1,000 over the same period, the corresponding estimate would be 8 per 1,000: a reduction of 2. The relative reduction is 20% in both examples, while the absolute difference is very different.

These calculations illustrate an assumption. They do not establish that the relative effect stays constant across populations. A reasonable application needs a suitable baseline risk and evidence supporting the transfer of the effect estimate. Preserve the follow-up period and communicate uncertainty around both ingredients.

Compare the consequences that matter

Describe benefits, harms, burdens and practical demands using comparable numbers where possible. Keep denominators and time periods consistent. A risk per 100 people over a month cannot be directly compared with another risk per 1,000 over ten years.

Natural frequencies can help make the discussion concrete. It can also help to state both sides of the same outcome: how many people experience an event and how many do not. Avoid presenting a best-case benefit beside a worst-case harm without explaining the different assumptions.

Some consequences resist a neat numerical summary. Travel, recovery time, uncertainty, inconvenience and effects on work or caring responsibilities may still matter greatly. Their importance should come from the person's circumstances, not from an assumption that everyone values them equally.

Make space for preferences and questions

Shared decision making combines relevant evidence with a person's informed preferences and the professional expertise needed to interpret the options. NICE recommends explaining available choices, including appropriate options to make no change, and checking understanding throughout the discussion.

The three-talk model offers a practical teaching structure: establish a supportive partnership, compare options and discuss what matters when reaching a decision. Real conversations can move backwards and forwards through those tasks. A decision aid can organise information, but cannot replace the discussion itself.

Ask what outcomes the person most wants to improve or avoid. Explain where evidence is uncertain, and check understanding by inviting the person to describe the options in their own words. Acknowledging uncertainty is more helpful than implying that an average estimate predicts an individual's future exactly.

Agree what happens next

A decision may include a review point, a plan to gather missing information or time to consider the alternatives. Record the rationale and any important assumptions. If circumstances change, the balance may change as well.

When summarising the evidence, retain its limits. State the expected absolute effects, the main uncertainty and the reasons the evidence may or may not transfer to the current situation. That gives a decision a clearer foundation than a headline claim about which option performed best.

Methodology references

  • NICE NG197: Shared decision making, recommendations
  • Elwyn et al. A three-talk model for shared decision making
  • Cochrane: Decision aids for people facing health treatment or screening decisions (2024)