Connect evidence to a decision
A test is most useful when the team can name the decision it would change and explain why that change is supported. You are allowed to value learning, too. It helps to say whether the goal is research, clinical care, or both.
Before you start: Clinical utility asks whether using information helps. Actionability keeps a measurement, a biological mechanism and a usable treatment option separate.
Where this step sits
This is the final lesson in judging a test or company claim. You have defined the claim and read its validation. Now ask what happens after each possible result.
Put the action before the test
It is easy to keep collecting information when the stakes are high. Each new result may feel like progress. A more helpful starting point is a sentence: “We are deciding whether to ____.”
Then ask how positive, negative, uncertain and failed results would change that decision. If every branch leads to the same care, the test may still have research value. Its immediate clinical value is harder to establish.
A plan also needs a time window. A careful result that arrives after the decision closes may not serve the intended purpose. For tissue tests, consider what the sample could support instead. Tissue consumed by one assay may be unavailable for another. These tradeoffs belong in the clinical team's shared plan.
Build a decision ledger
A short ledger makes the discussion easier to follow:
| Question | Example wording |
|---|---|
| Decision | Would this result change a named treatment or the choice of a trial? |
| Positive result | What exactly would happen, and what evidence supports it? |
| Negative result | Would care stay the same, or is treatment omission being proposed? |
| Uncertain or failed result | Would another sample or confirmatory test be needed? |
| Alternatives | Can existing pathology or another test answer the question? |
| Burden | What sample, time, travel, uncertainty and financial burden arise? |
| Owner | Who interprets the result and makes the clinical recommendation? |
| Stopping rule | What would make further testing unlikely to help this decision? |
The ledger does not calculate a treatment choice for you. It makes assumptions visible so the relevant people can assess them.
Worked example: a fictional protein test
A patient and clinician are discussing Drug A after surgery. A research assay reports whether a protein is abundant. The assay's biological rationale is plausible, but its treatment-selection evidence comes from a different cancer setting.
There are several reasonable learning branches. The team could ask whether an established clinical assay measures the same protein. It could inspect the exact Drug A evidence. It could consider a trial that tests the relevant strategy. Or it could decide the new test would consume tissue without changing the current choice.
What the result cannot do alone is turn protein abundance into proven postoperative benefit. Confirmation of the measurement answers “is the signal real?” Clinical evidence must still answer “does the proposed action help?”
If Drug A is approved for another indication, off-label use is a separate clinician-led question. A report that lists the drug does not guarantee access, payment or suitability.
Make room for research without disguising it as care
A research question can be valuable even when no treatment changes today. Examples include checking a mechanism, learning why two assays disagree, or generating a hypothesis for a future study.
State the goal honestly: “We want to learn whether this mechanism is present. The result will not independently select treatment.” Also name who will review it and what would count as a useful answer. That gives curiosity a direction and helps avoid an endless chain of follow-up tests.
When a company supplies unpublished performance data, record who supplied it and what was available to inspect. A statement that evidence is unavailable is different from a finding that the claim is false. Keep that distinction clear.
What can go wrong at this step
- An action changes, but benefit is assumed. A test-guided strategy can add unnecessary treatment as well as avoid it.
- A reassuring result supports untested omission. A favorable prognostic marker does not show that standard treatment can safely be skipped.
- A new test displaces a better-supported one. Consider sample opportunity cost and the named decision.
- A deadline substitutes for evidence. Practical urgency matters, but it does not validate the proposed action.
Try it
A fictional risk test will report high or low recurrence risk. The clinician says neither result would change the recommended treatment, and no study is open that uses the result. What should the reader conclude?
Answer: The immediate treatment value is limited by that decision plan. The test might still serve a clearly defined research or personal-information goal, but that goal should be weighed against burden, uncertainty and alternatives. “More information” is not automatically better care.
Explain it back
Complete: “Before ordering the test, we will decide what to do if the result is ____, ____ or ____.”
One possible answer: “Positive, negative or uninformative. We will name the evidence and clinical owner for each branch.”
Takeaway
Connect every test to a named decision, supported actions and a plan for uncertain results.
Next: Use the claim worksheet on a real report with its clinical or research owner. For a proposed investigational treatment, continue to accessing experimental therapy.
Sources and scope
Source check: October 9, 2026. The ledger and cases are teaching exercises, not a recommendation to order or omit a test. Expert and learner review remain pending.
- MedlinePlus Genetics: valid and useful testing — useful information is a separate test-quality question.
- FDA: companion diagnostics — evidence and intended use for treatment-linked testing.
- FDA: understanding off-label use — approval boundaries for a proposed medicine.