A diagnostic study compares an index test with a reference standard. To interpret the comparison, the reader needs to know what information each assessor had and how the reference classification was reached. Calling a comparator a “gold standard” does not supply those details.
Name the reference process
Was the reference a laboratory assay, an imaging assessment, a clinical follow-up rule or a panel judgement? Record its timing and decision rule. STARD's explanation asks authors to describe information available to interpreters of both the index test and the reference standard. STARD explanation.
Draw the information path
Imagine a fictional imaging study. Reader A interprets the new image; reader B assigns the reference classification using follow-up information. If reader B also sees reader A's result, the two assessments are not informationally independent. That does not establish the size or direction of bias, but it changes what the observed agreement can demonstrate.
Draw two boxes and list every input to each. A shared specimen is not the same issue as a shared interpretation. A common clinical history may be intentional, while knowledge of the other test result may conflict with the intended comparison.
Look for incorporation
Suppose the reference definition includes a positive result from the new test as one of its criteria. The test is then partly being compared with a rule that contains itself. A very high agreement could reflect that construction. The paper should explain why the reference was chosen and how such dependence was handled.
This example is about the logic of a comparison. It does not claim that every composite reference standard is invalid or that an alternative definitive test always exists.
Account for time between assessments
If the index test and reference assessment are months apart, ask whether the underlying state could have changed. Record interventions or events occurring between them where the report provides that information. Otherwise disagreement may combine measurement differences with change over time, and the table alone cannot separate them.
Separate reporting from judging quality
“Blinding not described” is an uncertainty in the report; it is not proof that blinding was absent. Similarly, a blinded reader can still use an imperfect reference method. Keep these as separate notes: reference validity, information available, timing and missing assessments. Together they make a more useful appraisal than a single gold-standard label. See our diagnostic study overview for the wider evidence chain.
Top comments (0)