QUADAS-2: How Reviewers Judge Whether a Diagnostic Study Can Be Trusted

QUADAS-2 is the tool systematic reviewers use to judge how much a diagnostic accuracy study can be trusted. It works through four domains, patient selection, the index test, the reference standard, and the flow and timing of the study, rating each for risk of bias and, separately, for whether the study fits the review's question. Understanding it lets any reader see where a diagnostic study is most likely to mislead.

What the tool is for

When a systematic review pools diagnostic accuracy studies, it cannot treat them as equally believable. Some were built in ways that inflate accuracy, and others answer a slightly different question than the review is asking. QUADAS-2 is the structured tool that reviewers use to grade those risks, one study at a time, so the review's conclusions rest on the sturdier evidence.

It is not a numeric score to be summed. It is a set of judgments, each backed by reasons, about where a given study might have gone wrong. That design is deliberate, because collapsing quality into a single number tends to hide the specific flaw that matters.

The four domains

QUADAS-2 walks through four domains. Patient selection asks how participants were enrolled and whether the design, a case-control contrast for instance, could distort accuracy. The index test domain asks whether the test under study was interpreted without knowledge of the final diagnosis, since knowing the answer can nudge a reader's call.

The reference standard domain asks whether the standard used to confirm disease is accurate enough to count as truth, and whether it was interpreted blind to the index test. Flow and timing asks whether every patient got the same reference standard, whether the interval between tests was sensible, and whether anyone dropped out of the analysis in a way that could skew the result.

Bias and applicability are different questions

The tool separates two things that are easy to blur. Risk of bias asks whether the study's design and conduct could have produced a wrong answer to its own question. Applicability asks whether the study's patients, test, and reference standard match the question your review is actually asking.

A study can be beautifully conducted, low risk of bias throughout, and still be a poor fit, because it tested a different population or an older version of the test. Keeping these judgments apart stops a reader from dismissing a rigorous study for being off-topic, or from trusting an on-topic study that was poorly run.

Signalling questions and how a rating is reached

Each domain comes with signalling questions, plain yes, no, or unclear prompts that surface the specific features that create bias. Did the study avoid a case-control design? Was the index test threshold set in advance? Were the two tests interpreted independently? The reviewer answers these, then judges the domain as low, high, or unclear risk based on the pattern of answers.

Because the reasoning is written down, another reader can see why a study was downgraded and disagree if warranted. This transparency is the real value: the rating is an argument you can inspect, not a verdict handed down.

Using it as a reader, not just a reviewer

You do not need to be conducting a formal review to use this lens. When you read any single diagnostic study, walk its four domains in your head. Ask how patients were selected, whether the test was read blind, whether the reference standard deserves to be called truth, and whether everyone was accounted for at the end.

Pair that with the reporting side. Guidance such as STARD tells authors what to disclose, and the Cochrane guidance for diagnostic reviews shows how these appraisals feed a synthesis. Together they let you locate a study's weakest domain, which is usually where an impressive accuracy number quietly springs its leak.