Why English? | Reading QUADAS-3 Diagnostic Accuracy Judgements Through Participants, Index Tests, Target Conditions and Analysis

People searching for QUADAS-3, diagnostic test accuracy risk of bias or how to appraise sensitivity and specificity need current terminology. English matters because the index test is not the target condition, a threshold is not a diagnosis, and an accuracy estimate from one carefully selected subgroup may not answer the review’s real clinical question.

The official University of Bristol QUADAS site, checked on 5 November 2026, identifies QUADAS-3 as the current recommended tool and states that it supersedes QUADAS-2. The official tool page records publication in February 2026 and describes six phases. Version 1.2 assesses risk of bias in Participants, Index test, Target condition and Analysis; the first three domains are also assessed for applicability to the synthesis question.

QUADAS-3 evaluates bias and applicability of selected diagnostic-accuracy estimates in systematic reviews. It does not prove that a test improves patient outcomes, is cost-effective or is the correct choice for an individual. This article owns the current appraisal intent. The existing STARD 2015 article owns reporting completeness, while future comparative-test questions may require the separate QUADAS-C extension.

• State the synthesis question and ideal test-accuracy trial.
• Draw study flow and select the exact estimates to judge.
• Assess four risk-of-bias domains with evidence-linked signalling answers.
• Judge applicability for participants, index test and target condition.
• Communicate the overall judgement without turning accuracy into guaranteed benefit.

---

Use the Current QUADAS Version

The University of Bristol identifies QUADAS-3 as current and says QUADAS-2 has been superseded.

Appraisal move for Use the Current QUADAS Version

Rewrite the synthesis question as a test pathway: who presents, where testing occurs, which version and threshold are used, what target condition matters and what role the test is intended to play. Keep **Use the Current QUADAS Version** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

A test evaluated only in clear severe cases and healthy controls may separate groups impressively but perform differently among consecutive patients with uncertain symptoms. Recruitment affects both bias and real-world relevance. For **use the current quadas version**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **use the current quadas version** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Record Version 1.2 and the Review Date

Tool versions can change; the official resources page listed version 1.2 when checked on 5 November 2026.

Appraisal move for Record Version 1.2 and the Review Date

Describe the ideal test-accuracy trial before looking at shortcomings in the real study. The ideal provides a consistent comparator and makes applicability assumptions explicit rather than intuitive. Keep **Record Version 1.2 and the Review Date** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If investigators choose the most favourable threshold after examining the same data, the reported sensitivity and specificity may be optimistic. A prospectively defined threshold answers a different evidential question. For **record version 1.2 and the review date**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **record version 1.2 and the review date** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Phase 1: State Every Synthesis Question

Define participants, index test, target condition, threshold and test role for each planned synthesis.

Appraisal move for Phase 1: State Every Synthesis Question

Select the exact accuracy estimate first. Record subgroup, threshold, index-test interpretation, target-condition method and analysis population so one study’s multiple estimates are not treated as interchangeable. Keep **Phase 1: State Every Synthesis Question** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

When index-test readers know the target-condition result, a subjective interpretation can shift toward agreement. The report should state information flow, not merely say blinded without naming who knew what. For **phase 1: state every synthesis question**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **phase 1: state every synthesis question** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Phase 2: Define the Ideal Test-Accuracy Trial

Describe the design that would answer the synthesis question without material bias and with direct applicability.

Appraisal move for Phase 2: Define the Ideal Test-Accuracy Trial

Draw participant flow with counts at recruitment, index testing, target-condition assessment and analysis. Mark indeterminate results, withdrawals and alternative verification routes rather than hiding them in footnotes. Keep **Phase 2: Define the Ideal Test-Accuracy Trial** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If positive index tests receive an intensive reference procedure while negative tests receive follow-up only, target status may be classified differently by index result. The appraisal must trace both verification routes. For **phase 2: define the ideal test-accuracy trial**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **phase 2: define the ideal test-accuracy trial** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Phase 3: Draw the Study Flow Diagram

Show recruitment, index testing, target-condition assessment, exclusions and analysis populations.

Appraisal move for Phase 3: Draw the Study Flow Diagram

Answer signalling questions with a source location and a mechanism. ‘High risk because recruitment was selective’ is stronger when it explains which patients were excluded and how spectrum could alter accuracy. Keep **Phase 3: Draw the Study Flow Diagram** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

Excluding indeterminate index results can create a neat two-by-two table while removing exactly the difficult cases seen in practice. Count and characterise those results before judging the estimate. For **phase 3: draw the study flow diagram**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **phase 3: draw the study flow diagram** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Phase 4: Select Accuracy Estimates

Identify the exact threshold, subgroup and estimate entering appraisal rather than rating a study as one object.

Appraisal move for Phase 4: Select Accuracy Estimates

Write bias and applicability in separate columns. A carefully conducted estimate can be indirect for the synthesis question, while a highly applicable sample can still be analysed in a biased way. Keep **Phase 4: Select Accuracy Estimates** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

The same study may report one estimate at a screening threshold and another at a confirmatory threshold. QUADAS-3 keeps the selected estimate tied to its intended role and synthesis question. For **phase 4: select accuracy estimates**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **phase 4: select accuracy estimates** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Phase 5: Assess Four Bias Domains

Judge participants, index test, target condition and analysis with signalling questions and evidence-linked reasons.

Appraisal move for Phase 5: Assess Four Bias Domains

Prepare the overall judgement in words before assigning a display colour. The rationale should identify the decisive domain, the estimate affected and what a better study would do differently. Keep **Phase 5: Assess Four Bias Domains** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

For a student reading a headline accuracy claim, ask five questions: who was tested, which test version, which threshold, how target status was decided and who disappeared before analysis. For **phase 5: assess four bias domains**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **phase 5: assess four bias domains** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Phase 6: Make the Overall Judgement

Combine domain information according to the tool while preserving the reasoning for each selected estimate.

Appraisal move for Phase 6: Make the Overall Judgement

Rewrite the synthesis question as a test pathway: who presents, where testing occurs, which version and threshold are used, what target condition matters and what role the test is intended to play. Keep **Phase 6: Make the Overall Judgement** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

A test evaluated only in clear severe cases and healthy controls may separate groups impressively but perform differently among consecutive patients with uncertain symptoms. Recruitment affects both bias and real-world relevance. For **phase 6: make the overall judgement**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **phase 6: make the overall judgement** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 1: Define the Participant Spectrum

Symptoms, setting, previous testing and disease severity influence both accuracy and applicability.

Appraisal move for Domain 1: Define the Participant Spectrum

Describe the ideal test-accuracy trial before looking at shortcomings in the real study. The ideal provides a consistent comparator and makes applicability assumptions explicit rather than intuitive. Keep **Domain 1: Define the Participant Spectrum** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If investigators choose the most favourable threshold after examining the same data, the reported sensitivity and specificity may be optimistic. A prospectively defined threshold answers a different evidential question. For **domain 1: define the participant spectrum**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 1: define the participant spectrum** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 1: Examine Recruitment

Consecutive or random sampling generally provides a clearer clinical spectrum than a selectively assembled sample.

Appraisal move for Domain 1: Examine Recruitment

Select the exact accuracy estimate first. Record subgroup, threshold, index-test interpretation, target-condition method and analysis population so one study’s multiple estimates are not treated as interchangeable. Keep **Domain 1: Examine Recruitment** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

When index-test readers know the target-condition result, a subjective interpretation can shift toward agreement. The report should state information flow, not merely say blinded without naming who knew what. For **domain 1: examine recruitment**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 1: examine recruitment** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 1: Avoid Distorting Case-Control Sampling

Artificially clear cases and controls can exaggerate apparent separation between target-condition states.

Appraisal move for Domain 1: Avoid Distorting Case-Control Sampling

Draw participant flow with counts at recruitment, index testing, target-condition assessment and analysis. Mark indeterminate results, withdrawals and alternative verification routes rather than hiding them in footnotes. Keep **Domain 1: Avoid Distorting Case-Control Sampling** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If positive index tests receive an intensive reference procedure while negative tests receive follow-up only, target status may be classified differently by index result. The appraisal must trace both verification routes. For **domain 1: avoid distorting case-control sampling**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 1: avoid distorting case-control sampling** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 1: Inspect Exclusions

Removing difficult, indeterminate or comorbid participants may create an easier sample than practice.

Appraisal move for Domain 1: Inspect Exclusions

Answer signalling questions with a source location and a mechanism. ‘High risk because recruitment was selective’ is stronger when it explains which patients were excluded and how spectrum could alter accuracy. Keep **Domain 1: Inspect Exclusions** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

Excluding indeterminate index results can create a neat two-by-two table while removing exactly the difficult cases seen in practice. Count and characterise those results before judging the estimate. For **domain 1: inspect exclusions**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 1: inspect exclusions** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 2: Describe the Index Test

Report device, version, specimen, operator, procedure, interpretation and intended role.

Appraisal move for Domain 2: Describe the Index Test

Write bias and applicability in separate columns. A carefully conducted estimate can be indirect for the synthesis question, while a highly applicable sample can still be analysed in a biased way. Keep **Domain 2: Describe the Index Test** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

The same study may report one estimate at a screening threshold and another at a confirmatory threshold. QUADAS-3 keeps the selected estimate tied to its intended role and synthesis question. For **domain 2: describe the index test**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 2: describe the index test** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 2: Prespecify Thresholds

A threshold selected after seeing the data can make accuracy look more favourable than a prospectively defined threshold.

Appraisal move for Domain 2: Prespecify Thresholds

Prepare the overall judgement in words before assigning a display colour. The rationale should identify the decisive domain, the estimate affected and what a better study would do differently. Keep **Domain 2: Prespecify Thresholds** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

For a student reading a headline accuracy claim, ask five questions: who was tested, which test version, which threshold, how target status was decided and who disappeared before analysis. For **domain 2: prespecify thresholds**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 2: prespecify thresholds** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 2: Check Information Available to Interpreters

Knowledge of target-condition information can influence subjective index-test interpretation.

Appraisal move for Domain 2: Check Information Available to Interpreters

Rewrite the synthesis question as a test pathway: who presents, where testing occurs, which version and threshold are used, what target condition matters and what role the test is intended to play. Keep **Domain 2: Check Information Available to Interpreters** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

A test evaluated only in clear severe cases and healthy controls may separate groups impressively but perform differently among consecutive patients with uncertain symptoms. Recruitment affects both bias and real-world relevance. For **domain 2: check information available to interpreters**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 2: check information available to interpreters** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 3: Define the Target Condition

The condition, state or outcome the test aims to detect must match the synthesis question.

Appraisal move for Domain 3: Define the Target Condition

Describe the ideal test-accuracy trial before looking at shortcomings in the real study. The ideal provides a consistent comparator and makes applicability assumptions explicit rather than intuitive. Keep **Domain 3: Define the Target Condition** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If investigators choose the most favourable threshold after examining the same data, the reported sensitivity and specificity may be optimistic. A prospectively defined threshold answers a different evidential question. For **domain 3: define the target condition**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 3: define the target condition** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 3: Evaluate the Reference Approach

The method used to classify target-condition status should be appropriate and applied comparably.

Appraisal move for Domain 3: Evaluate the Reference Approach

Select the exact accuracy estimate first. Record subgroup, threshold, index-test interpretation, target-condition method and analysis population so one study’s multiple estimates are not treated as interchangeable. Keep **Domain 3: Evaluate the Reference Approach** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

When index-test readers know the target-condition result, a subjective interpretation can shift toward agreement. The report should state information flow, not merely say blinded without naming who knew what. For **domain 3: evaluate the reference approach**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 3: evaluate the reference approach** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 3: Examine Incorporation and Review Bias

Using the index test inside the target-condition decision or interpreting one with knowledge of the other can distort estimates.

Appraisal move for Domain 3: Examine Incorporation and Review Bias

Draw participant flow with counts at recruitment, index testing, target-condition assessment and analysis. Mark indeterminate results, withdrawals and alternative verification routes rather than hiding them in footnotes. Keep **Domain 3: Examine Incorporation and Review Bias** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If positive index tests receive an intensive reference procedure while negative tests receive follow-up only, target status may be classified differently by index result. The appraisal must trace both verification routes. For **domain 3: examine incorporation and review bias**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 3: examine incorporation and review bias** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 4: Follow Participants Into Analysis

Account for every selected participant, every test result and every exclusion from the two-by-two table or model.

Appraisal move for Domain 4: Follow Participants Into Analysis

Answer signalling questions with a source location and a mechanism. ‘High risk because recruitment was selective’ is stronger when it explains which patients were excluded and how spectrum could alter accuracy. Keep **Domain 4: Follow Participants Into Analysis** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

Excluding indeterminate index results can create a neat two-by-two table while removing exactly the difficult cases seen in practice. Count and characterise those results before judging the estimate. For **domain 4: follow participants into analysis**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 4: follow participants into analysis** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 4: Examine Timing

The interval between index test and target-condition assessment should not allow the condition to change materially.

Appraisal move for Domain 4: Examine Timing

Write bias and applicability in separate columns. A carefully conducted estimate can be indirect for the synthesis question, while a highly applicable sample can still be analysed in a biased way. Keep **Domain 4: Examine Timing** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

The same study may report one estimate at a screening threshold and another at a confirmatory threshold. QUADAS-3 keeps the selected estimate tied to its intended role and synthesis question. For **domain 4: examine timing**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 4: examine timing** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 4: Handle Indeterminate Results

Excluding uninterpretable or borderline results can overstate accuracy and reduce real-world relevance.

Appraisal move for Domain 4: Handle Indeterminate Results

Prepare the overall judgement in words before assigning a display colour. The rationale should identify the decisive domain, the estimate affected and what a better study would do differently. Keep **Domain 4: Handle Indeterminate Results** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

For a student reading a headline accuracy claim, ask five questions: who was tested, which test version, which threshold, how target status was decided and who disappeared before analysis. For **domain 4: handle indeterminate results**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 4: handle indeterminate results** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Domain 4: Check Verification Patterns

Participants should not receive different reference approaches merely because of their index-test result unless the design handles this appropriately.

Appraisal move for Domain 4: Check Verification Patterns

Rewrite the synthesis question as a test pathway: who presents, where testing occurs, which version and threshold are used, what target condition matters and what role the test is intended to play. Keep **Domain 4: Check Verification Patterns** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

A test evaluated only in clear severe cases and healthy controls may separate groups impressively but perform differently among consecutive patients with uncertain symptoms. Recruitment affects both bias and real-world relevance. For **domain 4: check verification patterns**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **domain 4: check verification patterns** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Judge Applicability in the First Three Domains

Participants, index test and target condition are compared with the synthesis question as well as assessed for bias.

Appraisal move for Judge Applicability in the First Three Domains

Describe the ideal test-accuracy trial before looking at shortcomings in the real study. The ideal provides a consistent comparator and makes applicability assumptions explicit rather than intuitive. Keep **Judge Applicability in the First Three Domains** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If investigators choose the most favourable threshold after examining the same data, the reported sensitivity and specificity may be optimistic. A prospectively defined threshold answers a different evidential question. For **judge applicability in the first three domains**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **judge applicability in the first three domains** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Use Signalling Questions as Evidence Prompts

A signalling answer should point to a report location or explicit lack of information, not a reviewer impression.

Appraisal move for Use Signalling Questions as Evidence Prompts

Select the exact accuracy estimate first. Record subgroup, threshold, index-test interpretation, target-condition method and analysis population so one study’s multiple estimates are not treated as interchangeable. Keep **Use Signalling Questions as Evidence Prompts** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

When index-test readers know the target-condition result, a subjective interpretation can shift toward agreement. The report should state information flow, not merely say blinded without naming who knew what. For **use signalling questions as evidence prompts**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **use signalling questions as evidence prompts** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Distinguish No Information From Low Risk

Silence about recruitment, thresholds or analysis should not be converted into reassurance.

Appraisal move for Distinguish No Information From Low Risk

Draw participant flow with counts at recruitment, index testing, target-condition assessment and analysis. Mark indeterminate results, withdrawals and alternative verification routes rather than hiding them in footnotes. Keep **Distinguish No Information From Low Risk** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If positive index tests receive an intensive reference procedure while negative tests receive follow-up only, target status may be classified differently by index result. The appraisal must trace both verification routes. For **distinguish no information from low risk**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **distinguish no information from low risk** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Handle Multiple Thresholds and Subgroups

One study can contribute estimates with different bias and applicability profiles.

Appraisal move for Handle Multiple Thresholds and Subgroups

Answer signalling questions with a source location and a mechanism. ‘High risk because recruitment was selective’ is stronger when it explains which patients were excluded and how spectrum could alter accuracy. Keep **Handle Multiple Thresholds and Subgroups** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

Excluding indeterminate index results can create a neat two-by-two table while removing exactly the difficult cases seen in practice. Count and characterise those results before judging the estimate. For **handle multiple thresholds and subgroups**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **handle multiple thresholds and subgroups** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Explain Any Algorithm Override

A proposed judgement supports consistency, while a justified departure requires a transparent rationale.

Appraisal move for Explain Any Algorithm Override

Write bias and applicability in separate columns. A carefully conducted estimate can be indirect for the synthesis question, while a highly applicable sample can still be analysed in a biased way. Keep **Explain Any Algorithm Override** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

The same study may report one estimate at a screening threshold and another at a confirmatory threshold. QUADAS-3 keeps the selected estimate tied to its intended role and synthesis question. For **explain any algorithm override**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **explain any algorithm override** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Did You Know? QUADAS-3 Has Six Phases

The current structure moves from synthesis questions and an ideal trial to estimate selection, four domains and an overall judgement.

Appraisal move for Did You Know? QUADAS-3 Has Six Phases

Prepare the overall judgement in words before assigning a display colour. The rationale should identify the decisive domain, the estimate affected and what a better study would do differently. Keep **Did You Know? QUADAS-3 Has Six Phases** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

For a student reading a headline accuracy claim, ask five questions: who was tested, which test version, which threshold, how target status was decided and who disappeared before analysis. For **did you know? quadas-3 has six phases**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **did you know? quadas-3 has six phases** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Did You Know? QUADAS-3 Changed the Unit of Assessment

The current tool assesses bias and applicability at the estimate level rather than only at the study level.

Appraisal move for Did You Know? QUADAS-3 Changed the Unit of Assessment

Rewrite the synthesis question as a test pathway: who presents, where testing occurs, which version and threshold are used, what target condition matters and what role the test is intended to play. Keep **Did You Know? QUADAS-3 Changed the Unit of Assessment** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

A test evaluated only in clear severe cases and healthy controls may separate groups impressively but perform differently among consecutive patients with uncertain symptoms. Recruitment affects both bias and real-world relevance. For **did you know? quadas-3 changed the unit of assessment**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **did you know? quadas-3 changed the unit of assessment** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Keep QUADAS-3 Separate From STARD

STARD supports complete diagnostic-accuracy reporting; QUADAS-3 judges bias and applicability of estimates in a review.

Appraisal move for Keep QUADAS-3 Separate From STARD

Describe the ideal test-accuracy trial before looking at shortcomings in the real study. The ideal provides a consistent comparator and makes applicability assumptions explicit rather than intuitive. Keep **Keep QUADAS-3 Separate From STARD** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If investigators choose the most favourable threshold after examining the same data, the reported sensitivity and specificity may be optimistic. A prospectively defined threshold answers a different evidential question. For **keep quadas-3 separate from stard**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **keep quadas-3 separate from stard** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Do Not Infer Patient Benefit From Accuracy Alone

Sensitivity and specificity do not by themselves prove that testing improves outcomes or is the best local choice.

Appraisal move for Do Not Infer Patient Benefit From Accuracy Alone

Select the exact accuracy estimate first. Record subgroup, threshold, index-test interpretation, target-condition method and analysis population so one study’s multiple estimates are not treated as interchangeable. Keep **Do Not Infer Patient Benefit From Accuracy Alone** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

When index-test readers know the target-condition result, a subjective interpretation can shift toward agreement. The report should state information flow, not merely say blinded without naming who knew what. For **do not infer patient benefit from accuracy alone**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **do not infer patient benefit from accuracy alone** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Help Students Read Test Claims

Ask who was tested, how status was established, which threshold was used and which participants disappeared before analysis.

Appraisal move for Help Students Read Test Claims

Draw participant flow with counts at recruitment, index testing, target-condition assessment and analysis. Mark indeterminate results, withdrawals and alternative verification routes rather than hiding them in footnotes. Keep **Help Students Read Test Claims** tied to the selected estimate rather than giving the whole paper a single reputation.

Diagnostic-accuracy example

If positive index tests receive an intensive reference procedure while negative tests receive follow-up only, target status may be classified differently by index result. The appraisal must trace both verification routes. For **help students read test claims**, compare the observed design with the ideal test-accuracy trial and state the plausible effect on the estimate or its applicability.

Interpretation boundary

QUADAS-3 uses **help students read test claims** to inform an evidence-linked judgement about bias or applicability. It does not prove patient benefit, affordability or individual suitability. Preserve the signalling answers, proposed judgement, any override and the short rationale another reviewer would need to audit the decision.

---

Frequently Asked Questions

What is QUADAS-3?

The current Quality Assessment of Diagnostic Accuracy Studies tool for risk-of-bias and applicability appraisal in systematic reviews.

Does QUADAS-3 replace QUADAS-2?

Yes. The official Bristol site says QUADAS-2 has been superseded.

Which version was current on 5 November 2026?

The official resources page listed version 1.2.

How many phases are there?

Six: synthesis questions, ideal trial, flow diagram, estimate selection, domain assessment and overall judgement.

What are the four risk-of-bias domains?

Participants, Index test, Target condition and Analysis.

Which domains also receive applicability judgements?

Participants, Index test and Target condition.

Why is appraisal at the estimate level important?

A study can report several thresholds, subgroups or estimates with different risks and applicability.

How is QUADAS-3 different from STARD?

STARD is reporting guidance; QUADAS-3 is a risk-of-bias and applicability tool.

Does high diagnostic accuracy prove patient benefit?

No. Accuracy alone does not establish improved outcomes, feasibility, affordability or individual suitability.

What should students ask first?

Who was tested, what the index test was, how target status was established, which threshold was used and who reached analysis.

---

Useful Next Reading

• Start with the official QUADAS-3 overview.
• Use the current six-phase QUADAS-3 tool page.
• Find version 1.2, guidance and worked examples.
• Read diagnostic-accuracy reports with STARD 2015.
• Practise critical reading of claims, evidence and bias.

Closing thought: QUADAS-3 makes an accuracy estimate auditable by anchoring every judgement to a synthesis question, an ideal trial, a real participant flow and the exact estimate under review.