QUADAS 2 (Quality Assessment of Diagnostic Accuracy Studies)

QUADAS-2 is a structured quality assessment tool designed to evaluate the risk of bias and concerns about applicability in primary diagnostic accuracy studies included in systematic reviews. It assesses four key domains: patient selection, the index test, the reference standard, and flow and timing. Each domain is evaluated for risk of bias, and the first three are also assessed for applicability concerns.

Diagnostic accuracy studies are susceptible to specific biases that can inflate or deflate estimates of sensitivity, specificity, and predictive values. If a systematic review of diagnostic test accuracy does not assess the quality of included studies, the pooled estimates may reflect methodological artifacts rather than the true performance of the test. QUADAS-2 provides a standardized framework for this assessment, developed through expert consensus and piloting, enabling reviewers to identify which studies are at high risk of producing biased results and to interpret the overall evidence accordingly. The Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy (version 2.0, 2023) uses QUADAS-2 for this assessment. In February 2026 the developers published its successor, QUADAS-3, in Annals of Internal Medicine, and they now recommend QUADAS-3 for systematic reviews.

QUADAS-2 evaluates each included study across four domains using signaling questions that guide the assessor's judgment.

The patient selection domain examines whether patients were enrolled consecutively or randomly, whether the study avoided inappropriate exclusions, and whether the selection process could introduce spectrum bias (enrolling patients who are easier to diagnose). A case control design that compares clearly diseased patients against clearly healthy controls, for instance, tends to overestimate diagnostic accuracy.

The index test domain assesses whether the index test was interpreted without knowledge of the reference standard results (blinding) and whether the threshold for a positive result was prespecified. If the threshold is selected after seeing the data, the accuracy estimates may be optimistic.

The reference standard domain evaluates whether the reference standard is likely to correctly classify the target condition and whether it was interpreted without knowledge of the index test results. An imperfect reference standard (imperfect gold standard bias) can distort accuracy estimates in complex ways.

The flow and timing domain examines whether all patients received the same reference standard, whether all patients were included in the analysis, and whether the interval between the index test and the reference standard was appropriate. Differential verification bias occurs when different reference standards are used for positive versus negative index test results.

For each domain, the assessor classifies the risk of bias as low, high, or unclear. Applicability concerns are rated for the first three domains using the same scale. The results are typically presented in a summary table and can be visualized graphically.

QUADAS-3, published in February 2026, keeps four risk of bias domains with applicability concerns rated for the first three, but makes several changes. Its domains are participants, index test, target condition, and analysis, with the new analysis domain replacing flow and timing. Assessment starts by defining an ideal test accuracy trial for the review question, risk of bias is judged for each accuracy estimate rather than for the study as a whole, assessors record the rationale for each judgment, and the tool ends with a formal overall judgment of risk of bias and applicability.

Use QUADAS-2 when conducting a systematic review that includes diagnostic accuracy studies, whether as the primary focus or as a component of a broader review. It should be applied to every included study by at least two independent reviewers, with disagreements resolved through discussion or a third reviewer. For a new review, consider QUADAS-3, which its developers now recommend in place of QUADAS-2. QUADAS-2 is specific to diagnostic accuracy studies; for randomized trials, use the Cochrane Risk of Bias tool, and for observational studies, use tools such as the Newcastle Ottawa Scale.

A common error is applying QUADAS-2 mechanically without understanding the underlying biases each domain is designed to detect, leading to ratings that do not reflect the actual methodological quality of the study. Another mistake is using the original QUADAS checklist instead of a later version: QUADAS-2 improved upon the original by introducing structured signaling questions and separating risk of bias from applicability concerns, and QUADAS-3 (2026) now succeeds QUADAS-2. Reviewers sometimes rate a domain as "unclear" when insufficient information is reported, without attempting to contact the study authors for clarification. Finally, failing to incorporate QUADAS-2 results into the interpretation of the review findings, treating the quality assessment as a formality rather than a substantive input to the synthesis, limits the value of the assessment.

A systematic review evaluating the accuracy of point of care ultrasound for detecting pneumothorax includes 24 primary studies. Two reviewers independently apply QUADAS-2. In the patient selection domain, 8 studies are rated high risk because they used a case control design. In the index test domain, 5 studies are rated high risk because the ultrasound threshold was determined after seeing the reference standard results. In the reference standard domain, 3 studies are rated high risk because the reference standard (chest radiograph) is itself imperfect for small pneumothoraces. The review team presents these results in a traffic light table and conducts a sensitivity analysis excluding high risk studies, finding that the pooled sensitivity drops from 91% to 86%.