7.1 STARD 2015, CARE, and Specialized Reporting Guidelines

STARD 2015, CARE, and Specialized Reporting Guidelines

Reporting guidelines help authors make methods and findings understandable; they do not repair a weak study or impose one universal method. The publication professional should select the guideline that matches the design, map its items to the manuscript, and describe what actually happened without turning checklist language into invented procedures.

STARD 2015 for diagnostic accuracy

The Standards for Reporting Diagnostic Accuracy Studies (STARD) 2015 statement contains 30 essential items. A diagnostic-accuracy report typically compares an index test with a reference standard in people suspected of having a target condition. Readers need enough information to judge spectrum effects, selection, verification, interpretation, and applicability.

Key elements include:

• identify the work as a diagnostic-accuracy study and state the intended use and clinical role of the index test;
• describe eligibility, setting, participant identification, sampling, and whether data collection was prospective or retrospective;
• define the index test, reference standard, rationale, cutoffs, and whether thresholds were prespecified or exploratory;
• state whether performers or readers of the index test had access to clinical information and reference-standard results, and vice versa;
• explain handling of missing, indeterminate, and outlier results;
• report participant flow, time interval and interventions between tests, and adverse events from testing;
• provide a cross-tabulation of index-test results against reference-standard results, or their distribution for continuous tests;
• report estimates of accuracy with precision and distinguish prespecified from exploratory analyses.

STARD asks authors to provide a participant flow diagram. Calling the diagram “mandatory” in every legal or editorial sense is too broad: journals determine submission requirements, while STARD is a consensus reporting recommendation. The practical answer is to include the diagram or follow the venue's documented handling of the item.

Indeterminate and missing results

Excluding uninterpretable tests can make accuracy look better than it is. If 45 of 1,200 specimens are indeterminate because of processing artifacts, report how those results arose, how many occurred, how they were handled in the primary analysis, and how alternate handling affects estimates where relevant. Do not silently remove them.

Sensitivity and specificity are conditional probabilities. Positive and negative predictive values also depend on prevalence in the evaluated population. A receiver operating characteristic curve may be useful for a continuous score, but it does not replace clinically justified thresholds, confidence intervals, and a description of the intended use.

A reference standard is the best available method for classifying the target condition; it is not automatically a perfect gold standard. If the reference standard is imperfect, composite, or interpreted with knowledge of the index test, disclose this and discuss the likely bias.

Masking and test review

Review bias can arise when the reader of one test knows the result of the other. Report what information was available to each reader and how interpretation occurred. Do not claim “double blind” unless that accurately describes the design. In imaging or pathology, also describe number, expertise, training, adjudication, and repeat-reader procedures when relevant.

CARE for clinical case reports

The CARE guideline is a 13-item framework for complete and transparent case reports. It covers title, keywords, abstract, patient information, clinical findings, timeline, diagnostic assessment, intervention, follow-up and outcomes, discussion, patient perspective, and informed consent information.

A useful timeline aligns symptoms, examinations, tests, diagnoses, treatment changes, and outcomes. The discussion should explain the clinical reasoning, strengths, limitations, and relevant literature without generalizing one patient's course into proof of efficacy or incidence. De-identification should be thoughtful: a rare diagnosis, precise dates, location, occupation, image, and narrative can identify a person even after the name is removed.

Consent to publish

Consent for clinical care or participation in research is distinct from consent to publish identifiable case details. Follow the journal's case-report policy, local law, and institutional rules. CARE asks the report to indicate informed consent, but the exact acceptable process and whether a waiver is possible are not uniform across every jurisdiction and journal. Never fabricate consent or imply that de-identification alone satisfies a venue that requires written publication consent.

Adverse-event reporting

A case report about a suspected adverse drug reaction may also trigger pharmacovigilance duties. Publication does not replace reporting to the applicable sponsor safety system, institution, or regulator. The responsible team should promptly route the information under the governing law, product status, agreement, and safety procedure, document what was done, and preserve patient privacy. Avoid the overbroad claim that every medical writer has an identical independent duty to report directly to every regulator.

Other design-specific guidance

The EQUATOR Network catalogs many reporting guidelines. Common examples include:

Use the current version and applicable extension. A checklist citation without item-level coverage is not enough. Conversely, avoid forcing a study into a guideline merely because an acronym appears familiar.