Editverse Free Calculator

Free STARD Flow Diagram Generator for Diagnostic Accuracy Studies

Generate the official STARD 2015 flow diagram for your diagnostic test accuracy study. Live auto-calculation through enrolment, testing, and analysis. 300/600 dpi PNG and fully editable PowerPoint exports. No data leaves your browser.

In Simple Words: Diagnostic accuracy studies evaluate how well a new test (the index test) identifies a condition compared to a reference standard (the gold standard). Journals require a STARD flow diagram that shows how participants moved from initial enrolment through receiving the index test and reference standard to the final analysis dataset. Without this diagram, readers cannot assess whether results apply to their patient population or are affected by verification bias.

This tool generates the STARD 2015 flow diagram — the current international standard for diagnostic accuracy reporting — from the counts you enter. It tracks the Enrolment, Testing (index test and reference standard), and Analysis stages, with auto-calculated totals at each step.

When to Use the STARD Flow Diagram

Use STARD 2015 whenever your study evaluates the diagnostic accuracy of a test, biomarker, imaging modality, clinical prediction rule, or screening instrument. STARD applies whether the index test is a laboratory assay, imaging finding, questionnaire, or AI algorithm. The 2015 update is the current version; STARD-AI, published in 2025 in Nature Medicine, extends it for AI-based diagnostic tests.

For systematic reviews of diagnostic accuracy studies, use PRISMA 2020 with the DTA extension. For interpreting sensitivity and specificity from your own data, see the diagnostic accuracy calculator or the ROC / AUC calculator.

🔒 100% private. Diagrams are generated entirely in your browser: your counts never leave your device.
Edit the counts: the diagram updates live. Boxes auto-calculate.
Suggested caption
Tip: click any box in the preview to edit its text. Your inputs autosave in this browser · data never leaves your device.
Need help with your review or manuscript? Statistical analysis from $350 · Manuscript writing from $800 · Publication support from $150 Get a quote →

Worked Examples

Example 1 — Serum biomarker vs histology reference standard for hepatic fibrosis

A diagnostic study enrolled 600 patients with chronic liver disease referred for staging. 100 were excluded (contraindications to biopsy: 60, withdrew consent: 40). 500 received the index test (FIB-4 score). 30 did not receive the reference standard (liver biopsy) because of coagulopathy or patient preference. 470 received both tests. 20 were excluded from analysis due to inadequate biopsy sample. 450 were included in the 2×2 table for sensitivity/specificity calculation. Enter these numbers — the diagram auto-calculates at each stage.

Example 2 — Point-of-care CRP for lower respiratory tract infection diagnosis

A prospective diagnostic study recruited 400 patients presenting to primary care with cough and fever. 80 were ineligible (age under 18, prior antibiotic use). 320 received POC CRP (index test); 10 had failed capillary samples. Of 310 who received the index test, all received the reference standard (chest X-ray + microbiological culture). 15 had technically inadequate cultures; 295 were analysed. Enter 400 enrolled, 80 excluded, 320 index test, 10 failed, 310 reference standard, 0 not received, 295 analysed — the diagram populates across three stages.

Assumptions and Requirements

  • Index test and reference standard are both required in STARD. The flow must show separately how many participants received the index test and how many received the reference standard. Patients who received only one are at risk of verification bias — document them explicitly.
  • Reason for not receiving reference standard must be stated. STARD 2015 item 10 specifically requires the reason for non-receipt of the reference standard (e.g. refused, too ill, contra-indicated). Use the click-to-edit feature to add this detail to the testing box.
  • Analysis exclusions must be justified. Participants who received both tests but were excluded from analysis (e.g. inadequate specimen) must appear in the final stage with a stated reason.

Common Mistakes

  • Ignoring differential verification. If only test-positive patients received the reference standard, this is differential verification bias — the most serious methodological threat in diagnostic studies. The STARD flow diagram makes this visible; your Discussion must address it.
  • Conflating test failure with exclusion from analysis. A failed or indeterminate index test result is not the same as being excluded from the analysis. Indeterminate results should be reported and their handling stated in Methods.
  • Using the incorrect reference standard. STARD requires the reference standard to be described precisely. If you used a composite reference standard (e.g. imaging plus clinical follow-up), state this in the Testing box — click to edit it.
  • Using STARD 2003 instead of 2015. The 2003 version is no longer current. STARD 2015 has a 30-item checklist and a revised flow diagram that separates index test and reference standard receipt more explicitly.

Frequently Asked Questions

My study has multiple index tests — do I need separate diagrams?
If all participants received all index tests, one diagram is sufficient — list both tests in the Testing box using the click-to-edit feature. If different subsets received different tests, consider separate diagrams or a multi-column flow with a note in the caption.
Can I use STARD for an AI diagnostic algorithm?
Use STARD 2015 as the base and supplement it with the STARD-AI reporting guideline (Nature Medicine, 2025), which adds 18 new or modified items covering dataset practices, the AI index test and its evaluation, and algorithmic bias and fairness: these AI-specific items are documented in Methods and Results rather than in the diagram itself.
What is verification bias and does my diagram need to address it?
Verification bias occurs when whether a participant receives the reference standard depends on their index test result. It inflates sensitivity. Your STARD flow diagram will reveal it: if the “did not receive reference standard” count is higher among index-test-negative participants than positive, you have differential verification bias. State this limitation explicitly in your Discussion.
Should I export at 300 or 600 dpi for journal submission?
For most journals (BMJ, Lancet, PLOS Medicine) 300 dpi at 170 mm (full-width figure) is sufficient. For high-impact journals with strict figure specifications (JAMA, NEJM) use 600 dpi at 85 mm (single column). Check the author instructions for your target journal.

Related Free Tools

References

  1. Bossuyt PM, Reitsma JB, Bruns DE, et al. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ. 2015;351:h5527. PMID 26511519.
  2. Cohen JF, Korevaar DA, Altman DG, et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open. 2016;6:e012799. PMID 28137831.
  3. Sounderajah V, Guni A, Liu X, et al. The STARD-AI reporting guideline for diagnostic accuracy studies using artificial intelligence. Nat Med. 2025;31(10):3283-3289. PMID 40954311.
  4. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155:529–536.

Need statistical analysis and STARD-compliant manuscript writing for your diagnostic accuracy study? Editverse provides complete reporting including sensitivity, specificity, ROC analysis, and full manuscript preparation.

Get Diagnostic Study Support →