Saved Search

Diagnostic Test Accuracy Analyzer

Estimate diagnostic classification measures and uncertainty from one aggregate binary 2 × 2 table.

1. Diagnostic 2 × 2 table

Dataset setup

Rows are index-test results; columns are reference-standard results. Enter non-negative whole-number counts, up to 1,000,000,000,000 per cell.

Editable diagnostic 2 × 2 table
Index testReference +Reference −Total
Index +

—
Index −

—
Total———

Orientation check: TP is Index + / Reference +, and TN is Index − / Reference −. Use aggregate, non-identifiable counts only.

2. Study setupSelect the sampling design
These correspond to single-gate and two-gate diagnostic-study designs and control interpretation warnings.

Each unit should contribute once and be independent.

Report labels optional
3. Analysis settings95% confidence interval
Use a fixed pre-test probability justified for the intended clinical setting.

About this analysis

Example clinical scenario

A diagnostic study cross-classifies an index test and a prespecified reference standard for each independent participant, producing aggregate TP, FP, FN, and TN counts.

Formula and methodological note

Sensitivity, specificity, predictive values, the observed reference-standard-positive proportion, and overall accuracy use Wilson score confidence intervals. Positive and negative likelihood ratios use Simel log intervals. The diagnostic odds ratio uses a log-Woolf interval. No continuity correction is applied. At zero-cell boundaries, the observed estimate is retained as zero, infinity, or not estimable, and an unavailable log interval is stated explicitly.

Assumptions and limitations

  • One binary index test is compared with one binary reference-standard classification at a prespecified threshold.
  • Each analytical unit is independent and contributes once.
  • Predictive values, sample prevalence, and overall accuracy depend on the analyzed case mix and sampling design.
  • Case-control or two-gate studies do not provide routine-population predictive values or prevalence from the entered sample composition.
  • The analysis does not correct verification, incorporation, spectrum, threshold-selection, or imperfect-reference-standard bias.
  • Indeterminate results, missing classifications, repeated or clustered units, paired-test comparisons, ROC curves, and meta-analysis are outside this release.

Interpretation and safety

Diagnostic accuracy describes classification relative to the entered reference standard. It does not establish clinical utility, patient benefit, cost-effectiveness, treatment effectiveness, or an individual diagnosis. This research tool does not replace study-specific statistical and clinical expertise.

Privacy

Calculations use aggregate counts in browser memory. The app does not intentionally transmit or store entered counts or labels.

Validation and quality checks

Version 1.2.0 retains the original statistical method and reproduces the prespecified Wilson, Simel, log-Woolf, and probability-updating benchmarks and includes automated boundary, reporting, interface and privacy tests; manual accessibility and print acceptance remain under review. Availability does not constitute regulatory validation or guarantee that the method is appropriate for a particular study.

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.
  2. Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial proportion. Stat Sci. 2001;16:101–133.
  3. Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence. J Clin Epidemiol. 1991;44:763–770.
  4. Glas AS, Lijmer JG, Prins MH, et al. The diagnostic odds ratio. J Clin Epidemiol. 2003;56:1129–1135.
  5. Whiting PF, Tomlinson E, Rutjes AWS, et al. QUADAS-3. Ann Intern Med. 2026. doi:10.7326/ANNALS-25-02104.

Suggested citation

Yaacob, NM & Che Soh, NAA. Diagnostic Test Accuracy Analyzer, version 1.2.0. NajibMY; 2026. Available from: https://najibmy.netlify.app/tools/diagnostic-test-accuracy/

This research tool does not replace statistical consultation or clinical judgement.