HCSL Software Repository

DiagAccU

Chatzimichail RA, Chatzimichail T, Hatjimihail AT. DiagAccU: A unified framework for uncertainty estimation of diagnostic accuracy measures. Ver. 1.3.4. Hellenic Complex Systems Laboratory; 2026.

DOI: 10.5281/zenodo.21901487

Software Synopsis

DiagAccU implements a threshold-based computational framework for quantifying the sampling, measurement, and combined uncertainty of sixteen diagnostic accuracy measures used in screening and diagnostic testing, together with their confidence intervals. Class-conditional distributions of the measurand in the diseased and nondiseased populations are modelled as normal, lognormal, or gamma, and disease prevalence uncertainty is estimated using the Agresti–Coull approximation. Measurement uncertainty is represented by homoscedastic or heteroscedastic (linear and nonlinear) functions of the measurand. Uncertainty is propagated by first-order Taylor-series expansion, with effective degrees of freedom from the Welch–Satterthwaite formula and confidence intervals from Student's t distribution. Optimal thresholds are located numerically for nine measures and reported with their uncertainties. The program supports the selection of clinically meaningful thresholds, including confirmatory and exclusionary decision thresholds; transparent uncertainty reporting; sensitivity analyses across the distributional assumptions; and publication-quality plots and comprehensive tables for reporting and study-design applications.

Snapshot

Snapshot of the program