Diagnostic Performance of Artificial Intelligence Pain Assessment Systems in Clinical Patients: A Systematic Review and Meta-Analysis — Dataset

This record contains the extracted dataset and analysis-ready supplementary tables underlying the systematic review and diagnostic test-accuracy (DTA) meta-analysis "Diagnostic Performance of Artificial Intelligence Pain Assessment Systems in Clinical Patients" (PROSPERO CRD420261354420). Eleven databases were searched (1 January 2015 – 9 May 2026) and 189 studies were included, of which 117 contributed to at least one meta-analytic pool. Pooled estimates were computed in R 4.5.1 using metafor (v4.8-0) and mada (v0.5.12): a bivariate/HSROC model for the 2×2 pool; logit inverse-variance pooling for AUROC; Fisher-z pooling for correlation and ICC; and inverse-variance pooling for MAE standardized to a 0–10 scale. Risk of bias was assessed with QUADAS-2 and certainty with GRADE adapted for diagnostic test accuracy. Contents:- Supplementary_Data_Extracted_Dataset.xlsx — complete study-level extracted dataset (189 rows × 103 columns; sheet "Studies"). Every value is extracted from previously published reports; no patient-level or otherwise non-public data are included.- Supplementary_Tables_S1-S14.xlsx — analysis-ready supplementary tables S1–S14.- DATA_DICTIONARY.md — description of the key columns.- README.md — overview, citation, methods summary, and link to the analysis code. The R code that reproduces every pooled estimate, subgroup analysis, and figure is archived separately (see Related works / the repository linked in the README). Released under CC-BY-4.0.