Bridging the Trust Gap: A Comprehensive Dataset on Responsible AI in Clinical Decision Support Systems

This dataset was created and analyzed for the article Responsible AI in Clinical Decision Support Systems: A Large-Scale Hybrid Review Using Topic Modeling and Qualitative Synthesis. It consists of a structured collection of 15,438 unique abstracts retrieved from PubMed and Scopus (2019–2025) on Artificial Intelligence in Clinical Decision Support Systems (AI-CDSS). The data underwent deduplication, cleaning, and preprocessing (tokenization and stop-word removal) before being analyzed using the BERTopic modeling pipeline (MiniLM embeddings, UMAP dimensional reduction, HDBSCAN clustering, and c-TF-IDF/KeyBERT for keyword extraction).

The quantitative phase produced 37 thematic clusters mapping the AI-CDSS research landscape. From these, two high-relevance clusters (focused on trust, explainability, and usability in CDSS and AI integration in oncology/imaging) were selected for qualitative synthesis. A purposive sample of 24 influential articles (dual-indexed in Scopus and Web of Science, ≥10 citations, full-text available) was coded across 11 predefined dimensions (e.g., domain, clinical task, integration, trust, explainability, ethics).

This dataset provides a transparent, reproducible foundation for large-scale hybrid systematic reviews, supporting both computational topic modeling and expert-driven qualitative evaluation. It is intended for use in auditing Responsible AI adoption in healthcare, particularly regarding trust, explainability, and ethical compliance.