Systematic Review of AI Clinical Decision Support Studies

A 2024 systematic review and meta-analysis included 50 studies evaluating AI-based clinical decision support systems (AI-CDSS) across 17 medical specialties. The analysis, following PRISMA guidelines, screened 3,296 records and pooled performance metrics, finding moderate discriminatory ability (pooled AUC: 0.652) and high specificity (0.819). The dataset contains extracted study characteristics from this review.

Use Cases

• Analyze the distribution of studies across the 17 medical specialties to identify research concentration gaps.
• Compare pooled performance metrics like AUC, specificity, and sensitivity across different study designs (e.g., prospective vs. retrospective).
• Examine the reported implementation details, such as the 24% prospective deployment rate, to assess real-world validation trends.
• Investigate the association between study quality (assessed via QUADAS-2) and reported performance measures like accuracy or heterogeneity (I²).

Strengths

• Structured extraction from a systematic review of 50 studies, providing a curated sample.
• Performance metrics are pooled from a meta-analysis, including AUC, specificity, accuracy, and sensitivity with confidence intervals.
• Includes quality assessment via the standardized QUADAS-2 tool for diagnostic accuracy studies.

Limitations

• Small scale dataset (21.5 KB), indicating limited extracted variables or a summary-level dataset rather than raw study data.
• Substantial heterogeneity (I² ≥ 98.9%) reported in the source meta-analysis, indicating high variability between the included primary studies.
• Based on a literature search ending December 2024, so it does not include studies published after that date.

Provenance

Source

Systematic review and meta-analysis published on figshare by author William J. Waldock.

Collection Method

Literature search of PubMed and Cochrane Library through December 2024, following PRISMA guidelines; 3,296 records were screened.

Time Range

Review covers literature through December 2024.

Freshness

Last updated March 2026, based on a review of literature through December 2024.