Comparative Study of Medical Information Retrieval Approaches: A PRISMA-Based Systematic Review of Query Expansion, Term Weighting, Knowledge Modeling, and Bayesian Methods for Enhanced Clinical Decision Support
Medical information retrieval (IR) enables clinicians, scientists and policymakers to find relevant, reliable, and contextual evidence from clinical practice guidelines, diagnostic tools, systematic reviews and patient records. Retrieval in this field is challenged by semantic variability, abbreviations, variability in data formats and patient privacy. This review adheres to the PRISMA standards and describes four exemplary directions for methods that advance medical IR: 1) external collection cluster-based query expansion, 2) part-of-speech pre-weighting of terms, 3) domain knowledge modeling via data-driven mining of sublanguage patterns, and 4) knowledge fusion of domain's heterogeneous knowledge bases with a Bayesian integration model for precision medicine. All four methods are shown to outperform common IR benchmarks with respect to mean average precision, normalized discounted cumulative gain, recall, and annotation quality. We organize the four directions into a taxonomy named MERIT, the Medical Evidence Retrieval Improvement Taxonomy. Based on the reviewed evidence, a reliable evaluation of retrieval techniques requires the use of standard medical IR benchmarks as well as standard baselines. Hybrid IR methods that fuse external evidence, linguistic structure, mined patterns, and probabilistic knowledge may promote clinical decision support in the future.
Authors
- Grace Katuri (ORCID: https://orcid.org/0009-0002-3178-5560)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-20
- DOI
- https://doi.org/10.21203/rs.3.rs-11802469/v1
- Primary Topic
- Information Retrieval and Search Behavior
- Type
- preprint