Evidence Map · e0005
Multimodal and generative artificial intelligence in clinical decision support: an evidence map
Bolat Sadykov, Timur Saliev, Dimash Davletov. Multimodal and generative artificial intelligence in clinical decision support: an evidence map. Global Medical Reviews. 2026;1(1):78–96.
Abstract
Background
Multimodal and generative artificial intelligence systems are increasingly evaluated for clinical decision support, but technical performance does not establish patient benefit. This review aimed to map clinical evidence while separating multimodal systems from adjacent generative decision support.
Methods
A PRISMA 2020-informed evidence map considered peer-reviewed reports from 2018 through 9 July 2026, the date of final source verification and citation updating. Screening and extraction were conducted in a single review stream. Descriptive appraisals and a review-specific evidence-maturity framework informed the narrative synthesis. The map was not prospectively registered and no public protocol was deposited.
Results
The set comprised 25 assessed reports: 17 retained and 8 excluded or used as background. Eleven reports formed the multimodal core and six the generative contextual layer. Core reports addressed triage, prognosis, pathology, radiology, echocardiography, oncology, ultrasound, emergency care, simulated telehealth and dentistry. Diagnostic discrimination, expert concordance, report quality and usability predominated. Contextual reports contributed more direct workflow and patient-outcome evidence. Related reports were not assumed to represent independent patient datasets. Safety, equity and longer-term outcomes were incompletely covered.
Conclusions
Evidence supports further evaluation of assistive systems with clinician oversight. Autonomous deployment and broad claims of patient benefit remain insufficiently supported. Prospective trials, external validation, safety monitoring and subgroup analyses are needed.
Keywords
Article history
- Received
- Revised
- Accepted
- First published online
Peer-review model: Double-blind external peer review · Version: Version of Record
Supplementary materials
Supplementary tables and materials are included within the Version of Record PDF and HTML full text where present.