To systematically map the implementation strategies and outcomes of prospective deep learning (DL) implementation studies in healthcare, focusing on specific strategies such as stakeholder engagement and workflow integration.
Key Findings:
20 articles were included, covering various specialties: 3 in radiology, 1 in otolaryngology, 3 in dermatology, and 13 in ophthalmology.
All studies assessed clinical outcomes, demonstrating effectiveness and feasibility of DL integration into clinical workflows.
Adoption and appropriateness were the most frequently evaluated implementation outcomes.
Only one study evaluated implementation costs, and none assessed sustainability.
Stakeholder acceptability was evaluated in only 8 studies, indicating a need for broader assessment.
Interpretation:
The review highlights a significant gap in real-world DL implementation research, emphasizing the urgent need for further studies using hybrid effectiveness-implementation designs to enhance clinical adoption.
Limitations:
Limited number of studies evaluated implementation costs and sustainability, and there was insufficient assessment of stakeholder acceptability across studies.
Conclusion:
Continued research is essential to facilitate the effective adoption of DL systems in clinical practice, specifically addressing the identified gaps in current studies.
