Machine Learning and Artificial Intelligence for Supporting Systematic Reviews

Michaela Lunan-Taylor, PhD
Senior Associate, Evidence Synthesis
RTI Health Solutions

Transcript:

Literature reviews (systematic, targeted or narrative) are a cornerstone of health economics and outcomes research and evidence-based decision making. However, the volume of scientific data is ever increasing, with the average systematic literature review (or SLR) estimated to take over 15 months to complete. Machine learning and artificial intelligence (AI) are being hailed as the answer to streamlining literature reviews through reduced timelines and lower consumption of research resources. However, these methods need evaluation to ensure continued robustness and appropriate integration into existing workflows. 

In our research, we identified recent evidence of the use of machine learning and AI to support systematic reviews. Barriers to the uptake included a lack of regulatory guidance, costs, training requirements, user-friendliness of tools, and transparency concerns.

Further research is needed on the best approaches to maximise recall of AI. There is also a need to expand the data sources used for training machine learning and AI tools. Given the high expectations of AI, performance assessment against current practice, rather than perfection, may be appropriate to enable efficiency and increased uptake of these methods.

Machine Learning and Artificial Intelligence for Supporting Systematic Reviews: A Systematic Review of Recent Methodological Developments and Recommendations for Implementation.
Marcano Belisario J, Lunan M, Hawe E, Thurairajah S.