AI to Expedite Sex/Gender Analysis in Evidence Synthesis (AI-SGES)
A use case of digital health interventions for cancer patients
Why Sex and Gender-based Analysis in Evidence Synthesis Matters
Sex and gender are often overlooked in evidence synthesis. Our general aim is to accelerate the integration of sex/gender analysis within evidence synthesis by using AI.
Evidence synthesis is the systematic process of identifying, appraising, and integrating findings from research to answer a specific question. It supports decision-making by summarizing large bodies of evidence. Common methods include systematic reviews and clinical practice guidelines.
Sex and gender can influence health risks and treatment responses. Their systematic consideration in evidence synthesis can reveal outcome differences, reduce bias, and help address health inequities that might otherwise be overlooked. This improves the accuracy, relevance, and applicability of the evidence, ultimately supporting better healthcare decisions for diverse populations.
Using AI in Evidence Synthesis: Opportunities and Risks
Despite its importance, sex/gender analysis is still limited in systematic reviews and clinical guidelines. Artificial intelligence (AI) can automate labour-intensive tasks like selecting studies and extracting data, which can speed up evidence synthesis. However, if not carefully implemented and evaluated, AI may reinforce existing sex/gender biases and lead to inequitable conclusions.
There is a need to evaluate the performance of AI for accelerating the consideration of sex/gender in evidence synthesis.
Project Aims
This project aims to accelerate the integration of sex/gender analysis within evidence synthesis by using AI. We will tackle the following specific objectives.
- To map existing AI tools used to speed up systematic reviews and outline how they can support sex/gender analysis.
- To evaluate how sex, gender, and equity issues are handled in clinical guidelines for digital health tools for cancer.
- To evaluate AI tools and large language models like ChatGPT to see how well they help systematic reviewers identify and assess sex/gender information in research.
- To develop tools, including an AI framework and a custom GPT, to help integrate sex and gender analysis into evidence synthesis efficiently, transparently, and without amplifying bias.
Methodological Approach
Three year mixed-methods project using scoping reviews, diagnostic accuracy studies, agreement studies, surveys, and consensus methods. The engagement of the key stakeholders, including healthcare professionals and patients, will be integral to this project from inception to conclusion.
Funding
Project funded by the Swiss National Science Foundation (SNSF) under the National Research Programme (NRP) 83, Gender Medicine and Health.
Project protocols at Open Science Framework: https://osf.io/qp2er/
Project Team
Project Applicants
Claudia M. Witt – University Hospital Zurich, and University of Zurich
Janna Hastings – University of Zurich
Jesús López-Alcalde – University Hospital Zurich, and University of Zurich
Project Partners
Javier Zamora – Hospital Universitario Ramón y Cajal, IRYCIS, CIBERESP (Spain)
Gabriel Rada –Epistemonikos Foundation (Chile)
Jennifer Petkovic – Campbell and Cochrane Health Equity Thematic Group, the MuSE Consortium, Bruyere Research Institute, University of Ottawa (Canada)
Gemma Villanueva – Cochrane Response (UK)
Project Information: https://data.snf.ch/grants/grant/227036
