Important Ethical, Technical, and Epidemiological Considerations in an AI Tool Production (ETEPAI): Scoping Review

Comprehensive Summary

This study explores how an Ethical, Technical, and Epidemiological Considerations in an AI Tool Production (EETEPAI) framework can be used to guide the development of artificial intelligence (AI) tools in healthcare settings. Chew and Ngiam conducted an anlysis of academic databases, regulatory documents, and expert recommendations to gather already existing AI guidelines and frameworks. From the data collected, they combined 30 separate frameworks into a single model structured around four stages of design, development, deployment, and postdeployment, as well as three domains including ethical, technical, and epidemiological. The ETEPAI framework brings together the most essential guidelines and can be applied throughout the entire AI development process, providing essential questions, performance indicators, and highlighting common pitfalls. It also addresses important considerations such as bias, data quality, model selection, and challenges with implementation, while aligning with standards like the European Union’s trustworthiness guidelines. Overall, Chew and Ngiam present ETEPAI as an approach that integrates ethical, technical, and epidemiological perspectives into AI development and implementation in healthcare.

Outcomes and Implications

This research is significant as AI tools are increasingly being used in healthcare, where mistakes, overlooked bias, and poor design can directly affect patient outcomes. A standardized framework such as ETEPAI has the potential to help ensure these tools are developed safely, reliably, and ethically. Chew and Ngiam evaluated ETEPAI to be a successful framework for guiding the development of AI systems that can be effectively integrated into clinical settings and decision-making. It supports better validation, transparency, and ongoing monitoring methods, all of which are critical in hospitals. The authors stress that while ETEPAI could be used right away in research and development, its full impact in clinical practice will take time because it still needs more testing and wider adoption by healthcare organizations.