Interpreting Health Events in Big Data Using Qualitative Traditions

Area of Science:

• Artificial Intelligence
• Smart Health
• Clinical Informatics

Background:

• Effective artificial intelligence (AI) in healthcare requires real-world context and validated ground truth.
• Current AI training often lacks sufficient contextual clinical knowledge.
• Qualitative methods offer a robust approach to establishing reliable ground truth for AI development.

Purpose of the Study:

• To illustrate the application of qualitative descriptive methods for generating ground truth in AI training.
• To demonstrate a hybrid approach combining sensor data and participant experience for AI model development.
• To advocate for the inclusion of qualitative research expertise in smart health AI design.

Main Methods:

• Utilized qualitative descriptive methods to gather contextual clinical knowledge.
• Integrated sensor-based data with participants' subjective descriptions of Restless Leg Syndrome episodes.
• Employed an interdisciplinary, inter-methodological research team.

Main Results:

• Successfully trained an intelligent agent to detect Restless Leg Syndrome using a combined data approach.
• Qualitative data provided crucial ground truth, enhancing the AI's contextual understanding.
• The study highlights the value of integrating diverse data types for AI efficacy.

Conclusions:

• Qualitative methods are vital for creating accurate and contextually relevant smart health AI.
• Incorporating clinician expertise in qualitative research improves AI design and end-user experience.
• A mixed-methods approach, combining quantitative and qualitative data, is optimal for training sophisticated health AI.