AI-Based Medical Decision Support: Exploring the Data Gap.

Area of Science:

• Biomedical Engineering
• Clinical Informatics
• Artificial Intelligence in Medicine

Background:

• Clinical decision-making relies on good judgment, increasingly augmented by artificial intelligence (AI).
• Current AI impact on patient care is modest due to limitations in data quality, structure, and completeness from electronic medical records (EMRs).
• EMRs are designed for billing, leading to fragmented, inconsistent, or missing clinical information, hindering AI effectiveness.

Purpose of the Study:

• To highlight the limitations of current data sources for AI in healthcare.
• To propose a pathway for improving AI-driven clinical decision support.
• To emphasize the need for enhanced data acquisition strategies in medicine.

Main Methods:

• Analysis of current AI limitations in healthcare, focusing on data quality issues in EMRs.
• Exploration of natural language processing (NLP) and large language models (LLMs) for data extraction.
• Comparison with data integration strategies in other industries, like autonomous vehicles.
• Identification of emerging multimodal wearable technologies as a solution.

Main Results:

• Algorithmic capability is less of a limitation than the quality and structure of available data.
• NLP and LLMs improve data extraction but are constrained by underlying data quality and privacy concerns.
• A critical gap exists in acquiring quantitative physiological data, especially for the musculoskeletal system.
• Other industries successfully use continuous, multimodal sensor data for real-time decision-making.

Conclusions:

• Meaningful progress in AI-enabled healthcare requires a transformation in data acquisition.
• Integrating continuous, multimodal sensor data from wearable technologies can provide richer physiological datasets.
• This data transformation is essential for enabling more accurate, continuous, and clinically relevant AI-driven decision support.