AI-Based Medical Decision Support: Exploring the Data Gap

Good judgment remains fundamental to clinical decision-making, and yet it is increasingly augmented by data and artificial intelligence (AI). Although AI holds promise for real-time clinical-decision support, its impact on patient care has been modest. The principal limitation is not algorithmic capability but the quality, structure, and completeness of the data available for training and deployment. Most AI systems rely on electronic medical records (EMRs), which were designed primarily for billing rather than clinical insight. Consequently, important clinical information is fragmented, inconsistently documented, or absent altogether. Natural language processing and large language models (LLMs) improve data extraction, and yet they remain constrained by the underlying data quality and important privacy concerns. A critical gap persists in the acquisition of quantitative physiological data, particularly for the musculoskeletal system, where current practice relies on qualitative or semiquantitative assessments collected at single time points. In contrast, other industries-such as the autonomous vehicle industry-have advanced further by integrating continuous, multimodal sensor data to inform real-time decisions. Emerging multimodal wearable technologies offer a pathway toward similarly rich physiological data sets in medicine. Meaningful progress in AI-enabled health care will require such a transformation in data acquisition, enabling more accurate, continuous, and clinically relevant decision support.