Developing a machine learning model to detect diagnostic uncertainty in clinical documentation.

Expert labeling, natural language processing, and machine learning methods combined with human validation resulted in highly predictive models to detect diagnostic uncertainty in clinical documentation and represent a promising approach to detecting, studying, and ultimately mitigating diagnostic uncertainty in clinical practice.