AI-based assessment of pulmonology inpatient consultation note completeness: predicting documentation gaps and...

Inpatient consultation notes frequently suffer from incomplete documentation, which may delay clinical decision-making and compromise patient care. Although consultations are central to multidisciplinary coordination, there is still no widely adopted framework to ensure standardized documentation across specialties. This study aims to evaluate the relationship between the completeness of referring physician documentation and consultation response time in a high-volume inpatient setting and to develop an artificial intelligence (AI) tool for detecting missing information in real time.

We retrospectively analyzed 1219 inpatient consultation notes from 20 representative days sampled across seasons and weekdays. Each note was evaluated using a modified QNOTE-based framework covering seven core clinical elements. The association between missing components and consultation response time was assessed. A natural language processing (NLP)-based machine learning model was developed to predict the absence of key content from free-text notes.

Only 0.3% of notes contained all essential components. Missing information-particularly auscultation findings, thoracic imaging, and laboratory data-was associated with response delays of up to 35 min per consultation. Documentation quality and response times varied by time of day and day of the week, with more complete notes and faster responses observed during night shifts and weekends. The NLP model achieved high accuracy in identifying missing elements, with F1-scores exceeding 0.90 in several categories.

Incomplete referring notes-lacking essential clinical content provided to the consultant-are significantly associated with delayed responses, indicating an urgent need for improved documentation practices. Our AI-assisted NLP model enables real-time detection of missing, specialty-specific content from free text-without relying on rigid templates. While this study focused on pulmonology, the approach is scalable to other fields and may serve as a blueprint for standardizing inpatient consultations. Though not yet integrated into clinical workflows, it represents a practical step toward smarter and more efficient documentation.