MedKnowts: Unified Documentation and Information Retrievalfor Electronic Health Records

Title of the Paper

MedKnowts: Unified Documentation and Information Retrieval for Electronic Health Records

Paper Information

• Domain: Medical documentation and information retrieval in Electronic Health Records (EHR)
• Keywords: Electronic Health Records, documentation, information retrieval, concept-oriented view, autocomplete, medical notes, structured data

Research Background and Problem Statement

• Problems and Challenges:

  • Although widely used EHR systems aim to improve healthcare quality, support collaboration, and reduce medical errors, studies have often found that they lead to excessive time spent by physicians on documentation, potentially causing cognitive overload and professional burnout.
  • Data in medical documentation is often scattered across various formats and interfaces (structured and unstructured data separation), increasing the difficulty for physicians to read and synthesize information.
  • To simplify workflows, physicians and healthcare staff frequently rely on copy-pasting and pre-filled templates, which can result in redundant entries, information overload, or even the propagation of incorrect data.
  • In emergency departments, where time is limited and patient histories are complex, physicians must quickly integrate information, but existing EHR tools lack adequate support.

• Significance:

  • Optimizing EHR documentation can not only enhance healthcare efficiency but also improve the continuity and quality of medical services.
  • Reducing cognitive load associated with information retrieval and documentation switching can enhance physician well-being and mitigate professional burnout.

• Motivation and Related Work:

  • A significant body of research focuses on developing EHR systems capable of real-time term recognition, simplified data entry, and easy retrieval.
  • Existing systems, such as "Active Notes" and "Doccurate," have contributed to tagging and information visualization but often require manual operations or lack well-designed user experiences.
  • The authors aim to improve this process through the "MedKnowts" system, which integrates documentation and information retrieval seamlessly.

Proposed Solution

• Proposed Approach:

  • Developed a platform called "MedKnowts" that integrates medical documentation and information retrieval into a unified interface.
  • Utilized concept-oriented views to capture structured data through autocomplete and post-recognition technologies.
  • Introduced a sidebar and preview window to proactively display patient history information relevant to the documentation, reducing the need for physicians to switch between documentation and data retrieval.

• Innovations:

  • Enabled the insertion of structured data without requiring special trigger symbols, significantly simplifying data entry.
  • Supported real-time semantic highlighting, providing color-coded cues based on keyword types (e.g., symptoms, lab values, medications) and helping users quickly identify negations (e.g., "No Fever").
  • Used "cards" to organize and display patient medical data by conceptual content, including lab results, medication history, and note excerpts, overcoming the traditional EHR limitation of displaying data by format.
  • Introduced a shared card feature to support multi-user collaboration, allowing physicians and assistants to share critical information within clinical teams.

• Implementation Steps and Techniques:

  • Implemented multiple trigger methods for autocomplete (e.g., context inference, deep learning models).
  • Used a customized parser for post-recognition to automatically annotate medical terms in free text.
  • Built an interactive card interface to help physicians access key medical information with minimal clicks.
  • Deployed the system prototype in an emergency department for over a year, iteratively refining it based on user feedback.

Research Outcomes

• Specific Results:

  • During over a year of deployment in an emergency department, the team documented nearly 1,500 patient cases, and the tool's overall usability received high ratings from physicians and documentation assistants.
  • Experimental results showed that the system's embedded highlighting, autocomplete, and card features significantly reduced physicians' information retrieval time and improved documentation efficiency.
  • The system achieved a System Usability Scale (SUS) score of 83.75 (a score of ≥70 is considered good).

• Advantages Over Existing Solutions:

  • Compared to systems like "Active Notes," which require manual triggers, MedKnowts offers highly automated and seamless user experiences.
  • The card view concept breaks the limitation of traditional EHRs that organize patient records by data format.
  • Supports complex contextual scenarios and dynamically adjusts recommendations based on user input.

• Experimental or Evaluation Results:

  • The autocomplete feature was widely appreciated by healthcare workers of varying experience levels, particularly aiding less experienced documentation assistants in reducing term spelling errors and cognitive load.
  • Semantic highlighting and auto-generated default text significantly reduced repetitive tasks, especially in lengthy and repetitive sections such as physical exams and system reviews.
  • Various cards demonstrated notable utility in emergency settings, such as enabling quick access to critical patient medical histories.
  • The system effectively adapted to dynamic medical scenarios, particularly in acute patient management.

• Limitations and Future Directions:

  • The scalability of auto-generated cards needs improvement; while the system currently supports core concepts, creating more detailed cards still requires significant manual effort.
  • Semantic analysis and accuracy in complex scenarios (e.g., handling negations and modifiers) require further optimization.
  • Future plans include adding more interactive pathways, such as drag-and-drop or custom grouping to create advanced information views.
  • Explore better user adaptation strategies to help beginners quickly master full functionality while providing advanced customization options for experienced users.