Enhancing SDC workflows with AI: Automation and beyond by Maria Ryzhikova | FHIR DevDays 2025

Enhancing Structured Data Capturing with AI

Introduction to the Session

• Maria welcomes participants and addresses a common issue of time-consuming repetition in creating questionnaires.

• She introduces herself as a product manager for Box Forms at Health Samurai, with experience in health IT since 2020.

Overview of the Agenda

• The session will cover:

• A review of FHIR SDC (Structured Data Capture) and its roles.

• The design process of FHIR questionnaires and how AI can enhance it.

• Simplifying multilingual form creation using AI.

• Other aspects where AI can assist in SDC.

Understanding FHIR SDC

• FHIR SDC is an official extension developed by HL7 that enhances basic FHIR questionnaire functionalities.

• It adds features like conditional logic, UI rendering rules, answer pre-filling, and structured data extraction to improve form functions.

• Key roles defined in SDC include:

• Form Designer

• Form Manager

• Form Filler

• Response Manager

• Receiver

• Archiver

Workflow in SDC Implementation

• There are two primary workflows: form creation and form filling.

• The form creation process involves collaboration between the form designer and manager to iteratively create and update forms.

Components of a FHIR Questionnaire

• A questionnaire consists of metadata (version, name, title, status) and items (displays, groups, questions).

• Designing a questionnaire requires defining structure, writing rules with "FHIR Path," field pre-filling, and data extraction—areas where AI integration is beneficial.

Integration of AI into Form Creation Process

• Current AI tools allow users to define the structure of a FHIR questionnaire and generate templates that require further refinement but save time on initial setup.

Demonstration of the Tool's Capabilities

• Maria demonstrates uploading a PDF form (GET7), which is converted into the corresponding FHIR format using OCR technology for text extraction.

Verification Process After Conversion

• Once generated, JSON files are validated automatically; errors are highlighted for user correction before final submission.

User Interface Features

• The interface includes:

• A tree view for item management on the left side.

• Real-time rendering of forms on the right side for immediate testing.

Form Creation and Multilingual Support in Data Extraction

Enhancing Form Layout and Score Calculation

• The speaker discusses improving the layout of a form by placing two fields on one line and adding a total score calculation.

• A table format is suggested for questions, utilizing a choice matrix widget to enhance visual organization.

• The process of assigning scores to answers is described, emphasizing the potential for automation in this task.

• To finalize score calculations, fields can be set as read-only with expressions added via the Fire Pass editor.

• AI assistance is acknowledged in streamlining form creation processes, particularly in calculating total scores.

Implementing Multilingual Forms

• The need for localization support in forms is introduced, presenting two options: creating separate forms for each language or developing a multilingual form.

• The advantages of using multilingual forms are highlighted, such as simplifying maintenance when edits are required across multiple languages.

• The speaker explains how to define default languages and utilize translation extensions within the Fire questionnaire framework.

• An example of setting up translations from English to French is provided, showcasing user interface capabilities for managing translations.

• After generating translations with AI assistance, users can review and adjust them before finalizing the form.

Future Applications of AI in Form Development

• Upcoming plans include generating expressions and prefill logic from text prompts to enhance form creation efficiency.

• There are intentions to guide users through forms using UI interactions and voice commands, making data entry more intuitive.

• Efforts are underway to convert PDFs into structured questionnaire responses to streamline medical workflows while ensuring accessibility for non-tech users.

• The presentation concludes with an invitation for further discussion on these topics during the week.