Initial Production Release
Castor Catalyst is a clinical data review platform that enables research teams to ingest source data from structured EMR records and uploaded PDF documents, review and approve extracted data points, and submit them to Castor CDMS.
Key Capabilities
Authentication and Access
- Users log in exclusively via their Castor Identity account
- Upon login, only studies and sites the user has access to are available.
Deep Linking
- Catalyst can be launched from within the Castor CDMS interface, directing the user to the corresponding study and participant context.
- Deep-link URLs are preserved through the login flow.
FHIR Data Pipeline
- Retrieves and imports structured EMR data in FHIR R4 format.
- Maps FHIR resources to the study eCRF structure using deterministic, version-controlled mappings. Data points that cannot be mapped deterministically are mapped using AI.
- Configurable rules de-identify PII from incoming FHIR data before processing and storage.
- The pipeline is modular by clinical domain (e.g., Lab Results, Vitals), changes to one domain do not require re-validation of others.
Document Processing Pipeline
- Users upload PDF source documents for automated AI-powered data extraction.
- Users select the specific clinical domain pipelines (e.g., labs, vitals) to execute on the uploaded document.
- A simplified drag-and-drop upload interface is available for site users.
Data Review and Submission
- All processed data from both pipelines is presented in a unified review interface before submission to Castor CDMS.
- Users navigate through each data point and can approve or manually override extracted values. Supported field types: Numeric, String, Multi-line text, Date, Date & Time, Time, Dropdown, Radio button, Checkbox.
- A 100% review policy is enforced for document pipeline data, no data is submitted without explicit user confirmation.
- PDF-sourced data points are visually highlighted on the original document using a bounding box. FHIR-sourced data displays the structured source element.
- Data types and values are validated against destination Castor CDMS fields before submission.
- Validation errors are displayed per field with actionable messages.
- The system prevents duplicate data point creation and data overwriting in Castor CDMS.
Audit Trail
- A centralized, 21 CFR Part 11 compliant audit trail records all user actions including reviews, manual overrides (with original AI-suggested and user-corrected values), and approvals.
- For document-sourced data, the audit trail includes an immutable reference to the source document, page number, and approved value.
- All submissions are attributed to the authenticated user in the Castor CDMS audit trail.
Security and PII Management
- PII is filtered and redacted at the ingestion boundary. Raw source data is segregated from processed application data with least-privilege access enforcement.