Purpose: Create a head and neck patient database consolidating patient information from across medical record platforms to track patient outcomes, foster future research projects, and aid professional and technical development.
Methods: Consultants, including medical physicists, physicians, and medical software developers, were interviewed to determine what data would be most impactful to aggregate into a source-available cross-platform document-oriented database. This group was responsible for defining the information to be collected, the structure of the data, the collection and storage methods, and the database interface. Only data with clinical or utilization value was defined to reserve database space as well as data entry time. Patient data was collected from three separate medical software interfaces — Epic, Mosaiq, and Pinnacle. All files were converted into JSON format and integrated into a NoSQL relational user model database. All patient information was fully anonymized prior to database insertion. This activity was reviewed by our Institutional Review Board.
Results: A head and neck database was created. Data collection and anonymization was consistent across all patients. The database includes relevant medical history, treatment, and post-treatment data including the relevant simulation and daily image sets. Data abstraction continues to take manual collection for database organizers. This collection could be streamlined to increase data entry.
Conclusion: The organizational structure and intent of a clinical database has been proposed and implemented for HN patients. The next step will be to use machine learning techniques in tandem with this database to predict follow-up complications and track patient outcomes. Potential improvements continue to be evaluated such as increasing the frequency and type of cancers recorded as well as opening access to the database and its construction to multiple centers.