Assessment of Performance Indicators of a Radiation Therapy Department Using an Electronic Medical Record System

Abstract

Purpose/Objective(s): A wealth of information is entered into hospital electronic health care records (EHR), radiation oncology information systems (ROIS), treatment planning systems (TPS) and spreadsheets every day as part of routine care. However, how do you design a system that makes routine aggregation and use of this big data a practical reality in the clinic that is also cost effective, expandable and productive on a scale of months rather than years? The danger is the belief that the solution is primarily about technologies. We are succeeding with a more comprehensive design strategy combining optimally selected technical solutions with standardizations, practice changes and education for stake holder buyin in an agile development framework. Materials/Methods: A multi-tiered approach to aggregation, analytics and exchange was developed by segregating distinct objectives and functional requirements. Capabilities of multiple database technologies were examined for ability to enable quickly and inexpensively progressing with aggregation and use of the data and potential roles in analytics. Use-case surveys were used to define and prioritize translational research data elements, relationships and source systems. A strategy for iterative expansion was implemented for building an infrastructure that "always" adds to clinical value (vs. having to wait for the complete system to do anything) as data elements are added in successive layers of utility priority. A database system was designed to encapsulate and organize the information to reflect conceptual hierarchies and inter-relationships and develop a radiation oncology translational research ontology. Results: The aggregation tier, Radiation Oncology Analytics Resource (ROAR), was constructed to enable extraction and integration of data from the EHR, ROIS, TPS and research spread sheets with periodic refreshes to remain current. It was designed to accommodate both fully specified data sources, development of new data sources and capture of research data. Current ROAR data for patients (17,609) treated over a 15-year period, includes treated plan (36,771) and fields (2,216,165) analysis, alignment image details, charges DVH curves. Much of the key data is in the EHR (EPIC). Current elements from our EHR include demographics, survival, longitudinal labs (3,015,187) and toxicities. Addition of longitudinal data is preceding for other identified data categories including: patient reported outcomes, medications, treatment related interventions/hospitalizations, recurrence, medical and surgical oncology. Graphical interfaces were created for end user self-service. The data base is used to support research and practice quality improvement efforts including development of practice quality assessment metrics. Conclusion: Learning from the details of our history is key to mapping a path to improve the outcomes in our future. Creating automated systems for routine aggregation and analysis is a vital and achievable step on that path.