The adoption of electronic health record (EHR) systems is growing at a fast pace in the United States, resulting in very large quantities of patient data becoming available in electronic format. The volume of EHRs represents a tremendous potential for secondary use, which is essential to achieve the potentials for effective clinical research. Most detailed clinical data found in EHR systems is captured in free text without structure or coding. Secondary use of this text requires either manual chart abstraction (a tedious and expensive process) or automatic information extraction approaches based on software-based natural language processing. These approaches offer great advantages in terms of precision, speed, and scalability but currently suffer from important limitations.
We propose to address some of these limitations through new methods allowing for efficient annotation of clinical texts, new methods for reuse of existing annotated clinical texts, and new methods to extract information from clinical text more accurately. We will validate these new methods with a patient-centered outcomes research use case: extracting and deriving the Glasgow coma scale (GCS). The GCS is a critically important neurological assessment tool embedded in both clinical practice and patient-centered outcomes research among patients suffering neurological injury. The GCS provides a practical method for assessing the level of consciousness of a patient based on clinical evaluation of eye opening, verbal response, and motor response. The GCS is a central element for risk adjustment in outcome studies in critical care patient-centered outcomes research.
Our main outcomes will include new tools and resources to improve the accuracy of information extracted from EHR textual content (e.g., GCS) and to allow for more efficient use of clinical text annotations. The efforts we are proposing will provide researchers with richer and more accurate clinical information, which is a key resource for their work. They will directly benefit from this research and will guide our efforts throughout the project by meeting with our research team regularly and providing crucial feedback. Patients, as owners of the EHR content from which information will be extracted, are also key stakeholders. They will guide our efforts by ensuring that the information we would extract, and its interpretation and possible uses, match their own interests and needs.
*All proposed projects, including requested budgets and project periods, are approved subject to a programmatic and budget review by PCORI staff and the negotiation of a formal award contract.