PROJECT SUMMARY The free text notes in electronic health records (EHRs) contain details vital to cancer research that often are found nowhere else in the EHR. Such details include social and behavioral determinants of health, family histories, cancer staging, tumor markers, disease progression, response to treatment, overall outcomes, and more. Utilizing the free text (also called unstructured data) can be challenging, and researchers need tools to help them leverage those data in a meaningful way. The electronic medical record search engine (EMERSE) has been in operation, and under continuous development, for 17 years and was designed to help cancer researchers meaningfully use unstructured EHR data. EMERSE started as simple information retrieval system but has since gained many features for supporting clinical research. It is distinguished from other text-processing research tools, in part, by its ease of use and other enterprise software features such as maintaining audit logs and providing administrator functions. With the support of an ITCR grant that is nearing completion, EMERSE has grown from a tool used at only one site (University of Michigan) to one that is running, or being implemented, at multiple cancer centers and other academic research institutions across the country. EMERSE also now supports the capability of searching across a network of instances to securely obtain obfuscated patient counts at other sites, which can be valuable for identifying rare cancer cohorts. The innovations in this proposal include: (1) Building new functionality into EMERSE to expand its research capabilities such as timeline data visualizations, self-service data extraction workflows from templated notes, support for optical character recognition (OCR), and integration with other ITCR tools such as DeepPhe through an application programming interface (API); (2) Incorporating natural language processing (NLP) into EMERSE including named entity recognition (mapped to the Unified Medical Language System, UMLS), negation (e.g., “patient denies pain”), uncertainty (e.g., “possibility of recurrence”), and experiencer (e.g, “breast cancer in her mother”); (3) Expanding the EMERSE network in breadth, capability, security, and trust by partnering with external sites to collectively develop a consensus and a viable approach for broadly enabling this novel network technology; and (4) Continuing to evaluate EMERSE with a focus on networking, security, and scientific outcomes by conducting additional user-centered studies to develop a deeper understanding of how EMERSE is being, or could be, used with a goal of continuous improvement. The EMERSE team routinely receives feedback from users and site administrators and the proposed work is highly responsive to our growing user base and will enable research that is currently impractical, if not impossible, for many researchers to accomplish. The enhancements and knowledge gained from this effort will make EMERSE even more powerful and capable of supporting a wide range of clinical and translational cancer research for a growing user base across the nation.

PROJECT NARRATIVE Unstructured (free-text) clinical notes in electronic health record (EHR) systems contain valuable phenotypic, genomic, and other data needed to support cancer research ranging from identification of rare cancer cohorts to data abstraction. These notes remain underutilized, in part, because of the complexity of accessing and extracting details from the unstructured data. EMERSE (The Electronic Medical Record Search Engine), an easy-to-use, mature, information retrieval software system that is being used or implemented at research institutions across the nation, simplifies this process. The aims of this proposal are to develop novel, valuable, user-prioritized features to extend the power and reach of EMERSE, including incorporating Natural Language Processing (NLP) capabilities.