Migration from a historical to a new EHR system at Mayo Clinic was associated with deterioration of performance of a rule-based NLP for identification of PAD cases from clinical narratives, installed in the institutional near-real time NLP infrastructure. After refining the algorithm and keywords for the new EHR system, the revised PAD-NLP system had accuracy of 98%, sensitivity of 96%, specificity of 100%, positive predictive value of 100%, and negative predictive value of 96.15%.
Learning Objective: Migration of EHR systems significantly impacts operation and performance of analytic tools installed in the near real-time NLP infrastructure. This study describes the process to restore functionality and for further refinement of the NLP algorithm mitigating the impact of the EHR migration.
Sungrim Moon (Presenter)
Vinod Kaggal, Mayo Clinic
Sunghwan Sohn, Mayo Clinic
Hongfang Liu, Mayo Clinic
Rajeev Chaudhry, Mayo Clinic
Adelaide Arruda-Olson, Mayo Clinic