PD35-01: Feasibility and validation of large-scale data acquisition from the electronic health record to a secure research database for nephrolithiasis

Wilson Sui, Joshua K. Calvert, Nicholas L. Kavoussi, Adam Lewis, Nicole L. Miller, Cosmin A. Bejan, Ryan S. Hsi

Introduction: Electronic health records (EHRs) are an underutilized source of clinical data for research. A major barrier is the difficulty to securely and efficiently extract large amounts of identifiable data. Here we demonstrate feasibility and validation of utilizing an automated data extraction tool from the EHR to Research Electronic Data Capture (REDCap) for the study of nephrolithiasis.

Methods: We identified 2,257 consecutive patients with nephrolithiasis who underwent 24-hour urine studies from 2001 to 2018.  We applied the functionality in REDCap allowing us to extract clinical data directly from the EPIC electronic health record using medical record numbers. This technology allowed for the automatic import of over 3000 data points from the EHR including demographic, clinical, and laboratory information. The data was directly extracted to REDCap, a secure web platform for research data. Validation of the data extraction was performed by comparing to manual chart review in 50 patients. We examined sex differences across comorbidities, medications and 24-hour urine by students’ t-tests and chi-square analysis.

Results: We constructed the database using six source fields linked to the CDP including birthdate, gender, race, ethnicity, past medical history, and medications. Time from REDCap project creation to data linkage was two hours. Data abstraction required on average eight seconds per patient. In comparison, manual chart review required 29.2 ± 12.7 seconds per patient to abstract five index comorbidities and five specific medications. Validation of the CDP performance showed a PPV of 100% in all categories except hyperlipidemia (92%), and NPV ranged 89-100% across medical history and medications. We report sex differences in medical history and 24-hour urine abnormalities among nephrolithiasis patients undergoing metabolic evaluation (see Figure). Female stone formers were more likely to have hyperparathyroidism and also hypocitraturia and high urine pH (p-values < 0.001).

Conclusions: We demonstrate feasibility of a rapid, efficient, and large-scale data extraction from the EHR to a secure research database for nephrolithiasis research. Replication of this tool at other sites may enable large networks for nephrolithiasis research. Source of

Funding: None