West Coast Ballroom
Harmonizing Multi-Site Electronic Health Records Data for Critical Care Comparative Effectiveness Studies: How do we ensure that data quality is equivalent to traditional clinical trials? (307822)
*Annie N Simpson, Medical University of South CarolinaDaniel Brinton, Medical University of South Carolina
Kit N Simpson, Medical University of South Carolina
Dee N Ford, Medical University of South Carolina
Katie Kirchoff, Medical University of South Carolina
Andrew Goodwin, Medical University of South Carolina
Keywords: Electronic Health Record, Data Quality Practice, Comparative Effectiveness Studies, Critical Care Medicine
Electronic health record (EHR) data offers tremendous potential to examine best practice adherence rates and to test new effectiveness interventions; however, it is underutilized. In preliminary work, we developed a single site computable phenotype for a cohort of ventilator dependent respiratory failure (VDRF) patients. However, to fully realize the potential for study development, the phenotype required multi-site testing 1) to examine its feasibility outside of our single site, 2) be validated to establish its accuracy and generalizability, and 3) demonstrate that data can be harmonized into an analytical dataset. Based on sparse literature from PEDSNet, data from three sites were converted to common-name data elements, and were analyzed for missing or extreme values; and when needed E-charts were examined to ascertain if missing values existed elsewhere in the EHR. A random sample of 70 patients’ data per site were validated with independent chart review. Concurrently, the lead site harmonized the three sites’ data into an analyzable data set. Details from this data quality study will be reviewed. We will discuss improvements in best-practice use of these data for CER studies.