The integrity of health information in Electronic Health Record (EHR) is critical to delivering evidence-based care and improving population health through research. EHR errors are observed on fundamental data elements such as height/weight, causing both biologically implausible and patient-level conflicting data. We used a nonparametric regression with a Gaussian moving window to systematically identify outliers on population and individual level from IBM Explorys Therapeutic Dataset. This framework proves effective in improving data quality of EHR.
Learning Objective 1: Discuss the state-of-art informatics approaches to improving data quality in Electronic Health Record
Wei Yao (Presenter)
IBM Watson Health
John Borsi, IBM Watson Health
Yifan Xu, IBM Watson Health
Nnenna Ibeanusi, IBM Watson Health