An assessment and cleaning framework for electronic health records data

As a result of policies, such as the Health Information Technology for Economic and Clinical Health Act, that were designed to spur adoption and meaningful use of Electronic Health Records (EHR) to improve healthcare, an abundance of clinical data has been generated. Similar to other real-life big data, EHR data can be very “dirty,” which makes quality assessment and cleaning vital for producing accurate and complete EHR data sets that can be reused for clinical research. However, real-world quality assessment outcomes and cleaning methods for EHR data still remain largely undocumented to date. This study aims to contribute to such literature by: i) developing a data quality assessment and cleaning framework for EHR-based secondary analysis; ii) applying the framework to a case study of identifying hip fracture readmission risk factors based on data extracted from Cerner Health Facts, one of the nation's largest relational EHR data warehouses; and iii) reporting data quality problems identified and the cleaning methodologies that addressed the problems. Given the considerable similarities among various EHR systems, it is expected that the framework and findings based on Health Facts can be extended to cleaning relational EHR data in general.