Oh Joy! Birth of a New Process to Automate Extraction of Birth Certification Information from an EHR

BACKGROUND: Clinical information in birth certificate records is critical for public health practice, policy, and research. There are known problems with data quality, and gathering birth certificate data is labor intensive and potentially error-prone. Informatics strategies for automated extraction of birth data have been ‘showcased’, but not tested within an actively-used electronic health record (EHR). Our objectives were to: a) implement a standards-based approach to extract birth certificate information from a commonly-used EHR, b) describe workflow changes when automating the process, and c) evaluate impact on data quality.

METHODS: The Utah Department of Health (UDOH) partnered with University of Utah Health Care (UUHC) which uses Epic for its in- and out-patient EHR.  In December 2015, UUHC implemented an interface to create a Labor and Delivery (L&D) Summary that meets the HL7 and IHE vital records standards and automates extraction of birth data. To evaluate changes in workflow, we reviewed current and future processes for entering and gathering information. To evaluate the quality of the electronic L&D Summary, UDOH identified 60 previously-registered births and pulled the previously-reported data. An electronic L&D Summary was created for each of these 60 births. Then, two UDOH birth certificate experts created a ‘reference standard’ by independently reviewing EHR records for the 60 births and adjudicating differences. Using the reference standard, we assessed accuracy and agreement of data previously-reported or included in the L&D Summary by comparing values for 12 data elements selected for their importance for public health policy and surveillance.

RESULTS: The electronic L&D Summary contains maternal and newborn information in one xml-based file, without requiring changes to data entry screens and processes. Most needed data are in 4 sections: Pregnancy history, Admitting diagnoses, Labor and delivery, and Newborn delivery information.  While the file generated by the interface includes needed content, many data fields do not currently display when viewed using the style sheet for rendering CDA-based xml documents. This impacts usability, but does not limit use by electronic systems that can use the file to populate vital record databases.  The Summary does not include paternity information. Strengths and limitations and the quality of the data in the L&D Summary will be presented.

CONCLUSIONS: Automated extraction of birth certificate information available in an EHR is feasible and supports, and may enhance, the data abstraction tasks currently performed by hospital birth clerks. Enhancements are required to support human review of the extracted information.