Do You Know Where Your ACK Messages Are? Assessing and Correcting Fatal Errors from Vaccine Submissions to a State Immunization Registry

Purpose/Objectives: The COVID-19 pandemic has highlighted the importance of timely, accurate, and complete reporting of immunization information from electronic health records (EHRs) to existing immunization registries. However, vaccine submissions to registries are not always successful, with errors being reported back to the sending interface using HL7 acknowledgement (ACK) messages. Little attention has been paid to a more formal characterization of these fatal error messages, including types of errors and how best to handle them. We describe a collaboration between the Tennessee state immunization registry (TennIIS), EHR technical support engineers, and institutional informaticians to develop and implement a process to visualize, analyze, and correct fatal error messages. This work helped inform the activities of a national CDC-sponsored HIMSS Immunization Integration Program initiative to improve EHR-IIS interoperability.

Design/Methods: In July, 2020, TennIIS began generating weekly "Summary of Errors" reports with figures for total vaccine (VXU) messages submitted, number of error messages generated, and a list by frequency of types of errors. This report was iteratively tweaked over the next several months to include additional statistics, and it is currently emailed to our institution each Monday. Concurrently, EHR technical support engineers worked with informaticians to develop workqueue and reporting tools. Weekly text file extracts of specific data elements from ACK messages and corresponding VXU messages are automatically exported every Monday to a secure directory. Additional web-based tools have been developed to import these files for more detailed reporting and analysis. Results are communicated back to TennIIS and EHR technical support personnel to determine causes of errors, possible solutions, and resubmission of vaccine messages if feasible.

Results: Multiple types of systematic errors were identified, caused by issues on both the IIS and EHR sides. IIS issues included configuration errors and missing functionalities. An incorrect NDC code setting in the IIS resulted in 2,571 influenza vaccine messages being rejected. The setting was corrected and the vaccines successfully resubmitted. EHR issues included configuration/mapping errors, with one incorrect mapping causing 3,710 vaccines to be rejected. Missing demographics, an unusual 'Name Type Code' error, and invalid characters in patient names were among other discovered errors. 1,602 vaccines were successfully resubmitted after addressing some of these errors - including 519 COVID-19 vaccinations.

Conclusion/Discussion: The combination of TennIIS weekly summaries, vendor exporting functionalities, and web-based analysis tools provided previously unknown insights into the types of fatal errors that were occurring with vaccine submissions. These efforts were valuable not only in identifying and successfully resubmitting nearly 8,000 vaccines to the registry, but also in increasing our understanding of various root causes of errors and the different stakeholders who might be best suited to address them. Finally, it should be possible to scale these functionalities across other IIS and EHR vendors to support similar efforts.