Through the Looking Glass: Splitting ELR Messages for Consumption By Public Health Surveillance Systems

BACKGROUND: To process electronic laboratory result (ELR) messages into surveillance systems utilized by the Tennessee Department of Health (TDH), the messages are “split.”  Splitting is a term used by TDH and others to describe ELR message manipulation to contain only one order segment (OBR) and/or one observation segment (OBX) per message.  The HL7 Version 2.5.1 Implementation Guide: Electronic Laboratory Reporting to Public Health allows for multiple OBR and related OBX segments to be sent in one message, but TDH must split messages by OBR or by OBX due to limitations with surveillance systems.  TDH uses the NEDSS Base System (NBS), Patient Reporting Investigation Surveillance Manager (PRISM), and Enhanced HIV/AIDS Reporting System (eHARS), each with unique constraints that must be accounted for when processing ELR messages.

METHODS: Before splitting the ELR messages, TDH determined the requirements for processing ELR messages into each system.   This involved creating scenarios and test messages, processing into the systems, reviewing and decoding error messages from systems when applicable, revising messages, and starting again.  Once requirements were known, including which fields the systems use for snapshot processing to determine if a message is an update, TDH used Rhapsody to manipulate the messages received.  

RESULTS: TDH determined numerous constraints within the surveillance systems.  NBS is limited to one order/observation group per message, except for parent-child relationships for scenarios like drug susceptibilities.  Rhapsody routes must scan messages for multiple order/observation groups, determine if a parent-child relationship is present, and split the message when one is not.  Similarly, Rhapsody routes must split ELR messages containing multiple observations in certain instances, such as when consuming applications only process one observation at a time or when program area must be derived.  Considerations are made before splitting and creating new messages, including how to duplicate segments across messages without affecting snapshot processing.

CONCLUSIONS: ELR on-boarding does not end once a near perfect HL7 message is achieved by a trading partner.  Resources must then go into configuring data received into consumable formats for systems utilized within the TDH.  Processes must be reviewed and updated as consuming applications evolve over time, requiring strong informatics capacity to understand each scenario and determine how to approach each processing issue that arises.  Fidelity is crucial with transformation or reconfiguration of these data, and it must be ensured that no meaning is being lost and nothing is inappropriately inferred from data used within systems for daily public health practice.