Tuesday, August 26, 2008: 4:10 PM
Surveillance for notifiable conditions can be improved by monitoring existing clinical data flows. Building on standards for message structure and content (HL7 and LOINC®), we have implemented and maintained an automated notifiable condition reporting system for the last 9 years in the Indiana Network for Patient Care, an operational health information exchange. The system receives real-time HL7 clinical results from a variety of health information exchange stakeholders, translates these disparate proprietary codes into LOINC codes, determines whether the results carried by the message indicates a notifiable condition by checking the abnormal flag sometimes contained in the message, or by comparing the test results with criteria in the PHIN notifiable conditions mapping table. The system treats any result having an abnormal flag as an indication of a notifiable condition. If the abnormal flag is not set, the software compares the test result to a threshold value from the notifiable condition mapping table or to specific organisms listed in the PHIN notifiable condition table. The system has proven reliable in delivering results and has scaled to multiple clinical data sources over several years of use. A recent analysis of the system revealed that automated electronic notifiable condition reporting identified 4.4 times as many reports and identified those reports 7.9 days earlier than traditional spontaneous, paper-based reporting methods. We will describe: 1) the strategic rationale for implementing a notifiable condition processor in the context of health information exchange rather than at local hospitals; 2) the technical architecture including message normalization prerequisites and NLP methods for accommodating free text results and negation; and 3) operational aspects of the system including processes for receiving feedback from public-health and maintaining notifiable condition tables.