Data Adjudication Architecture for Health Information Exchange (HIE): a Case of Adjudicating and Storing Hemoglobin A1C Values.

Increased participation in Health Information Exchange (HIE) has created the opportunity to integrate data from external sources into organization’s own workflows. Lack of proper data adjudication of incoming data can lead to erroneous data unfit for use in the clinical setting. In this project, we analyzed data in Intermountain Healthcare’s Enterprise Data Warehouse (EDW) to identify error patterns. From this analysis, we extracted requirements for data adjudication and designed architecture for an inferencing solution integrated in the overall system architecture to detect errors and facilitate corrective actions. Background Intermountain Healthcare is participating in multiple HIE projects including the Care Connectivity Consortium (CCC) at the national level, the Utah Clinical Health Information Exchange (cHIE) at the state level and various private exchanges. Intermountain’s experience in implementing HIE suggests that data received through HIEs may not be interoperable and present many unique data integrity challenges. Method We analyzed 2.2 million Hemoglobin A1C (HgbA1C) result records from Intermountain Healthcare’s enterprise data warehouse (EDW) to detect potential error patterns. The data in the EDW is a copy of the data from the Clinical Data Repository (CDR) which stores longitudinal patient data. Data sources for HgbA1C results include Intermountain and non-Intermountain laboratories as well as manually entered point-of-care laboratory results. From the results of the error pattern analysis and our experience with HIE, we created requirements and designed the architecture for the data adjudication system for ensuring data integration and interoperability. Results and Discussion Based on preliminary analysis, we classified the errors into the following categories: A) Completeness (presence of mandatory data); B) Physiological compatibility (HgbA1C 20%); C) Redundant data (duplicates). We found an error rate of 0.43% (431304 errors per million records), amongst which 33.9% of errors were from physiological incompatibility and 66.1% errors were due to incomplete data. Although the number of duplicate records were negligible, we believe that the error rate will increase as the number of external organizations participating in the HIE increases. Figure 1 shows the architecture for the data adjudication system. The patient information is currently exchanged in either the HITSP/C32 v2.5 or HL7 v2.x formats. The Message Orchestrator manages the workflow for the data adjudication process. Relevant data is extracted from the incoming external data and sent to the Clinical Decision Support (CDS) Data Adjudicator. The CDS Data Adjudicator then applies rules to this data to ensure data completeness, physiological compatibility and non-redundancy. Any discrepancies in data will be added to the “Critique Information” report for resolution. If the data passes all validity rules, it is stored to the CDR. Figure 1. Architecture for HIE Data Adjudication Message Orchestrator Data Extractor Critique Information CDR (Patient Data) A. Rules for Data Completeness B. Rules for Physiological Compatibility C. Rules for Redundancy Clinical Decision Support (CDS) Data Adjudicator (JBoss Drools®) External HIE Data