Background: The automated transfer of ordering information and the return of discrete molecular data from pathology labs into the electronic health record (EHR) is becoming best practice. Current genomic testing orders are often hindered by redundant entry of patient information and the complexity of existing data transfer standards. This increases the work that hospital IT and clinical groups need to contend with as more assays become increasingly available. The FHIR standard deployed through an application programming interface (API) provides a next-generation data transfer standard. It enables a more efficient and accurate ordering experience compared to traditional Health Level 7 v2 (HL7 v2) deployments. To better understand the data transfer savings that impact clinician ordering, we compared real-world genomic ordering experiences using FHIR and HL7 v2 systems that are live in over 150 clinical sites each.

Methods: Two sequencing panel orders, Tempus xT (somatic DNA) and xG (germline DNA), were evaluated. For each DNA sequencing panel, the “Ask On Entry” (AOE) data elements that are entered by the clinical team were noted. For each of these AOE data elements, mappings to FHIR-based DNA sequencing orders and their matched HL7 v2 counterparts were compared. For both the FHIR and the HL7 v2 mappings, these were previously generated by a team of clinical informaticists and reviewed by two physicians for accuracy. The number of overall data element groups and specific data element groups were quantified to represent data transfer complexity. Descriptive statistics were used for evaluation.

Results: A comparative analysis was performed between the conventional HL7 v2 ordering methodology versus FHIR-based ordering for the Tempus xT and xG assays. The somatic xT assay included 18 total AOE questions. The FHIR approach to map these questions used six base resources (collections of related data elements). In contrast, the HL7 v2 standard required 11 equivalent segments to capture the same data. This streamlined structure in FHIR reduced the number of required fields by 45%. For the germline xG assay, 28 AOE questions were noted. Ten base resources were used in FHIR as compared to 26 segments in HL7 for a reduction by 61%.

Conclusions: In this real-world application of genomic ordering for two common DNA sequencing assays, a notable average reduction in complexity of 53% was seen when comparing FHIR-based API models to their HL7 v2 analogues. These findings support our prior anecdotal observations of lower integration times for FHIR- compared to HL7-based deployments. Given the increasing number of genomics-based assays, adoption of the FHIR standard for genomic ordering may reduce the data complexity that hospital IT teams and EHR systems face thereby decreasing the time to integration and ongoing maintenance.

VIEW THE PUBLICATION