Barriers to automated visual field perimetry data extraction, harmonization, and representation in OMOP

Abstract

Purpose : Preserving the visual field (VF) is the endpoint of glaucoma care. Perimetry data are not typically represented in existing “big data” resources, hindering our ability to study functional progression. We analyzed different methods for extracting and representing perimetry data elements to determine existing barriers.

Methods : Glaucoma specialists from multiple institutions and their informatics teams attempted to extract bulk data from their perimeter devices. A list of all extraction methods and resulting output files was created and compared among the institutions. Data elements were extracted from materials published by the vendors, VF reports, and public data repositories. The extracted data elements from different vendors were compared to evaluate interoperability and searched against existing concepts within the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) and standard vocabularies using the Athena tool.

Results : The bulk data extraction from Zeiss Humphrey Field Analyzer (HFA) and Haag-Streit Octopus (OPS) perimeters lacked standardization and was not specified in the user manuals. Extraction from the devices yielded reports or raw test data in image information object definition (IOD), comma-separated values (CSV), DICOM files in encapsulated or raw, and Extensible Mark-up Language (XML) formats. Zeiss provided an ophthalmic visual field (OPV)-specific DICOM conformance document, while Haag-Streit has not implemented OPV DICOM standard for their perimeters. Fifty-seven and 46 data elements were identified for HFA and OPS, respectively. These elements were categorized into eye-level, cluster-level, and point-level concepts. OMOP CDM contained 19.3% (N=11/57) and 13 % (N=6/46) of the data elements for HFA and OPS, respectively (Table 1). Notably, the CDM included VF interpretation concepts (e.g., paracentral scotoma of the right eye).

Conclusions : Barriers hindering the representation of perimetry data include the lack of a standardized method for bulk data export, limited accessibility to advanced data export tools, and limited concept coverage within the OMOP CDM. Addressing these challenges is crucial for achieving data harmonization, promoting interoperability, and empowering future multicenter clinical research.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.