When the American Academy of Ophthalmology IRIS® Registry (Intelligence in Research and Sight) was established in the 2010s as a means of allowing Academy members to meet Medicare quality reporting system requirements, those who helped conceive its founding understood the database’s potential to serve as a foundation for research. Today, more than 18,000 clinicians contribute de-identified patient data to the IRIS Registry, which comprises over 400 million visits from over 70 million unique patients. Verana Health, the Academy’s end-to-end data and analytics partner, has been instrumental in ensuring that the IRIS Registry’s vast pool of electronic health record (EHR) data is organized in a way that makes it functional to researchers seeking to understand real-world practice patterns. In fact, with its VeraQ™ population health data engine connecting, de-identifying, harmonizing, tokenizing, and linking disparate data, Verana Health is able to produce high quality, curated Qdata™ to power research insights.
Standardizing Data from Dozens of EHRs
Among the first steps of curating the specialty-specific IRIS Registry data is standardizing data contributed from various EHR software platforms. As the partnership between the Academy and Verana Health has matured, the standardization of objective, quantitative data points from multiple different EHR platforms, such as intraocular pressure measurements and visual acuity measurements, has become easier, and the ability to link to other de-identified data sources has become more streamlined.
Linking EHR and Imaging Data
The next phase of deepening the utility of the IRIS Registry data relies on the ability to connect additional quantified data points with patient EHR data. Imaging modalities such as optical coherence tomography (OCT) and fundus photography are particularly relevant to ophthalmology, as our field relies heavily on these imaging modalities to diagnose conditions, monitor progression, and guide treatment plans. As a medical director at Verana Health, I have helped to articulate the methods and importance of linking ophthalmic image data to patient EHR data in the IRIS Registry, as illustrated in two recent media outlets: a video interview with the editorial staff at Ophthalmology Times, and a blog post I co-authored with Theodore Leng, MD, MS, for Ocular Surgery News.
I summarized my presentation from the 2021 Association for Research in Vision and Ophthalmology (ARVO) annual meeting in a recent interview with Ophthalmology Times. In this discussion, it was explained that use of a specific algorithm designed to connect images and EHR patient data allowed approximately 1.5 million images to be linked with patient profiles in the IRIS Registry. This demonstrates the proof of concept that accurately pairing patient EHR data with images via an algorithm-based DICOM (Digital Imaging and Communications in Medicine) standardization process was possible in ophthalmology.
This study, presented at the 2021 ARVO annual meeting, was explored in further detail in the blog post. There, Dr. Leng and I explained that part of the challenge of linking ophthalmic images with patient profiles is the heterogeneity of embedded metadata structures in various ophthalmic imaging platforms and software programs. Because of this, we developed a process to reconcile platform-specific metadata with DICOM standards, which in turn permitted linkage of patient profiles and images.
Among the findings of the study I shared at the 2021 ARVO annual meeting were that:
- Harmonization, the process of aligning data from disparate sources, is possible with ophthalmic images using DICOM metadata. Armed with this knowledge, the IRIS Registry and Verana Health could decide to move forward with a large-scale implementation of an algorithm-based approach to harmonization.
- At the time of study, approximately 83% of images were linked to patient profiles in the IRIS Registry. Approximately 1.5 million images among 1.8 million images were linked to patient profiles. Of the 58,500 unique patients whose images were used in this study, 48,500 patient profiles were linked to images by the end of the study.
- Room for improvement in DICOM-based data harmonization exists. Improving DICOM standard adherence by imaging device manufacturers could be key in enhancing the image-based breadth and depth of the IRIS Registry.
Further analyses—demonstrating improvement in image-to-patient profile linkage rates—have been conducted and are planned for future publication. As we continue to make inroads in connecting the dots between real-world data sources, such as image and EHR data, the ability to extract meaningful insights to advance patient care will continue to progress and provide tools for the medical community at large.