Abstract
Purpose :
Current electronic health record (EHR) systems utilize a billing-centric interface, posing challenges to those seeking to perform individual and large-scale data analyses. Standardization among testing methodologies as well as a patient-centric approach to the EHR is proposed to better capture key clinical information across specialties and to enhance the process of compiling and comparing data for research, quality improvement, and use by artificial intelligence.
Methods :
The data center project was divided into four sections: data preparation, data structuring, data security and access, and data analysis. Data preparation: We developed a schema to extract and de-identify clinical data from Epic’s EHR and to anonymize testing images and their raw data. The images and respective clinical data were then associated with one another. Data structuring: Within the data lake, structured zones were built to provide spaces to view source data, manipulate and compare data, and to create published reports. Data Security and Access: All information was organized using complex rules to restrict access based on data share agreements, user status and authorization. Data analysis: A user-friendly interface was then created to allow for effective filtering and sorting of data as needed. Finally, the information could be exported to other analytical software, like R, Python or Tableau, or downloaded for direct use. As a proof of concept, we ran a pilot study to determine if the center four-point means (C4PM) on the visual field (VF) can predict visual acuity (VA).
Results :
Leveraging the data center, we identified 348 eyes with glaucoma. We were able to pull 24-2 Humphrey VF automated perimetry, foveal sensitivity information and computation of C4PMs for each eye from imaging data and associate the results with corrected VA, intraocular pressure and demographics. Individuals with macular disease were excluded. This process of identifying patients, extracting, reviewing and analyzing the data took less than 5 days.
Conclusions :
Building an EHR-integrated data center is not only possible, but it can facilitate research and enhance our understanding of different types of diseases and their comorbidities. This allows clinicians and researchers to hypothesize clinical relationships and to pursue validation for relationships in clinical findings in a quick and efficient manner.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.