Big data to transform patient care

Faculty are spearheading efforts within the Byers Eye Institute at Stanford to organize electronic health records(EHR) and imaging data to facilitate a broader goal of precision health. Data registries for specific ocular disease outcomes are stored in the cloud in a secure database, primed for analysis. The challenge in big data is integrating and organizing EHR information, such as physician progress notes and visual image interpretations. Recent advances in algorithms and computing power enable finding patterns in big data that humans cannot otherwise see.

“Integrating and analyzing these extensive patient data will help improve the accuracy of disease prognosis as well as predict which therapies would be best for each patient,” Sophia Wang, MD, assistant professor of ophthalmology said. “The hope is that having access to algorithms that can predict a patient’s prognosis could help physicians everywhere prevent eye diseases from progressing any further.”

Creating an internal data warehouse

Wang is involved in multiple big data projects, collaborating with Tina Hernandez-Boussard, MD, PhD, MPH, associate professor in medicine (biomedical informatics), and Suzann Pershing, MD, MS, associate professor of ophthalmology. Hernandez-Boussard and Wang are working together to validate algorithms which can parse extra pertinent information from free text in the EHR. This utilizes a subfield of artificial intelligence known as natural language processing, which allows a computer to understand human language as a person speaks or writes. 

Wang and Robert Chang, MD, associate professor of ophthalmology, are also working on an internal data warehouse to investigate glaucoma surgical outcomes using real-world patient care data at Byers. A technology known as pictures archiving and communication systems (PACS) stores digital electronic images from clinical reports on a secure server. These images include tests such as fundus imaging, visual field data, and optical coherence tomography data for glaucoma. Their goal is to link together EHR and the PACS with longitudinal outcome information. 

Chang noted that more diverse datasets are needed to help eliminate bias. If a patient receives care at Stanford and later receives care at a different hospital, linking this data from one hospital to another could improve their care and understanding of their disease course. The challenge ahead is figuring out best practices for aggregating and linking data across centers, which presents security challenges when it is identifiable data, known as protected health information, rather than de-identified data. To overcome this issue, Chang is exploring the use of blockchain technology, privacy-preserving computation, and federated learning.

Teaming up in multi-center studies

Pershing and Wang are also contributing to a multi-center collaboration led by the University of Michigan to develop the Sight OUtcomes Research CollaborativE (SOURCE) Registry: aggregated, de-identified EHR and imaging data from nearly a dozen U.S. academic centers. For nearly two years, Wang and Pershing have worked on extracting data and ensuring de-identified linkage to protect privacy and security of patient information. The SOURCE eye registry will contain detailed eye health and clinical care data on millions of patients, enabling innovative research.

Our focus will be on linking all this data from different sources together. We want to be at the forefront for aggregating data for diverse patient populations. We cannot take a one-size-fits-all approach for patient care anymore.

In addition to the SOURCE registry, Pershing is also spearheading Stanford’s research efforts using data from the American Academy of Ophthalmology Intelligent Research In Sight (IRIS®) Registry, the largest national registry of eye health data. The continually-growing registry includes aggregated de-identified data on almost 60 million patients, and nearly 350 million patient visit records derived from over 16,000 ophthalmology practices across the country. 

The IRIS registry could also potentially reduce costs by using massive real-world data to reduce the need for expensive clinical trials, and assist in candidate recruitment and data collection, an emerging approach called registry-embedded clinical trials.

Sights on the future

Byers Eye Institute faculty efforts in big data extend beyond the U.S. as well. For example, Pershing is collaborating with Geoffrey Tabin, MD, Fairweather Foundation professor of ophthalmology and global medicine, to collect, analyze, and act on eye care data from his work in Asia and Africa.

Moving forward, artificial intelligence algorithms will be used to mine each of these large, diverse datasets for clinical insights. As patients begin to monitor their own eye health at home in between office visits, even more data will provide a much richer clinical picture of eye health, disease, and the relation to general systemic health.

“Our focus will be on linking all this data from different sources together,” Pershing said. “We want to be at the forefront for aggregating data for diverse patient populations. We cannot take a one-size-fits-all approach for patient care anymore.”

By KATHRYN SILL

Kathryn Sill is a web and communications specialist for the Byers Eye Institute in the Department of Ophthalmology, at Stanford University School of Medicine. Email her at ksill@stanford.edu.