Inspired by patients, USC clinician-scientist Kyle Bolo aims to improve glaucoma screening

For glaucoma specialist Kyle Bolo, MD, his research mission starts with patients he sees in the clinic every day. All too often, they’ve already experienced some vision loss from an irreversible disease of the optic nerve that, without intervention, can lead to blindness.

“Glaucoma is sort of a silent disease,” said Bolo, an assistant professor of clinical ophthalmology at the Keck School of Medicine of USC. “A lot of patients have been either undertreated or not treated at all, and they come in with advanced disease. That always hurts, because I can do everything in my power to help them maintain what they have, but there’s less to save, and it’s harder to save what remains. That strongly motivates my research.”

He’s working toward a day when early detection makes advanced glaucoma an unlikely diagnosis, with support from the American Glaucoma Society, industry collaborators at Zeiss and National Institutes of Health funding through the Southern California Clinical and Translational Science Institute. Bolo and his teammates at the USC Roski Eye Institute are using machine learning algorithms to comb through long-term data about current glaucoma screening practices. They aim to bring light to unanswered questions about the effectiveness of current screening in improving patient outcomes.

“There are some gaps in our literature,” Bolo said. “Screening has a long history in glaucoma, but the health outcomes and any potential for unintended effects have never been studied thoroughly. We’re trying to analyze large amounts of data to make conclusions about the practice of screening that can be applied broadly.”

An opportunity to shed new light on glaucoma screening

The attempt to preserve people’s vision through widespread screening comes with its own risks. False positives lead to overtreatment, unnecessarily exposing some patients to side effects from medicine or surgery. Even if a specialist determines a patient does not have glaucoma, the referral itself can create turmoil.

“Sometimes patients come to us very scared,” Bolo said. “Overdiagnosis itself can be emotionally burdensome.”

The few studies evaluating glaucoma screening have tended to be short-term and limited in scope. This is due in part to the difficulties of following patient outcomes after screenings come up negative. Even patients who test positive are difficult to follow after initial treatment.

Bolo and his colleagues are able to investigate rich longitudinal data through a partnership with the Los Angeles County Department of Public Health’s ophthalmology department. Over more than a decade, 85,000 diabetes patients have been tested for glaucoma alongside complications of diabetes that can affect vision. This trove of data brings power to the researchers’ findings. It also calls for the use of advanced statistical techniques and machine learning.

(Regarding patient privacy, Bolo emphasizes that all data is stripped of patient identifiers, that machine learning is used solely for complex statistical analysis, and that data is never exposed or used to train public algorithms.)

Patients who screened positive for glaucoma in the L.A. County program underwent diagnostic evaluation with eyecare providers. The judgments of those experts will play a key role in Bolo’s analysis.

“By having glaucoma specialists review a portion of the diagnostic tests and label which evaluations correspond with true glaucoma, we can develop statistical models that we can apply to the whole dataset to determine, with a high degree of certainty, which patients have glaucoma,” he said. “That allows us to avoid the immense task of reviewing thousands and thousands of cases — a foundational step to do this large, data science-driven research into whether screening works.”

Bringing the engineer’s perspective to biomedical science

It only makes sense that Bolo pursues research at the intersection of medicine and computer science. Although raised by two physicians who encouraged interest in medicine, he majored in math as an undergraduate.

“Medicine was always in the background of my life, but as far back as I can remember, I’ve very much had an engineering mindset,” he said. “I loved to tinker.”

His first job channeled his enthusiasm for computer science with the electronic health record company Epic Systems. This proved to be a formative experience that influenced his calling.

“I was exposed to ideas about how to organize patient data at Epic that inform how our team handles complex glaucoma data for reliable data-driven science,” he said. “Moving into medicine seemed like a way to take that engineering mindset and make an impact that helps people directly.”

After earning his medical degree at Columbia University, he became a resident at the Keck School of Medicine, where he conducted research with Ben Xu, MD, PhD, associate professor of ophthalmology, clinical scholar and chief of the USC Roski Eye Institute’s Glaucoma Service. Returning after his fellowship training to become a USC faculty member seemed like the natural choice for Bolo.

“Ben is a very effective mentor, and the quality of mentorship at USC overall was a huge draw for me,” he said. “There’s a group of faculty here that give me a good trajectory to follow as a clinician-scientist.”

Bolo continues to partner with Xu, director of data science and AI at the USC Roski Eye Institute, on a larger slate of high-tech investigations. In collaboration with faculty from the USC Viterbi School of Engineering, they explore how telemedicine and artificial intelligence might be deployed to improve glaucoma care.

“If done cautiously, there’s just so much potential for AI to be applied beneficially in our everyday clinical workflows,” Bolo said.

His future ambitions include launching trials that compare outcomes between patients screened for glaucoma and those who aren’t. He wants to answer open questions about how screening compares to standard care, whether it actually affects the early detection of disease, and whether it’s beneficial to patients in the long run.

“Figuring out how to make screening effective starts with understanding how it’s working now,” Bolo said. “I’m open-minded — I have to be as a scientist — but I hope we find that screening is effective, for the sake of patients.”

Learn more about research at the Keck School of Medicine.