PROJECT SUMMARY
Swarup S. Swaminathan, MD is an Assistant Professor of Ophthalmology at the Bascom Palmer Eye Institute
with a career goal of becoming an independent clinician-scientist in the field of glaucoma clinical research. His
overall research focus is to utilize novel statistical and data science methodologies to improve assessment of
progression in glaucoma and early detection of those patients at greatest risk for irreversible vision loss. The
primary objectives of this K23 career development proposal are: 1) to compare currently available methods
used to monitor glaucomatous disease progression with higher-order Bayesian prediction models equipped
with data from electronic health records (EHR), and 2) to provide an academic glaucoma specialist with the
mentored research experience and formal training to conduct independent clinical research. Achieving these
objectives will provide the critical skills required to establish an independent research program focused on
applying data science principles to improve the clinical assessment of progression in glaucoma patients. The
proposed K23 application will provide valuable mentorship and formal training in biostatistics, analysis of large
databases containing longitudinal data, application of Bayesian statistics in the medical sciences, and artificial
intelligence and machine learning data analysis. The extensive technical resources available at the Bascom
Palmer Eye Institute and University of Miami Institute for Data Science & Computing, the mentorship and
expertise of his advisory committee, and the dedicated institutional commitment will provide Dr. Swaminathan
with the support needed to transition into an independent clinician-scientist. He will regularly meet with his
mentors and advisors to discuss career development, attend pertinent university seminars and workshops,
present ongoing research at national meetings, and consistently submit his work for publication. This proposal
will test the hypothesis that EHR-equipped Bayesian models outperform ordinary least square (OLS)
regression in accuracy and their ability to detect progression earlier. In Aim 1, Bayesian models equipped with
EHR population-level imaging and functional data will be constructed to calculate the rate of change in optical
coherence tomography and standard automated perimetry metrics of individual patients. In Aim 2, patient-
specific risk factor data will be incorporated into Bayesian models to further refine these individualized
predictions. These models will be compared to OLS regression, with the hypothesis that Bayesian models will
be superior. Finally, in Aim 3, an interactive application will be developed to gather data from clinical practice in
order to validate the use of Bayesian models in clinical care. An expert clinician panel will compare masked
OLS and Bayesian estimates from these cases. The results of the proposed research will provide the
foundation for an R01 grant examining the use of EHR data to improve clinical decision-making for the
longitudinal care of glaucoma patients.