Project Summary/Abstract
The candidate and principal investigator (PI) Geoffrey Tison, MD, MPH, is an Assistant Professor in the Divi-
sion of Cardiology at the University of California, San Francisco. The long-term goal of the PI is to become an
independent clinician-investigator with the training necessary to perform technology-leveraged clinical research
to both investigate and facilitate cardiovascular disease prevention. Specifically, the training aims of this award
will allow the PI to build upon his existing clinical research and data analysis skills to employ machine learning
and other technology-based solutions, like mobile health tools, to advance heart failure prevention. The candi-
date will complete coursework to develop his skills in machine learning, medical informatics, and clinical trial
design and implementation, taking part in the UCSF Medical Informatics Training Program. To achieve these
training goals, the candidate has assembled a mentoring team with extensive and complimentary expertise in
clinical trials, epidemiology, and technology-enabled research (Dr. Jeff Olgin, the primary mentor, Dr. Mark
Pletcher, Dr. Veronique Roger), biomedical/clinical informatics and novel data analysis (Dr. Atul Butte) and
heart failure clinical and research expertise (Dr. Liviu Klein, Dr. Veronique Roger, Dr. John Spertus). This pro-
ject seeks to take advantage of our current digital medical era to remotely capture individualized up-to-date
patient data and predict dynamic risk, addressing the unmet need to improve remote heart failure management
and decrease heart failure hospitalization. The project will test and develop tools to predict dynamic heart fail-
ure risk based on real-time data measured in a free-living heart failure population—using a novel smartphone-
based tool—and from patterns in up-to-date EHR data. The specific aims are: Aim 1–! Examine changes in
functional status, measured by serial Self-Administered 6 Minute Walk Test, as a predictor of near-term HF
hospitalization. Aim 2–! Develop a “dynamic” heart failure risk model that incorporates four types of up-to-date
EHR data as it becomes available—including encounters, medication refills/changes, labs and vital signs. This
research is expected to produce two validated methods to estimate dynamic, up-to-date heart failure risk to
enable the provision of earlier, more effective outpatient interventions that decrease hospitalization. This con-
tribution has the potential to improve remote management for heart failure patients, while shifting the clinical
care paradigm to utilize dynamic, longitudinal and free-living data for clinical decision-making. This award will
directly enable a future R01-level randomized pragmatic clinical to trial test the hypothesis that delivery of up-
to-date risk information to outpatient clinicians can decrease future HF hospitalizations. This award will provide
the PI with a unique combination of skills: a strong clinical background, a rigorous clinical research foundation,
advanced analytic skills in machine learning and fluency to utilize health-related technologies to derive insights
and deliver preventive interventions.