Zoom Meeting

Advance Registration Required: Register Here.

Alex F. Bokov, PhD, MS

My training is in the biology of aging. I then became interested in data science and spent 2014-2018 helping stand up PCORNet and ACT infrastructure at my site. Today I am developing better predictors of poor patient outcomes using EHR data. I adapted the Rockwood frailty index to i2b2 data which originated from Epic and Allscripts Sunrise EHR systems. Preliminary results suggest that this frailty index can identify patients at risk for ED visits, hospitalization, and 30-day readmission without needing to know their age (which in the aging field is considered a high bar that few if any biomarker approaches meet). Yet the bulk of my work, and probably that of other data scientists, is extracting and validating the input data. Therefore half the talk will address methods: how to actually turn a database table into risk scores and do so in a reproducible, testable, and reusable way.