Interview With A Former SHM Longitudinal Scholar Grant Recipient: Monisha C. Bhatia, MD, JD, MPH And Physicians In Training (PIT) Committee Member
MB: My project investigated the creation a model which can predict whether patients will require post-acute care placement (PAP) following discharge. The project has been over a year in the making, as I first thought of applying for SHM's Longitudinal Scholar Grant during my third year of medical school at Vanderbilt University, and I am now a first-year resident at Jackson Memorial Hospital.
PIT: How has starting your medical training influenced your project? What are some of the implications of your work?
MB: Starting my medical training has only further illustrated the importance of identifying candidates for PAP early during a patient’s hospital admission to help medical teams become more efficient in their discharge planning. Ideally, it can help social workers and case managers identify patients most likely to require their assistance, and it can also aid in shared-decision making with patients when deciding whether post-acute care setting is appropriate for them based on their risk factors.
PIT: Briefly describe how you went about your project and what the primary outcome was.
MB: From an initial set of approximately 78,000 eligible admissions during a one-year period at Vanderbilt University Medical Center, 6,000 of those admissions were needed to generate the model. Using an undersampling approach so as to account for the relative rarity of discharge to post-acute care settings, we evaluated numerous variables that are available in the electronic health record within 24 hours of admission. Our ultimate goal with this 24-hour predictive model was to be able to make this prediction soon upon admission when we can best empower our interdisciplinary team.
PIT: How did the prediction model perform?
MB: The prediction model our team generated had outstanding performance when we tested it against 2,000 observations that were held from the original dataset for calibration, predicting the patient’s destination correctly in about 80% of cases.
PIT: Any future directions?
MB: We are hoping to make a few adjustments to the model to further improve its accuracy, but it is already live in the electronic medical record at Vanderbilt University Medical Center.
PIT: How was your experience at HM19 presenting your poster?
MB: At the SHM poster presentation, I had the opportunity to discuss the project with many researchers from around the country who offered valuable feedback and ideas for applying the project in new ways going forward.