COVID+DS: Using data science to optimize scheduling elective procedures in the time of COVID
As the COVID-19 pandemic hit Durham County, the Duke University Health System (DUHS) responded by postponing most non-urgent (i.e. elective) operations. As we developed a better sense of the hospital resources needed to manage COVID infections, hospital leadership became more comfortable with re-commencing these elective procedures. However, in order to be prepared for any potential local surges and to be able to properly manage potentially limited resources such as ICU beds and ventilators, leadership wanted to have an understanding of what resources a given procedure may require.
Enter Data Science!
A cross-disciplinary team of data scientists, informaticists and clinicians came together to develop and implement a set of predictive models. The models utilize EHR data on surgical cases scheduled over the next 30 days and generate predictions for: anticipated length of stay, anticipated ICU length of stay, need for a ventilator and need to be discharged to a skilled nursing facility. The models have been integrated into a Tableau Dashboard and are refreshed every morning at 6am so that surgical leadership can decide which cases can proceed safely and which may not be rescheduled for another day.
In this talk I discuss the team and the process that brought this together in less than 10 weeks.
This session is part of the Duke+Data Science (+DS) program virtual series on COVID-19 + Data Science.