Please consider downloading the latest version of Internet Explorer
to experience this site as intended.
Tools Search Main Menu

Calendar

Data Science Summer 2021 Colloquium Series: Louis Ehwerhemuepha

On the Deployment of Predictive Models in Health Care – Case Studies of Hospital Readmission and ED Severe Sepsis and Decompensation

This virtual talk is a featured presentation of the National Science Foundation Research Experience for Undergraduates (NSF REU) on Computational Methods for Understanding Music, Media, and Minds. The talk is free and open to all faculty, staff, students and community members.

Abstract: Predictive models can only begin to make meaningful impact on deployment. In this research talk, we will discuss the development of 2 predictive models, processes for adoption of the models, deployment, associated clinical interventions, and measurement of corresponding impact on care delivery.

We measured clinical impact of the deployment of a readmissions model at CHOC Children’s health system in Orange, California. The 30-day all-cause readmission rates during the periods before and after model deployment and clinical interventions (and corresponding 95% CI) were 0.125 (0.122, 0.128) and 0.111 (0.108, 0.115) respectively. Statistical process control charts indicate non-random reductions in readmission rates over time with an estimated $2,673,264 (95% CI: 2,612,431, $2,735,364) in healthcare savings. 

In a similar way, we deployed a real-time tool for severe sepsis and decompensation to estimate risk of these undesirable outcomes at ED triage. We deployed the model and discuss some practical considerations and some oddity in selecting model performance metrics in the presence of extreme class imbalance.

Bio: Louis Ehwerhemuepha, PhD is a data scientist at CHOC Children’s health system in Orange, California. He has led the technical aspects of the development of CHOC’s data science program. He focuses on the development of wide ranges of statistical, machine-learning, and artificial intelligence applications as well as corresponding deployment of these models using local computing resources or advanced parallel distributed cloud computing. He has published multiple papers in the Nature Research journals such as Scientific Reports and Pediatric Research, with more than 12 publications in 2020, and more than 6 additional papers as of May 2021.

View the recording here

Dial-In Information

Register at the following link: https://rochester.zoom.us/webinar/register/WN_U_LpHgAuReaXc650au8HLg

Wednesday, July 21 at 12:00pm to 1:00pm

Virtual Event
Event Type

Lectures and Talks

Audience

General Public, Faculty, Students, Alumni, Staff, Health Care Professionals, Graduate Students/Postdocs

Tags

health, data science

Cost

Free

Group
Goergen Institute for Data Science
Contact Phone or Email

sylvia.francisco@rochester.edu

Subscribe
Google Calendar iCal Outlook

Recent Activity

You're not going yet!

This event requires registration.