Predicting Readmission Risk after Orthopedic Surgery

My colleagues and I from the Clinical Research Informatics Core at Penn Medicine gave poster presentations at the Public Health session of the Symposium on Data Science and Statistics last week.

Here's the abstract:

Our project examined hospital readmissions after knee and hip replacement surgeries that took place within the University of Pennsylvania health system. We used a variety of information available within patient electronic health records and an assortment of machine learning tools to predict the risk of readmission for any given patient at the time of discharge after a primary joint replacement surgery. We faced challenges related to missing data. We used a number of different machine learning models such as logistic regression, random forest and gradient boosted trees. We also used an automated machine learning pipeline tool, TPOT, that uses a genetic algorithm to search through the machine learning model/parameter space to automatically suggest successful machine learning pipelines. We trained multiple models that predicted readmissions better than the existing clinical methods, with statistically significant increases in AUC over the clinical baseline. Finally our models suggested a number of features useful for readmission prediction that are not used at all in the existing clinician model. We hope our new models can be used in practice to help target patients at high risk of readmission after joint replacement surgery, and to help inform which interventions may be most useful.

SDSS Poster Presentation

Machine Learning for Healthcare

Yesterday I gave a dev talk at Philly Tech Week on machine learning for healthcare, slides embedded below.

Description: "How are machine learning and data science being adopted in healthcare? From diagnostics, risk predictions, and more, this session will provide an overview of machine learning applications using electronic health records, walk through the process of how a model might be trained and used, and discuss methods for improving interpretability to augment medical decision-making."

Here's a link to the slides of you want to see my notes.

I think the talk went pretty well. In fact, I think I am actually a pretty good speaker, although I'm not sure how much I get out of speaking personally. The talk was pretty well attended, and I did receive a lot of positive feedback, so hopefully I inspired some people in healthcare or machine learning in some way or another.