Our work sits at the intersection of ICU data, machine learning, and clinical trials. The goal is practical: build tools and methods that help clinicians make better decisions for critically ill patients.
The ICU generates thousands of data points per patient per day, but most of it sits in the EHR unused. We build models and NLP pipelines that try to change that: readmission risk scores, early deterioration alerts, and AI-assisted substance misuse screening that runs at the bedside. The goal is practical: tools that help clinicians make better decisions with data they already have.
Good ICU research needs diverse data from many hospitals, but sharing raw patient records is a non-starter for privacy. The CLIF Consortium addresses this with a shared, open-source data standard. Rush is a founding site in the network, which now covers 12 institutions, 62 hospitals, and over 808,000 ICU patients. The idea is federation over centralization: models and results move across institutions, raw data never does.
Explore CLIF Documentation (opens in new tab)ICU care varies a lot between hospitals and between patients, and it is often unclear whether those variations matter. We run pragmatic trials embedded in everyday ICU workflows and use causal inference methods (target trial emulation, heterogeneous treatment effect modeling) to figure out which patients benefit from which interventions. Current work includes individualized oxygenation targets and ventilation strategies.
Models that estimate which patients are likely to benefit from a given intervention, for example personalized oxygenation targets for patients on mechanical ventilation.
Using observational ICU data to approximate the results of trials that would be impractical or unethical to run, a way to get causal answers from the data we already have.
Critical illness doesn't respect disciplinary boundaries. At Rush, we work with investigators in emergency medicine and respiratory therapy on shared trials and data science, alongside biostatisticians, data scientists, and clinical informaticists. The questions that matter at the bedside don't fit neatly into one department, so neither do we.
If your work touches ICU data, clinical trials, or federated research, we would like to hear from you.
Get in Touch