Systems Medicine

Person looking at a screen

The CRISMA Program on Systems Medicine performs studies at the intersection of quantitative methods and bedside care, leveraging approaches from mathematics, computational biology, and systems engineering. Program investigators use machine learning and other novel analytic techniques to develop and test strategies to monitor, diagnose, and treat patients with severe acute illness. 

Program Members


  • Gilles Clermont, MD, MSc
  • Michael Pinsky, MD


  • Salah Al-Zahati, PhD, RN, FAHA
  • Gregory Cooper, MD, PhD, Department of Biomedical Informatics
  • G. Bard Ermentrout, PhD, Department of Mathematics
  • Milos Hauskrecht, PhD, Department of Computer Science
  • Robert Parker, PhD, Department of Bioengineering
  • David Swigon, Department of Mathematics
  • Shyam Visweswaran, MD, PhD, Department of Biomedical Informatics


Recent and Ongoing Research Projects

Machine Learning of Physiological Variables to Predict, Diagnose and Treat Cardiorespiratory Instability 
This project applies machine learning modeling to complex multivariate high-density data streams from ICU patients in order to identify patients most likely to develop cardiorespiratory instability. This is a multidisciplinary project across computer science, data science, and critical care, leveraging advanced methods to create actionable predictive analytics. Funding: NIH/NIGMS R01GM117622 (PI: Pinsky)

Real-time Detection of Deviations in Clinical Care in ICU Data Streams
This project leverages the entire hospital electronic health record to identify whether therapeutic actions ordered by clinicians for a given patient at a given time are congruent with standard care. Therapeutic decisions that are atypical could indicate deviations from standard care and represent potential medical errors. The combination of advanced systems engineering and machine learning facilitates the production of such alerts in real-time. Funding: NIH/NIGMS R01GM088224 (MPI: Clermont, Hauskrecht, Cooper)

Model-Based Decisions in Sepsis  
This project approaches the pathophysiology of sepsis from a mechanistic computational modeling perspective. Advanced methods of systems dynamics allow quantification of uncertainty originating from imperfect physiologic understanding and sparse, imprecise data. This approach to sepsis endotyping thus happens in the space of mechanisms, rather than being purely descriptive. Funding: NIH/NIGMS R01GM105728 (PI: Clermont) 

Predicting Patient Instability Noninvasively for Nursing Care 
Forecasting cardiorespiratory decompensation is one step beyond early warning systems. This project uses machine learning approaches to detect pre-clinical sign of deterioration in step-down unit patients, with the objective of early escalation of care in patients on the path to decompensation. Which data and how much of it one needs to improve detection and rescue is at the core of this research. Funding: NIH/NINR R01NR013912 (PI: Hravnak)

Development and Evaluation of a Learning Electronic Medical Record System
The LEMR (pronounced ’lemur’) project explores a way the interaction between modern electronic health record systems and caretakers could be improved to facilitate data visualization, information transfer, and clinical tasks. Funding: NIH/LM R01LM012095 (PI: Visweswaran)

Endotypes of Thrombocytopenia in the Critically Ill 
This study uses the electronic health record to investigate the dynamics of thrombocytopenia in an attempt to determine how it affects the development and outcomes of disease. Funding: NIH/NHLB, R21HL133891 (PI: Clermont) 


For information on program activities or to inquire about collaborations and training opportunities, please contact Gilles Clermont, MD, MSc.