About The Program
Providing greater insights and solutions.
Predictive analytics has enormous potential to enhance clinical outcomes, patient experience and overall healthcare delivery. Building machine learning models based on past patient data enables us to plan for what may happen in the future. As a national leader in data science, the University of Chicago Medicine draws on many departments across campus with expertise in the development and deployment of these methodologies.
In 2019, the Analytic Interventions Unit (AIU) was formed to identify significant clinical applications of predictive analytics and support the development, deployment and evaluation of these initiatives at UChicago Medicine.
The AIU fosters innovation through provisioning of accurate and actionable predictive tools that improve clinical decision support, and ensures safe and effective application that protects quality of care for our patients.
A Diverse Portfolio
Current Focus Areas
The AIU is currently supporting a diverse portfolio of predictive analytics initiatives, either in-development or in-production, focused in the following clinical and operational areas:
- Sepsis risk (adult and pediatric)
- Hospital readmission risk
- Inpatient length of stay
- Inpatient risk of falls
- Ambulatory risk of no-show
- Scheduling optimization
Guiding the Project Team
Model Deployment Approach
The AIU works with clinical and operational leaders across the UChicago Medicine health system to identify priority areas where predictive interventions may help advance patient care. A clinical or operational sponsor works with the project team to guide the project: prioritization; development or configuration of the predictive model; incorporation within the clinical workflow; validation of model data inputs and outputs; and integration of the model into the health system’s technical architecture.
The AIU supports each stage of the model development and deployment process outlined below and coordinates resources needed for the successful implementation and evaluation of each predictive initiative:
Determine and Align on Clinical Priorities
- Identify priority areas where predictive data may provide clear steps for clinical intervention
- Determine whether implementation aligns with the goal of delivering safe, effective care
Source and Develop (for models that have not yet been built)
- Identify external sources for model procurement or development
- Develop model internally based on UCM data
Review and Validate
- Review the scientific validity and operational value of each model to be implemented
- Provide feedback and acceptance or concerns for deployment
Test, Deploy and Communicate
- Conduct integration / interface development
- Validate within Epic or another deployment system
- Identify appropriate security roles and access
- Shift model to full production
- Communicate and train stakeholders
Monitor Results and Effectiveness
- Track on-going results and clinical effectiveness metrics
- Conduct periodic review and evaluation of clinical outcomes as a result of predictive data
How Things Work
The Process
If you have questions, please reach out directly to the HDSI team.
AIU Team Members
Providing strategic direction and operational support.
The AIU is comprised of leaders in clinical quality, informatics, data and analytics, and information technology who provide strategic direction and operational support needed for implementation of predictive models across the healthcare system.
Associate Chief Medical Information Officer
Associate Professor of Medicine and Pediatrics
