AI to Predict Vulnerable Older Patients

Data to Enable a Learning Health System
In Progress
Artificial Intelligence, Prediction
April 2021 - September 2022


The overall objective of this study is to develop a prediction tool that processes primary care electronic medical record (EMR) data, and when implemented, helps primary care clinicians proactively identify older patients at risk of deterioration to direct limited primary care resources to these patients. In phase 1, we will develop and validate EMR search strategies to quantify emergency department (ED) visits and hospitalizations. In phase 2, we will apply the search strategies from phase 1 to develop and test a variety of models that use structured and unstructured EMR data to predict non-elective admissions to hospital.


Primary care clinicians currently do not know which older patients with multimorbidity will imminently require the ED or hospital. Accurately predicting which older patients with multimorbidity are at risk of requiring the ED or being admitted, and implementing effective interventions that avert these events, would improve health outcomes, reduce costs, and improve experiences of care. This work will create a foundation for future research on using EMR data to help direct primary care resources.

Team Members

Contact Information