Assessing and Identifying Clinical Dead-ends in Intensive Care Settings

type of treatment they will provide to patients. With technological improvements and the availability of a significant volume of data, it is increasingly difficult for care providers to properly evaluate and analyze the options available to them. The current health condition of the patient–reflected in the monitored observations which are recorded in EMR–may depend on all the relevant information from all prior observations and selected treatments, not just those most immediate (e.g., the trend of various health measures). This project is focused on developing algorithms that disentangle this history, physician decisions and the outcome (patient survival or death) to assess and identify when actions may lead to irrecoverable negative outcomes. With the identification of these suboptimal actions, correctly associated with the time they are taken, the algorithm can actively discourage doctor’s from repeating such actions when they encounter similar patient histories in the future.

Faculty Supervisor:

Marzyeh Ghassemi

Student:

Taylor Killian

Partner:

Vector Institute

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

University of Toronto