Machine learning algorithm for predicting the risk of ICU readmission within 48 hours of initial ICU discharge

  • Clinical Condition
    • ICU patient could be discharged prematurely.
    • Inadequate levels of monitoring and interventions.
    • Preventable clinical deterioration, increased mortality, morbidity and healthcare costs.

  • Current Practice
    • Care team makes all readmission decisions.
    • NO generally accepted decision algorithm for re-admissions to the ICU at MGH.
    • Decision parameters are subjective and variable.

  • Solution
    •  We propose to develop a machine algorithm based on MGH and BWU ICU data from 61,184 of patients from the last five years.
    • Determine the predictors of readmission.
    • Build a ML algorithm to predict patients at risk for readmission based on the vast previous data.