Discussion Board

Expand all | Collapse all

‘Nobody is catching it’: Algorithms widely used in hospitals are rife with bias

  • 1.  ‘Nobody is catching it’: Algorithms widely used in hospitals are rife with bias

    Posted 06-21-2021 17:51
    While those of us in this group are aware of how inaccurate and bias-filled AI and algorithms can be, this look at the subject by STAT may help get the word out. How can we identify which ones are and are not in error, contributing to disparate care, or reflecting the bias of its makers?
    https://www.statnews.com/2021/06/21/algorithm-bias-playbook-hospitals/
    Best,
    Helene

      
        Website 
        Twitter 
        LinkedIn
     
        Facebook


  • 2.  RE: 'Nobody is catching it': Algorithms widely used in hospitals are rife with bias

    Posted 06-21-2021 18:39
    If you have not seen this:
    The EPIC Sepsis prediction model is not very good:
    External Validation of a Widely Implemented Proprietary Sepsis Prediction Model in Hospitalized Patients | Critical Care Medicine | JAMA Internal Medicine | JAMA Network

    ------------------------------
    Edward Hoffer
    Massachusetts General Hospital
    ------------------------------



  • 3.  RE: ‘Nobody is catching it’: Algorithms widely used in hospitals are rife with bias

    Posted 06-22-2021 09:10
    Their definition of bias is somewhat non-standard because they include inaccurate targets as a bias. Furthermore, they do not compare statistical prediction models with clinician prediction. The models are usually more accurate than clinicians.

    Harry B. Burke, MD PhD