A Northwell Health meta-analysis shows Epic's predictive models don't hold up outside the lab — and the implications for every clinician-builder trusting vendor AI.
More seriously though, your first point about sepsis is closely related to my piece about data quality. Even if data quality stinks, these predictive tools can work but only if the data stinks in the same way as the location where the model was sourced.
I knew there was a reason I liked you. 😁
More seriously though, your first point about sepsis is closely related to my piece about data quality. Even if data quality stinks, these predictive tools can work but only if the data stinks in the same way as the location where the model was sourced.
Love your work!