Scales and Scores

Owned by John Svirbely
Last updated: Dec 17, 2020 by Davide Sottara
1 min read

“Scales and Scores” are simplified classification models designed to approximate complex clinical assessments/predictions that can be computed using simple pieces of data, and simple computations

These models pose several subtle challenges

  • Computability of the logic - e.g. “pain scale” vs “HAS-BLED score“

  • Availability of the all the (required) data / how to handle partial inputs

  • Precise semantics of the output concepts

  • Precise semantics of the input concepts

    • Refined definitions of how to query for the input concepts, based on the cohort that the model was trained/validated on