Assess whether a risk model is calibrated for this population (0f14d5)

Situation

Outcomes Review work in a unit with a new early-warning model now turns on a risk model is calibrated because a model vendor claiming AUC that the local validation cannot reproduce put risk-adjustment model calibration plot in play. Outcomes Review work in a unit with a new early-warning model now turns on a risk model is calibrated because a model vendor claiming AUC that the local validation cannot reproduce put risk-adjustment model calibration plot in play; quality-improvement physician should say what risk-adjustment model calibration plot proves for Healthcare.

Decision

Quality-improvement physician in a unit with a new early-warning model must choose Proceed under protocol / Pause the pathway / Escalate safety review / Hold using risk-adjustment model calibration plot after a model vendor claiming AUC that the local validation cannot reproduce.

Hypotheses to test

• The population in risk-adjustment model calibration plot is the one a model vendor claiming AUC that the local validation cannot reproduce named, so Proceed under protocol follows for this Outcomes Review file.
• The population in risk-adjustment model calibration plot is adjacent only to a model vendor claiming AUC that the local validation cannot reproduce; Pause the pathway is the honest Healthcare call.
• A unit with a new early-warning model already contained a model vendor claiming AUC that the local validation cannot reproduce before risk-adjustment model calibration plot arrived; no new Outcomes Review path.
• Provenance on risk-adjustment model calibration plot after a model vendor claiming AUC that the local validation cannot reproduce is broken; do not pick Proceed under protocol or Pause the pathway yet.

Analysis required

• Assess patient-safety and HIPAA / minimum-necessary implications of a risk model is calibrated.
• Quantify who is harmed if risk-adjustment model calibration plot is wrong.
• Separate a documented exception from an OCR-relevant gap in a unit with a new early-warning model.
• For this Healthcare Outcomes Review file, read risk-adjustment model calibration plot against a model vendor claiming AUC that the local validation cannot reproduce and write the one fact that would move a risk model is calibrated for quality-improvement physician.

Recommendation