AI Clinical Decision Support for Physicians Quiz

Answers and explanations appear as you take the quiz.

1. What mainly drives alert fatigue with drug interaction alerts?
   • Alerts are displayed in too small a font
   • So many alerts are clinically irrelevant that clinicians learn to override them, including the rare important one
   • Pharmacists see the alerts before physicians do
   • Alerts only fire after the order is signed
2. What did the 2021 University of Michigan external validation of the Epic Sepsis Model find?
   • It detected nearly every sepsis case with few false alarms
   • It worked only on pediatric patients
   • Discrimination was much weaker than advertised (AUC about 0.63), and it missed about two thirds of sepsis cases
   • It was withdrawn by the FDA as an unapproved device
3. A renal dosing rule suggests a lower dose inside the order without a pop-up. What kind of delivery is that?
   • Passive or non-interruptive delivery
   • An interruptive hard stop
   • A predictive machine learning alert
   • A CDS Hooks patient-view card sent to the patient
4. In CDS Hooks, what does an external CDS service send back to the EHR?
   • A replacement copy of the patient's chart
   • Billing codes for the encounter
   • A signed prescription
   • Cards that give information, suggest an action or link to a SMART on FHIR app
5. Why run a predictive model in silent mode before go-live?
   • To measure how well it performs on local patients before any alerts reach clinicians
   • To avoid needing EHR integration
   • To satisfy the five rights framework automatically
   • To train clinicians to ignore it
6. Which of these is part of the CDS "five rights" framework?
   • The right price
   • The right vendor
   • The right person
   • The right insurer
7. What kind of CDS did the 21st Century Cures Act mainly exclude from FDA device regulation?
   • All software used in hospitals
   • Any tool built by an EHR vendor
   • Only tools that use machine learning
   • Tools whose basis a clinician can independently review
8. Which governance metric most directly shows whether an alert is creating noise?
   • The number of FHIR resources in the EHR
   • Override rate together with override reasons
   • Server uptime
   • The alert's color scheme