AI in Medication Reconciliation Quiz

Answers and explanations appear as you take the quiz.

1. What does the guide say pharmacy fill data actually proves?
   • That the patient took every dose
   • That the prescriber still wants the drug
   • That the product was dispensed and picked up, not that it is being taken
   • That there are no over-the-counter products
2. According to the guide, what is a best possible medication history built from?
   • At least two sources, usually including a structured patient interview
   • The EHR list only
   • Only the most recent discharge summary
   • Insurance claims from the past decade
3. Why does the system normalize drugs to RxNorm and parse the directions?
   • To bill insurance correctly
   • So differently written but equivalent regimens, such as 500 mg x2 BID versus 1000 mg BID, are recognized as the same
   • To translate labels into other languages
   • To remove all brand names from the chart
4. A cardiology note says metoprolol was stopped for bradycardia, but it remains on the discharge list. Which AI capability catches this?
   • Barcode scanning
   • Refill synchronization
   • Image recognition of tablets
   • Extracting stop or hold intentions from notes
5. Which discrepancy counts as an error in the guide's framework?
   • An unintentional discrepancy
   • A documented intentional change
   • Any change from the home list
   • A drug the patient buys over the counter
6. Which misconception about AI reconciliation does the guide correct?
   • That transitions of care carry risk
   • That patients forget inhalers
   • That the software itself completes reconciliation, when it only prepares evidence for a clinician
   • That notes contain clinical decisions
7. Why should LLM-based extraction return the exact source text for each claim?
   • To make the output longer
   • So reviewers can quickly verify each claim and hallucinated medications are caught
   • Because RxNorm requires it
   • To avoid storing notes
8. Which evaluation approach does the guide recommend?
   • Counting how many alerts the tool generates, with more being better
   • Asking patients whether they like the interface
   • Comparing to last year's billing codes
   • Measuring sensitivity for important discrepancies and false flags per patient against a pharmacist-built BPMH