AI for Writing Medical Referral Letters Quiz

Answers and explanations appear as you take the quiz.

1. What should appear at the start of an AI-drafted referral letter so the specialist knows what is being asked?
   • The patient's full family tree
   • The specific clinical question and its urgency
   • The referring clinician's professional biography
   • A list of billing codes
2. Why is deleting only the patient's name not enough before pasting notes into a non-approved AI tool?
   • Names are the only identifier HIPAA cares about
   • AI tools automatically remove names anyway
   • A rare condition plus exact dates and a location can still identify the patient
   • Specialists need the name in the AI prompt
3. How should the prompt tell the AI to handle information that is missing from the notes?
   • Mark gaps as [MISSING] instead of guessing
   • Fill gaps with typical values for the condition
   • Skip those sections silently
   • Ask the patient directly through the chatbot
4. Why does the guide recommend an organization-approved tool, ideally EHR-integrated or covered by a business associate agreement?
   • Approved tools write longer letters
   • Public chatbots cannot write in English
   • Specialists reject letters from public chatbots
   • Consumer chatbots usually do not provide HIPAA protections by default
5. Under HIPAA's Safe Harbor method, how should dates in the input be handled?
   • Keep full dates so the timeline stays accurate
   • Remove all date elements except the year, using relative timing such as 'three months ago'
   • Convert dates to a different calendar system
   • Replace dates with the day of the week
6. In the physiotherapist example, what information does the orthopedic referral include?
   • The physiotherapist's clinic revenue
   • A request for a second physiotherapy opinion
   • Conservative treatments that failed, with when each was tried
   • The patient's insurance password
7. Why ask the AI for a separate list of assumptions and missing items alongside the letter?
   • So the clinician can verify gaps without guesses leaking into the letter
   • Because specialists require an appendix
   • To make the letter longer
   • So the AI can bill for extra output
8. After reviewing a draft made from de-identified input, what is the final step before sending?
   • Paste the letter back into the chatbot for approval
   • Send it without identifiers so the specialist can guess
   • Delete the chart notes
   • Put the identifiers back in inside the EHR, then sign