Quiz połączony z przewodnikiem · Ciężko Poziom
Sztuczna inteligencja do pisania quizu podsumowującego wypis ze szpitala
Testuje uzgadnianie leków w wersjach roboczych sztucznej inteligencji, sprawdza, dlaczego pominięcia są najbardziej ryzykownym błędem, a także listę kontrolną przeglądu lekarza dotyczącą podsumowań wypisów.
Pytanie 1 z 8
Which type of AI discharge summary error does the guide call most dangerous, because nothing on the page looks wrong?
Podgląd każdego pytania8 pytania
Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.
- Which type of AI discharge summary error does the guide call most dangerous, because nothing on the page looks wrong?
- Typographical errors
- Omission of a complication, medication change or pending result
- Formatting inconsistencies
- Summaries that run too long
- In the lisinopril example, what went wrong in the AI draft?
- It prescribed a new drug the patient never received
- It doubled the dose of lisinopril
- It listed lisinopril under allergies
- It listed lisinopril as continued when it had been held for kidney injury
- Why might an AI draft present an early differential diagnosis as the final diagnosis?
- Working hypotheses get copied forward through daily notes, so they look current
- The AI prefers rare diagnoses
- Final diagnoses are never documented in hospitals
- The AI reads only the first note of the stay
- According to the guide, where should discharge medications in the summary come from?
- The model's memory of typical regimens for the diagnosis
- The home medication list recorded at admission
- Structured discharge orders in the EHR, not regenerated by the model
- The patient's recollection at discharge
- Which item belongs on the physician's review checklist for pending results?
- Delete pending results to keep the summary short
- List every pending result and who is responsible for following it up
- Wait until all results return before discharge
- Mark pending results as normal
- Why are word-overlap scores such as ROUGE not enough to judge AI discharge summaries?
- They can only be calculated for English text
- They are too expensive to compute
- They require patient consent
- They reward text similar to a reference but don't show whether key facts are missing or invented
- How should the medication section of a discharge summary classify each drug?
- As new, changed or stopped, with reasons
- By the color of the pill
- Alphabetically, without noting changes
- By cost to the patient
- Who remains accountable for an AI-drafted discharge summary once it is signed?
- Dostawca sztucznej inteligencji
- The hospital's IT department
- The physician who reviews and signs it
- The patient's pharmacist