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.

Powiązane ścieżki prowadząceAI do pisania 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.

  1. 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
  2. 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
  3. 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
  4. 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
  5. 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
  6. 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
  7. 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
  8. 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