Assignment book

Twelve hands-on assignments at the level of the course. Do them at your own pace, in the tool that suits you best. No patient data in public chats, use fictional or anonymised cases. Tick off what you've done; notes are stored locally on your device.

Progress: 0 / 12 assignments completed

Differential diagnosis and clinical reasoning

1.
DD for thoracic pain, your own case

Take an anonymised case from the ED of last week (sex, age range, history, ECG finding in text, troponin trend). Have ChatGPT and Claude both produce a top 6 DD using the template from module 4.3. Compare which one ranks a single diagnosis differently and think about why.

2.
Extend DD to “what changes for women?”

Take assignment 1 and ask in the same chat: “how would this DD be different for a female patient of the same age with the same complaint pattern?” Note whether the tool mentions atypical presentation (fatigue, jaw/neck/shoulder, dyspnoea, nausea).

3.
Risk stratification and additional investigations

Give a fictional new-AF case (58yo, BMI 31, HT, T2DM, asymptomatic). Have the model: (a) calculate CHA2DS2-VASc, (b) weigh DOAC options, (c) propose 3 additional investigations. Mark which part you'd adopt without verification and which not.

Letters and patient information

4.
Discharge letter from your own raw points

Take one of your own recent discharge letters, copy just the raw bullet points (anonymised, no PII), and have Claude produce a draft using the discharge-letter template from module 4.4. Check literally: are doses unchanged? No added findings?

5.
Patient leaflet B1, new atrial fibrillation

Using the template from module 4.5, generate a leaflet for new AF after first outpatient appointment (DOAC + rate control). Compare to your current department leaflet. Which 2 sentences would you adopt into the official leaflet?

6.
Letter to GP in 3 tone variants

Take a fictional heart-failure decompensation case. Ask for three versions of the discharge letter: (a) concise-businesslike, (b) collegial-explaining, (c) very short (max 150 words). Which type fits which GP relationship?

Guidelines and literature

7.
Summarise your own guideline chapter with source

Copy a chapter from a public ESC guideline you need this coming week. Paste it verbatim into Claude (1M context) with the template from module 4.7. Add one extra question: “which of your bullets might a fellow cardiologist dispute?”

8.
Literature research for MDT via Deep Research

Formulate a research question for your next MDT (e.g. “what is the current state of imaging in AL amyloidosis, recent meta-analyses 2024–2026?”) and have Gemini Deep Research run it. Assess: how many of the cited sources can you actually open and verify?

Imaging and investigations

9.
Clean up echo report for MDT

Take a released or fictional narrative echo report (around 200–300 words). Use the template from module 4.6 to deliver it in 5 sections with measurements and MDT question. Compare Claude versus ChatGPT on care with numbers.

10.
Translate CMR findings into patient explanation

Write down 4 typical CMR findings in jargon (e.g. “subepicardial late gadolinium enhancement in the inferior LV wall”). Have ChatGPT translate them into B1-English, for the patient. Assess: medically correct + understandable for the patient?

Teaching and safety layer

11.
Case + 5 learning questions for year-2 residents

Use the teaching template from module 4.8 for a topic you're presenting next month (e.g. first-hour hs-troponin in NSTE-ACS, or HFpEF criteria 2025). Test the 5 questions on a real resident. Which 2 questions would you rewrite?

12.
Hallucination hunt in your own field

Deliberately tempt the model into a hallucination: ask for a specific class/level citation from a recent ESC guideline (without providing source). Identify the invented detail. Then ask the same question WITH the chapter pasted into the prompt. Compare the two answers.

Back to overview To evaluation