elisaterumi-ai/clinical-agent-skills/timeline-extraction

When extracting a clinical timeline, follow these steps:

1. Identify temporal information

Detect all time-related expressions, including:

• Dates (e.g., 12/03/2023)
• Relative time (e.g., "2 days ago", "last week")
• Event order (e.g., "after admission", "prior to surgery")

2. Extract clinical events

Identify relevant events such as:

• Symptom onset
• Medical visits or admissions
• Diagnoses
• Treatments or procedures
• Medication changes
• Test results

Only include explicitly mentioned events.

3. Link events to time

• Associate each event with its corresponding time reference
• If no explicit time is given, mark as "time": "unknown"

Do not infer missing temporal information.

4. Normalize order

• Arrange events in chronological order
• Use relative ordering if exact dates are not available

5. Preserve original meaning

• Keep event descriptions close to the original text
• Do not rewrite or interpret beyond what is stated

6. Output format

Return the result as JSON:

{
  "timeline": [
    {
      "time": "...",
      "event": "..."
    }
  ]
}

7. Constraints

• Only include events present in the text
• Do not hallucinate dates or events
• Do not infer temporal relationships not explicitly stated
• Do not modify clinical meaning

Keep event descriptions concise and factual