medical-entity-extractor

What This Skill Does

The medical-entity-extractor is a specialized AI agent skill designed to bridge the gap between unstructured patient communication and structured electronic health data. It acts as an automated triage and documentation assistant, parsing complex, messy, and natural language patient messages into actionable, categorized data. By identifying symptoms, medication details, lab results, vital signs, and patient requests, this skill helps reduce the administrative burden on clinical staff and ensures that critical patient information is surfaced immediately. It excels at interpreting the nuances of patient language, such as translating lay terms into medical conditions, identifying temporal relationships, and recognizing common clinical abbreviations.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/binubmuse/medical-entity-extractor

Ensure that your OpenClaw agent instance has the necessary permissions to access message streams or chat history modules, as this skill is designed to intercept and process incoming patient data automatically.

Use Cases

1. Automated Inbox Triage: Automatically categorize patient messages by urgency and topic, flagging P1 cases involving potential medication side effects or severe symptom escalation.
2. Pre-Visit Summarization: Generate concise, structured summaries of patient concerns and recent lab trends before a consultation begins.
3. Medication Management: Monitor patient reports regarding new prescriptions to detect adverse effects early and identify adherence issues.
4. Telehealth Integration: Populate digital intake forms directly from chat-based interactions, saving time for both patients and providers.

Example Prompts

1. "I've been dealing with a dry cough for the last two weeks, it's not getting better. I tried over-the-counter cough syrup but it hasn't helped much."
2. "My glucose levels have been trending high, hitting around 180 mg/dL every morning since I started my new insulin dose on Monday. Should I change the amount?"
3. "I have a sharp pain in my lower back that started yesterday after I was lifting boxes. It's a 7/10 pain. Do I need an urgent appointment?"

Tips & Limitations

• Precision: While highly accurate, the skill should be used as a decision-support tool, not a clinical diagnostic replacement. All extracted data should be verified by a licensed professional.
• Temporal Context: Ensure patient messages include date or time references for the best results in longitudinal symptom tracking.
• Handling Ambiguity: In cases where a patient uses very vague language, the skill will mark the entity as 'unspecified' rather than guessing. If the output appears empty, check the message for clear clinical identifiers.
• Updates: Regularly check the repository for updates to the underlying medical ontology to ensure the model stays current with new medication nomenclature and clinical guidelines.