Tencent Health AI Clinical Assistant: Medication Safety

Problem to solve

Medication questions in clinical care are rarely one-line lookups: the same Ibuprofen dose can carry different risks for an adult, an older patient, an anticoagulated patient, or a patient with CKD. This skill is written for clinicians, clinical pharmacists, and residents, providing CDS-style medication reference to reduce fragmented searches through labels, guidelines, and interaction resources. It does not replace the final prescribing decision, but organizes the usual safety reasoning into stable outputs: conclusion, dosing, mechanism, contraindications, monitoring, and red flags.

How it works

• Context reuse: maintains a separate medication context for age, sex, weight, pregnancy/lactation, allergy history, diagnosis, current medications, and renal/hepatic labs, then reuses known facts across turns without re-asking.
• Clinical phrasing: leads with prescriber-style notation such as Amoxicillin 0.5g po q8h × 7d; simple questions stay concise, while complex reviews expand into conclusion, mechanism, evidence, and intervention suggestions.
• Risk layering: raises a standalone ⚠️ **Clinical red flag:** block for NSAID plus anticoagulant bleeding risk, QT prolongation, serotonin syndrome, pregnancy contraindications, severe DDIs, and other high-risk combinations.
• Attachment parsing: reads medical records, prescriptions, and lab reports to extract demographics, allergies, history, medications, and abnormal values; the result is acknowledged before being folded into later medication reasoning.
• Standards-library option: the free tier uses general pharmacology knowledge and guidelines; with Tencent Health ACA access, it can add DDI database coverage, label-level incompatibilities, special-population risk, and off-label detection.

Boundaries

This is useful for dosing, DDI, contraindications, adverse effects, pediatric weight-based dosing, renal dose adjustment, and prescription review. It is not intended for drug pricing, inventory, insurance coverage, procurement, or commercial medical decisions. Unknown brand names are only inferred from the name and should be verified by active ingredients; final clinical decisions still require the treating clinician's judgment and complete patient context.