AI Clinical Assessment
MediFlow AI turns a patient's vitals and reported symptoms into a structured clinical assessment — differential diagnoses, urgency, red flags, and suggested tests — for a doctor to review, not to act on autonomously.
Every assessment carries the same disclaimer the model is instructed to return: "This is AI-generated decision support and must be reviewed by a licensed clinician." It is decision support, not an autonomous diagnosis — a doctor always reviews and decides.
What Each Assessment Includes
Differential Diagnoses
A ranked list of possible conditions, each tagged with a likelihood of low, medium, or high, based on the recorded vitals and symptoms.
Urgency Level
Each assessment is classified as routine, urgent, or emergency, helping care teams prioritize which patients need attention first.
Confidence Score
A numeric confidence score accompanies every assessment so clinicians know how much weight the model's own output places on itself.
Red Flags
Specific findings the model considers concerning are called out explicitly, rather than buried inside a general summary.
Suggested Lab Tests
Recommended investigations to help confirm or rule out the differential diagnoses, for the doctor to order at their discretion.
How It Works
Structured input
Recorded vitals (blood pressure, heart rate, temperature, etc.) and reported symptoms are compiled with the patient's summary.
Model analysis
The data is sent to Groq's llama-3.3-70b-versatile model, which returns a structured clinical assessment as JSON — not free-form text.
Doctor review
The assessment appears in the doctor's workflow. Nothing is acted on automatically — the doctor evaluates it alongside their own judgment.
Graceful fallback
If the AI service is unavailable, the workflow does not fail silently — it returns an explicit note that a clinician must assess the patient manually.