StewardIQ AI, Antibiotics Only When Needed
Helps clinics without a lab tell whether a sick child needs antibiotics, so fewer are handed out just in case.
In a lot of clinics around the world, a child comes in with a fever and there is no lab. The clinician cannot tell a bacterial infection from a viral one, so they prescribe antibiotics just in case. That is the single biggest reason antibiotics are losing their power.
The problem
- Antibiotic resistance directly caused 1.27 million deaths worldwide in 2019, and played a part in nearly 5 million more
- Almost three-quarters of that falls on poorer countries
- If nothing changes, resistant infections are expected to contribute to around 39 million deaths between 2025 and 2050
- Clinicians are not being careless. Without a test, prescribing is the safe choice for the child in front of them
- Most decision tools assume good internet and a lab. Neither is a given
How it works
The clinician enters what they can actually measure. Age, temperature, breathing rate, how the child looks, how long it has gone on, what they can see and hear on examination.
It works through the standard clinical guidelines for a child with a fever and gives a clear read on whether this looks bacterial, and how confident it is.
It always says when it is unsure. Unsure means see a clinician, not skip the antibiotics.
It runs entirely on the device, with no internet. That is not a nice extra. A tool that stops working when the connection drops does not get used.
The safety line
Danger signs are handled by fixed rules, not by the AI deciding. Trouble breathing, not drinking, unresponsive, convulsions, a rash that does not fade. Any of those means refer now, whatever else the tool thinks.
It can always push towards more caution. It can never push towards less.
What it changes
Where tools like this have been used well, unnecessary antibiotic prescribing drops sharply while the number of genuinely sick children who get treated stays the same or improves. That second half matters more than the first. A tool that cuts prescribing by missing sick children is worse than no tool at all.
Built with: an on-device model that needs no connection, Claude for the explanations, and syncing that catches up whenever a signal appears.
Interested in building something like this?
Get In TouchRelated Projects
MediScribe AI, Writes Up the Visit For You
Listens to the visit and writes the note, the codes and the follow-up instructions. Gives a doctor about two hours back a day.
ClinicalNotes AI, Notes and Codes, Drafted
Drafts the doctor notes from the visit and suggests the billing codes. A doctor reads it, fixes anything wrong, and signs.
MedicalCoding AI, Right Codes, Fewer Denials
Suggests billing codes from the notes, points out what the note is missing, and warns you which claims are likely to bounce.