Product
Mobile-first medicine: why clinical software must fit in a hand
Mobile is the primary computing surface for most of the world. Assuming a desktop designs for the past.
5 min read
A returned file is a consultation done once, handled twice and paid late. That rework is decided upstream, in the few seconds when someone attaches a code to the act — and a code read off the note is a different object from a code reconstructed from memory an hour later.
On procedure, deadlines, required documents and what any code covers, CNAM is the only source that binds: the question goes to the caisse, not to an article.
Two code systems, one act
A single consultation carries two codes, and treating them as one is the first source of rework. One says what the patient has; the other says what is billed. They are produced at different moments, by different logic, and are not interchangeable.
ICD-10 is the World Health Organization's diagnosis classification, used to name a condition in a standard form (World Health Organization, International Statistical Classification of Diseases and Related Health Problems, 10th Revision, 2019 version). The CNAM nomenclature code is something else: the billing code that appears on the invoice next to the patient share and the CNAM share. A well-coded diagnosis does not make a billable act, and a billable act does not describe the diagnosis.
The two also sit at different plan gates, worth settling before any demonstration. The ICD-10 picker and the ranked ICD-10 AI suggestions start on Pro. Mapping the AI output onto the insurance nomenclature — CNAM auto-coding — belongs to Premium alone. The chain that runs from the coded act to the bordereau is set out under CNAM billing and the bordereau.
Where the code leaves the act
A file comes back when the code no longer matches the act performed. One cause is mechanical rather than clinical: re-entry — the same act written down twice, by two people, at two different moments, with nothing but memory holding the two versions together. Whether it is yours is answerable only from your own returned files.
Follow one act end to end. It is performed and described in the note, in clinical language, with the patient still in the chair; later, at the desk, between two patients, someone reads that note — or a shorthand of it — and turns it into something billable. Nobody has done their job badly. The distance between the two moments is the defect.
We are not going to give you a rejection rate. No percentage read in an article applies to your specialty, your acts and your caisse; the figure that matters is yours, and producing it takes one evening.
Take three months of files. Count the ones that came back and write down why, one line each. Then price them: files returned in a month, multiplied by your own mean act value, multiplied by the extra weeks before payment. That is the only number that should decide anything here. What practice management software in Tunisia removes from that figure is re-entry — not clinical judgement, and not the caisse's procedure.
Ranked suggestions, not decisions
The AI proposes a code and ranks it, the doctor confirms it, and nothing is written until it is signed. That is the whole mechanism, and it is deliberately unspectacular.
On Pro, ICD-10 suggestions arrive ranked, each with its rationale and a confidence score, so a proposal can be accepted or corrected in a tap. The choice happens in the consultation workspace, beside the note itself — one view, not four windows and a paper list, and the act still in front of the doctor.
Every AI output is a draft, visually marked as one until it is approved; there is no automatic finalization and no silent write, and nothing enters a patient record until the doctor signs. When the doctor overrides a suggestion, the override is logged. The reasoning is set out in why the AI never signs.
| Coding layer | Essential | Pro | Premium |
|---|---|---|---|
| ICD-10 diagnosis picker | — | Yes | Yes |
| ICD-10 AI suggestions, ranked, with rationale | — | Yes | Yes |
| CNAM auto-coding to the insurance nomenclature | — | — | Yes |
Read that as a gate, not a promise: on Premium, what reaches the invoice is the nomenclature code the doctor has confirmed.
What we will not publish
We will not publish a list of official rejection reasons — the only list that binds is your caisse's. Any other list is a vendor stating a rule on the caisse's behalf, and that is how a practice ends up defending a procedure nobody imposed on it. What follows is different: the checks a practice can run on its own, without a clinical opinion and without anyone quoting a rule at you.
- Does the identity on the file match the documents exactly, with no abbreviation and no approximate spelling?
- Is the coded act the act that appears in the consultation report?
- Do the dates agree with each other — act, prescription, attachments?
- Is the file complete, or is it waiting on a document nobody actually asked for?
That is practice discipline, not a purchase — and the only part of this that works tomorrow morning whatever software you run.
A reason that appears three times is not an accident — it is a step in your own procedure, and it can be corrected once rather than every time.
Where the software stops
CNAM billing is built in, up to the bordereau. The act is entered at check-in; after the consultation the invoice is generated showing the patient share, the CNAM share and the CNAM code; the bordereau is generated from that invoice. Conveyance of the file to the caisse stays with the practice. We would rather write that limit down than let it be discovered after signature.
The billing layer is gated like the coding layer, and it is worth knowing which half you are buying. Invoicing with manual payments — cash, cheque, transfer, card, insurance — is on all three plans, with payments reconciled automatically against the invoice. Auto-invoicing straight from the consultation starts on Pro, as does revenue analytics. The ledger is everywhere; the automation is where the money is.
The same logic runs through the clinical layer, the subject of AI for Tunisian doctors: the model drafts, ranks and proposes, and a physician stays the author of anything that carries a signature.
Worth remembering
Choose the code once, in the consultation, from the note in front of you, and confirm it yourself; a code re-keyed later is a code that can diverge from the act. Doctopass proposes the code and constitutes the file up to the bordereau; it does not carry that file to the caisse.
Frequently asked
Is ICD-10 the same thing as the CNAM code?
No. ICD-10 is the World Health Organization's diagnosis classification and describes the condition; the CNAM nomenclature code is the billing code that appears on the invoice. Two systems, two purposes — and in Doctopass, two plan gates: ICD-10 on Pro, CNAM auto-coding on Premium.
Does the AI choose the code for me?
It proposes. Suggestions arrive ranked, each with the rationale behind it and a confidence score, and you accept or correct in a tap. Nothing is written to the record until you sign, the draft is marked as a draft until then, and if you override a suggestion the override is logged.
What do I do when a file comes back?
Two things, in order. Correct it and re-send it the way your caisse specifies — that step follows their procedure. Then record the reason where you can count it. And the boundary again, because it settles who does what: Doctopass constitutes the file up to the bordereau, and conveyance to the caisse stays with the practice.
Which plan carries the coding help?
Essential, at 69 DT, covers the record, the calendar, prescriptions and manual invoicing without the AI layer. Pro, at 159 DT, adds the ICD-10 picker and ranked ICD-10 suggestions with their rationale. Premium, at 269 DT, adds CNAM auto-coding to the insurance nomenclature. Per doctor, per month, billed annually, with a 14-day free trial to test it on your own acts; the full comparison is on the pricing page.
The denial you never trigger is the only one that costs nothing, and it is decided in the ten seconds when the code is chosen, not in the appeal. To see what a code picked from the note looks like by the time it reaches an invoice and then a bordereau, request a 30-minute demo — bring the last ten files that came back, they are the only test that settles it.