For clinicians
Written for the person whose patient arrives holding this.
Every line below is either a design constraint we hold ourselves to or a citation you can open. None of it is advice, and where the product's position is unsettled it says so rather than asserting it to you.
- What it never produces
- No diagnosis, no candidate condition, no prognosis, no dose advice, no tapering schedule, and no instruction to stop or change a prescribed medicine. A finding resolves to a question for the prescriber, which is the only output shape the product has.
- Attribution prompting
- Attributional re-examination is this product's premise, and the literature is explicit that it interacts badly with fluctuating insight. We do not claim to have removed that. The answer is steady mode: a self-selected framing that drops the attributional loop and keeps the shelf, the reminders and the record. It is chosen by the user and never inferred from their data, because inferring it would be the diagnosis we refuse to make.
- Withdrawal versus relapse
- The product never declares which is happening. It publishes the differentiators (dizziness, vertigo and nausea discriminate; depressive symptoms do not) and stops there. No schedule, no rate, and no comparator between stopping and continuing, because arranging two figures around that decision would be arguing it.
- No scoring
- Nothing produces a number about the person. No severity index, no wellness score, no risk stratification. Daily items are stored unscored and never summed.
- Crisis is siloed
- No medication data renders on any crisis surface, ever; a shelf is a means inventory. That silo is a precautionary call with no empirical data behind it and we would rather say so. What does have data: clinician-delivered safety planning cuts suicidal behaviour, RR 0.57, NNT 16, and no unguided self-completed digital version has been shown to carry that effect, which is why the app does not build one.
- Escalation is published
- We do not promise never to contact anyone. We publish what triggers escalation, because the absolute version of that promise is the one thing that cannot be honoured, and an earlier version of this product made exactly that mistake.
- Pharmacovigilance
- Suicidality, akathisia, serotonergic markers and sustained missed doses are monitored continuously against fixed thresholds, exempt from the pruning that applies to everything else.
- Not an instrument
- Five daily items: mood, interest, sleep quality, intentional movement, medicine taken. Deliberately not PHQ-9, GAD-7, WHO-5 or DASS-21, because administering a scored instrument is screening. Nothing is summed, nothing is compared to a cut-off, no item maps to a diagnostic criterion.
- Never called treatment
- A 2025 systematic review found no robust evidence that mood tracking alone treats depression. The check-in is a record, and that is the only thing it is ever called.
- Reassurance-seeking
- Designed against becoming a checking compulsion: the treatments that work for health anxiety work by removing reassurance, not supplying it. Medical reassurance fails to reduce anxiety long-term in 12 of 22 studies reviewed, and the review reporting that opens by asking whether it is established data or clinical lore, which is worth knowing.
- The friction is under-evidenced
- And the app says so on the screen. Our own run rated the mechanism medium confidence and marked the specific budgets and delays insufficiently evidenced, sourced to clinical writing rather than trials. We built it anyway, labelled.
- Search escalation
- Symptom searching escalates from a benign symptom to a severe rare diagnosis inside a single session, and CSS-15 scores climb with search frequency in a dose-response way. Search here is by medicine name or active ingredient only, never by condition and never symptom-to-diagnosis.
- Base rates do not reassure
- Salience reduced base-rate neglect in healthy participants and failed entirely in people with elevated health anxiety. The frequency design is not built on the assumption that a reassuring denominator does the work.
- Amplification
- Daily physical-symptom rating is banned outright, because somatosensory amplification from behaviour logging is a real and untested hazard for this population.
- No streaks
- No streak counter, no completion percentage, no missed-day language. "Not enough days yet" renders as an ordinary row with a reason.
- What arrives in your room
- A short sheet: the medicine, the labelled effect, the evidence tier, the direction-of-causation reading, and the source with its label version and date. It is a question, not a request, and it does not tell the patient what you should do.
- Regulatory position
- Built to sit outside the medical-device definition: no diagnosis, no triage, no prognosis, no symptom-based routing to care, no personal dosing. Whether that classification holds is unsettled and goes to regulatory counsel before it goes to code. We are not asserting it to you as decided, and a single regulated feature would pull in the whole application.
- Advertising rules
- Search by medicine name or active ingredient only; search by condition is discovery and recommendation, which the TGA treats as advertising a prescription medicine. Results come back alphabetically, which the Advertising Code requires.
- Coverage leads the summary
- Any summary a patient brings you opens with how many days are missing and which ones, ahead of any number, because missing days are informative and the averages they leave behind are biased optimistic.
- The change flag has no literature behind it
- The medication-change flag fires on more than one personal weekly standard deviation sustained for two weeks. That threshold is constructed, not cited; no published threshold exists. Read it as "something moved", not as a finding.
- eScript tokens
- The QR carries a Delivery Service Prescription Identifier, not the clinical detail; identifying the medicine needs a live call to the Prescription Delivery Service. Anything claiming to read a script offline is guessing, and we say so on the screen where a token is shared. Populating a shelf properly runs through the Active Script List, which depends on a commercial intermediary whose terms are not yet established.
- Frequency
- From approved product information, dated to the label version. Numbers only, both halves at equal weight, icon array under 1%; the verbal band is never shown. Australian PI is PDF-only with no bulk feed, so frequency is extracted document by document and that bounds how much of it exists.
- Interactions
- Surfaced with the tier and the source, never with a severity score. The never combine versus monitor distinction is preserved, because flattening it loses the only thing that matters. No free drug-interaction API survives, so this data is bought or extracted. A stated dependency, not a solved problem.
- Burden scales
- An anticholinergic meter is planned on CRIDECO, CC BY 4.0, whose licence covers its 217-drug table, subject to verifying that licence directly, which our own research flagged as the first thing to check before building on it. No sedative-load meter: the one static table is all-rights-reserved and the most complete modern catalogue is CC BY-NC. Serotonergic load is a flag rather than a meter, and there is no stimulant meter, because our search found no validated numeric scale, which is a different claim from none existing.
- Recalls and batches
- Batch matching depends on extracting the batch from prose, and often it simply is not there. Where the batch is unknown, the app says the batch is unknown.
- Complementary medicines
- On the same shelf as prescriptions. St John's Wort beside an SSRI is the case the product was designed around, and separating the two shelves is how that gets missed.
Every report is public. If you think one of them is wrong, we would genuinely like to know which part. How solid the sourcing is →