A medication database for clinic prescribing that your whole team searches the same way

The medicine search fires as soon as you type two characters (debounced so it doesn't hammer the server on every keystroke). One database function returns a single, unified result set: it matches the start of the brand name first, then the active ingredient, then, once you've typed a bit more, adds a fuzzy match that forgives a typo like 'Azithral' vs 'Azithro'. Each row shows the brand with its ingredient, strength, dosage form, and manufacturer so you can tell two similar products apart at a glance, and discontinued products are filtered out entirely.

Search results are reordered for your clinic specifically. A pediatric clinic and a cardiology clinic write completely different medicine mixes, so a one-size global 'most popular' order would be wrong for both. When the search runs with your clinic's context, the medicines your clinic prescribes most surface to the top; a clinic with no history falls back to plain alphabetical order. The ranking is rebuilt quietly once a day from your clinic's own recent prescriptions, so it keeps learning your habits without slowing a single save.

Every doctor, nurse, and front-desk user searches the same underlying medicine list, so the same brand resolves to the same entry no matter who's typing. When you pick a medicine, the prescription stores its real catalog id, not just the text, so the same medicine added by two doctors is recognised as one thing downstream, for recents and for the allergy check. That consistency is what makes 'most prescribed' ranking, repeat-prescription recents, and clean records possible in the first place.

If a medicine you prescribe isn't found, you add it inline without leaving the prescription: brand name, plus optionally dosage form, primary ingredient, and strength. Your additions are private to your clinic: another clinic's entry (or typo) never shows up in your search, and yours never leaks into theirs. A clinic administrator can later promote a custom medicine so it joins the shared list, and from then on it autocompletes for your team like any other entry.

The payoff of one consistent list is that the medicine you chose carries a real catalog id into the record. That id is what lets the patient's recorded allergies be checked against the exact medicine, not a guess at loose text. The match also falls back to the brand name and the active ingredient, with word-boundary-aware logic so a short fragment like 'Ace' won't false-alarm on 'Paracetamol' while a real drug-class fragment like 'Cillin' still catches the penicillin family. This is an operational prompt against what's already in the patient's own record, not a clinical decision-support engine.

Before you've typed anything, the search shows the medicines you most recently prescribed in this clinic, de-duplicated by name, so a repeat script is a tap rather than a re-type. Because each recent carries the medicine's id where the original recorded one, picking from recents flows through the same allergy check and the same consistent record as picking from a fresh search. And when a search has a lot of matches, the dropdown stays capped and readable with a 'refine your search' nudge instead of dumping hundreds of rows.