Read an interaction verdict

Add two or more medicines and every pair is screened. The result is a severity, a mechanism and a management note — each traceable to the source that asserted it. Reading it correctly means knowing what the severities mean and, just as much, what a clean result does not mean.

1
Build the list
In Interactions, search and click to add each medicine as a chip. Two is the minimum; the console screens up to 25 at once. Every pair is checked, so ten medicines is forty-five pairs.
2
Read the headline before the list
The card at the top carries the worst severity found anywhere in the list, and states it as a sentence rather than a colour. That sentence is the verdict; the list below it is the evidence.
3
Work down the pairs
Pairs are sorted worst-first. Each one gives you two things: What happens (the mechanism) and What to do (the management note), with the sources that asserted each and the date the entry was last reviewed.

The five severities

Contraindicatedcontraindicated

Do not give these together.

Severesevere

Serious interaction — one of these usually has to change.

Moderatemoderate

Usable together, but the dose or the monitoring needs to change.

Minorminor

Minor interaction — rarely needs anything changed.

Worth knowinginfo

Worth knowing about, no action usually needed.

One label does not match its value

The lowest severity displays as Worth knowing but its value in the data — and in the API response — is info. Everything else uses the same word in both places.

What the verdict is not

No interaction on file is not proof of safety

The console says this to you on screen and it is the single most important sentence on this page. A clean result means we hold no reviewed interaction for that combination. Interaction databases are incomplete by nature — a pair can be genuinely dangerous and simply not be in any of them yet.

It is not dose-aware. The screen is a property of the pair of substances, not of the regimen. A moderate interaction at a standard dose may be irrelevant at a low one, or serious at a high one, and the verdict does not know which you mean.
It is not patient-aware. Renal function, age, pregnancy and comorbidity all change what a given interaction means clinically. The console screen does not take them. (The API's prescription-screening endpoint accepts coarse, non-identifying bands for exactly this reason — see prescriptions.)
It is not a decision. Every result carries the same line: informational only, not a medical device, not a substitute for professional clinical judgment, verify against primary sources before clinical use.

Why a medicine might not be screened

Interactions are keyed on ATC codes, not on products. If a product in your list has no ATC code on file, the console tells you so explicitly — Not checked — no ATC code on file for X — rather than quietly dropping it from the pair count.

That is worth knowing rather than working around: it means the answer for that medicine is unknown, not clean. If the substance is one you can identify another way, searching by the substance rather than the brand will usually land on a record that does carry an ATC.

Where the data comes from

Two reviewed interaction sources feed this screen — ePHA (Swiss, clinically reviewed) and DDInter — merged on the ATC pair. Each interaction row carries its own provenance, shown as source chips under the mechanism and the recommendation.

Every entry is stamped with a review date. A pair reviewed three years ago is still shown; the date is there so you can weigh it.
Some text is not in English. Where no English translation has been ingested, the source's own wording is shown — often German, from ePHA. We show the source text rather than machine-translating a clinical instruction.
Some rows record no source. Those display source not recorded rather than being attributed to a source that did not assert them.
Missing fields say so. Mechanism not recorded for this pair and No management note recorded — check the product label are real states, not rendering bugs.

Access

Every plan can screen a list. Rows from DDInter 2.0 (severity only) are licensed CC BY-NC 4.0: they are free, unlimited and never use a lookup, and they come with their attribution. Rows with mechanism and management come from a commercially licensed source and correspond to the interactions:write scope (Growth and Enterprise, with lookups left); on other plans they appear as locked placeholders with a count, and their content is never sent. See your plan for what you are on.

The same thing over the API

POST /v1/interactions/check takes a list of drugs — by drug id, ATC code, or identifier system and code — and returns the same severities, mechanisms and recommendations, up to 50 at a time. See the interactions family.