Notes · Sterile compounding Published June 11, 2026 · Regaldi practice notes

USP <797> documentation: where errors actually happen

By the Regaldi practice · Reviewed by Veronica Taran, Florida-licensed pharmacist · This note is general information about a regulatory framework, not legal or pharmacy advice.

Most sterile-compounding citations are not about technique. They are about paper: the record that was not made, was made late, or does not match the work. USP General Chapter <797> asks for a documented system, and the documentation fails at predictable points.

The recurring failure points

Personnel records. Garbing and gloved fingertip sampling, media-fill tests, and competency evaluations each carry their own cadence. The failure is not the missed test; it is the operator whose qualification lapsed on Tuesday and who compounded on Wednesday, with nothing in the workflow that knew to stop them.
Environmental monitoring. Viable air and surface sampling run on schedules, and the results land in a binder that has no connection to any batch. When a question arrives about a specific preparation, the conditions it was compounded under have to be reconstructed by date arithmetic.
Beyond-Use Dates. The BUD is assigned under the chapter's categories and storage conditions, written on a label, and then lives apart from inventory. The error is dispensing past it, or picking the longer-dated unit while the shorter-dated one expires on the shelf.
Master formulation and compounding records. The chapter requires a master formulation record for batches and for preparations compounded from nonsterile components, and a compounding record for each preparation. The classic findings: records completed after the fact, steps initialed in a block at the end of the day, and a compounding record that cannot be read against the master version it was made under.
The second check. Where verification is required, the same initials appear in both boxes. Everyone knows why: the verifier was busy, and the form did not object.

The pattern underneath

Every one of these is the same failure in different clothes: the record is separate from the work. The log lives in a binder, the work happens at the hood, and a human is supposed to keep them synchronized under time pressure. Training narrows the gap. It does not close it, because the gap is structural.

What a system can prevent, structurally

When the record is the same surface the work runs on, classes of documentation error stop being possible rather than merely discouraged:

An operator whose qualification has lapsed is blocked from the step, not reminded after it.
Environmental samples are bound to the compounding session they cover, so the conditions of any lot are part of that lot's record.
The BUD is computed, surfaced at the point of pick, and escalated as it approaches; First-Expired, First-Out is the order the system offers, not a poster in the break room.
A compounding record opens against a specific master formulation version and cannot be completed with steps unsigned.
A verification in which performer and verifier are the same person is refused at the write.

What remains for training is what training is actually for: technique and judgment. The clerical class of error moves into the architecture, where it cannot be skipped on a busy day.