Coding the Event, Not the Diagnosis You Assume

Part of Pharmacovigilance and Drug Safety (PV) Training: Coding and the Narrative.

About 1 minutes. Free, in your language, no account needed.

The most consequential coding decision

When a reporter describes several symptoms, do you code each symptom, or a single diagnosis that would explain them?

The general rule: if the reporter, or a clinician in the record, has stated a diagnosis, code the diagnosis. If they have described symptoms and stated no diagnosis, code the symptoms.

Why this way round

Coding a diagnosis nobody made puts a medical conclusion into the database that no clinician reached. Later, it is indistinguishable from a real diagnosis.

Splitting a stated diagnosis into its component symptoms has the opposite failure: the diagnosis disappears from the counts, and an analysis looking for that condition will not find this case.

The reasonable middle

Where symptoms are clearly part of a single reported diagnosis, code the diagnosis and record the symptoms in the narrative. Where a diagnosis is provisional or disputed, code what is supported and describe the uncertainty.

Where you are unsure, ask. Coding conventions are the kind of thing every organization has a written convention for, and following it matters more than picking the theoretically better option unilaterally.

Consistency is worth more than perfection

Two defensible coding conventions applied consistently produce usable data. The better convention applied inconsistently does not.

If a case processor codes symptoms and their colleague codes diagnoses, the same clinical event enters the database as three counts from one of them and one from the other, and no frequency derived from it means anything.

Recoding happens

Dictionary versions change, and terms move. Databases are periodically recoded to the current version. This works only because the verbatim text was kept.