Adverse drug reactions and postmortems: the same document

Software thinks it invented the blameless postmortem. Pharmacovigilance has been running that program for decades: adverse drug reaction reporting is a global system built on exactly one principle — if reporting punishes the reporter, the reports stop, and the next patient is hurt by a lesson nobody learned.

The structural parallels

The report is about the system, not the person. ADR reports ask: what drug, what dose, what patient context, what timeline — the conditions under which harm occurred. Nobody writes "Nurse Joy was careless." They write what made carelessness (or mere normalcy) dangerous. Modern SRE postmortems say the same thing: describe the system that allowed the failure.

Signals emerge from aggregates. One reaction report is noise; a pattern across hundreds is a signal that pulls a drug from the market. Incident reports work identically — one is an anecdote, a pattern is an architecture problem.

Underreporting is the silent killer. The known failure of pharmacovigilance: most reactions go unreported, so signals arrive late. Software has the same disease — most incidents go unlogged, so organisations re-learn the same lessons annually.

What I carry into engineering

I write incident notes as if a regulator will read them — conditions, timeline, contributing factors, no blame. And I fight underreporting where I can: the cheaper and safer it is to admit a failure, the faster the system learns. Both fields run on the same truth: a culture that shoots messengers develops message scarcity, and message scarcity is how preventable harm compounds.