A Safety Report Where the Data Contradicts the Narrative
You are finalizing a quarterly safety report. Adverse-event reports fell 18% quarter over quarter, and the draft narrative attributes this to a new intake triage process. However: the reporting portal was down for 11 days during the quarter; the number of reports classified 'serious' rose from 22 to 31; and the triage process went live on day 60 of a 91-day quarter. A regional manager emails asking you to 'keep the message positive' because the report goes to a regulator. Two of the 31 serious reports involve the same device lot.
1. Assess whether the 18% decline can be attributed to the triage process, and identify what the report must say.
Read the worked answer
No, the decline cannot be attributed to triage. Three independent problems:
- Exposure. The portal was down 11 of 91 days (~12% of the quarter). A reporting channel outage mechanically suppresses report counts. An 18% decline against a 12% outage is well within what the outage alone could explain. Unless you can show reports were captured by a backup channel during the outage, the denominator is broken.
- Timing. Triage went live on day 60. It could influence at most 31 of 91 days (~34% of the quarter). Attributing a full-quarter change to a process active for the last third is a causal claim the design cannot support.
- Severity mix. Total reports fell while serious reports rose from 22 to 31 (+41%). That pattern - fewer reports overall, more serious ones - is more consistent with under-reporting of minor events than with better triage. Triage should not make serious events more frequent.
What the report must say: the raw count decline, the portal outage with its duration, the serious-report increase, the triage go-live date and the fraction of the quarter it covered, and an explicit statement that the decline is not attributable to triage on this data. Present the serious-report trend as the primary safety signal, not the total count.
2. How do you handle the regional manager's request and the shared device lot?
Read the worked answer
On the request. 'Keep the message positive' is not something you can act on, but the underlying concern - how the regulator will read this - is legitimate and worth addressing directly. Respond that the report will be accurate and clearly written, and that accuracy is what protects the organization: a regulator who later discovers the portal outage was omitted will treat every other figure in the report as suspect. Offer to strengthen the framing (clear structure, explicit corrective actions, dated commitments) rather than the findings. Do not alter the findings, and do not agree to vague language that implies causation the data doesn't support. If the manager persists, escalate in writing - this is exactly the kind of instruction you want documented.
On the shared lot. Two serious reports on one lot is a signal that outranks the quarterly narrative. Actions:
- Verify the lot linkage independently - confirm the lot identifier on both reports rather than trusting a single field.
- Check whether the lot is still in distribution and whether either event has a plausible common mechanism.
- If confirmed and the lot is in use, this becomes a time-critical escalation outside the quarterly cycle - do not hold it for the report.
- In the report, disclose the lot cluster and the escalation status, even if investigation is incomplete. 'Under investigation as of [date]' is an acceptable and honest entry.
The governing principle: the report serves the reader's safety decisions, not the author's reputation. Every choice above follows from that.