Contrary Adverse Reporting

Dear all,

I’d be keen to seek some thoughts on the up-to-date prevalence of contrary adverse reporting (this is where the report incorrectly omits a NOT or NO, or something small which changes the fundamental opinion issued like IS/ISN’T, not through the fault of a reporter, but a transcription error or auto-correct change).

Around the early 2010’s there was a regional study carried out which, from memory, showed roughly 2% of abnormal radiology reports had a contrary adverse report issued at first (but were later addended, amended or even spotted before final validation). I can no longer find this study, and obviously VR systems have changed from the standard SpeechMagic-for-all we pretty much had back then. One of the suggestions I made as feedback to the study was a standard tick-box or field in RIS (then passed to external systems) which clearly indicated ‘Findings Present’ (this hasn’t been picked up by any vendors in the years since seemingly!)

Do reporters still note ‘clipping’ of the ‘NO abnormality’ etc. with the newer generation of VR systems?

Amazingly though, this seems to be an area where AI isn’t being widely deployed to assist and point out possible contrary reports before issue!

Thanks,
Al.

Our current supplier uses Augnito. Whilst the report is preliminary all of the negation terms/semantic modifiers are highlighted to be checked before the report is verified.

1 Like

We use Nuance integrated, and yes, this is still an issue. We have noticed that it happens with some people more than others: reports marked as “non‑acute … ” sometimes comes as “acute…".

1 Like