Ann: Voluntary Suspension, page-112

I agree with the facts here, and the comment by @pilko3 "Why and how on Earth wouldn't the USA trials deliver the same results?"

Yet there are differences between the US and Aus trials, with the key difference being the adjudication.  Bear in mind that ResappDx has been trained and tested against Australian adjudication standards.  Last time we identified significant differences between the different adjudication teams, so we centralised adjudication into a single team.  The clear implication is that some adjudication teams produced a greater percentage of results inconsistent with ResappDx.

So if the centralised adjudication team has used criteria and methods that produce this higher degree of inconsistency, then that will reduce both positive and negative percentage agreement.  This seems like a credible possibility as our algorithm has been trained using Australian adjudications.

Getting the bottom end of the 95% confidence interval above 75% is a tough call.  Anything that reduces our agreement rate will have a negative impact on our score.