Math problem with results

Hi guys, thought I'd start a new thread for this.
Sorry for the absence RAP veterans; been mulling this over since last week.
Welcome aboard new holders - positive or negative; discussion is welcome because we need it now as we needed it before US results came out.

A former RAP holder and HC poster called me over the weekend to alert me to some discrepancies in the results presentation by RAP management in regards to the Aussie trials. I will now attempt reproduce/present the flawed math behind their calculations and let's have a chat about it?

For the purposes of calculation consistency - I'm just looking at Viral Pneumonia, Cough Alone.

We had an update from RAP (31/4/16) re: clinical results for Paediatric patients. They had 43 patients diagnosed with RAP software at an accuracy of 91% for viral pneumonia, cough alone. Excellent. (this means 39.13 patients came back with positive diagnosis so we'll say 39 patients cos 0.13 patients is goofy math)

The final results tally (22/6/17) states that only 70% agreement was achieved for pneumonia, cough alone with a sample size of 123 patients. This means that 86.1 (let's say 86) patients were correctly diagnosed for the disease.

This is where I've got an issue... RAP's investor presso (forgot the date but it's definitely there on page 7) states that the later stages of the trials were "optimised for US protocols". If we listen to the conference call last week - TK states that US protocols means that the research staff can not be near the patients during diagnosis - this then means that the FIRST 43 patients had RAP staff present fiddling around whilst the latter 80 patients (123 minus initial 43) had no RAP interference. Now, if we take away 39 patients from the 86 patients (47) and put THAT number over the 80 patients to get a percentage - we get 47/80 = 58.75 or 59%. This means that WITHOUT RAP staff tweaking the diagnostic process - the algo was only able to achieve an accuracy of 59% here in Aus on a sample size of 80 patients.

RAP management elected to COMBINE the first 43 patients (WITH RAP influence) with the subsequent 80 patients (US PROTOCOLS - ie: no RAP influence) to get an overall result of 70% but this is not what's done in the scientific community. One does NOT remove/add/alter a variable in the middle of study/trial and then collate results from before and after. This contaminates the data significantly and is not an accurate representation of the facts. RAP involvement is a clear variable that impacted results and they can see this when they are putting the data together.

with RAP involvement - 91%
without RAP involvement - 59% (ie: US protocols)

this didn't raise any flags?

How can management expect a 75% standard of accuracy from US trials when they can't even get above 60% here in Australia? They should not have included the first 43 patient data because it just doesn't count. TK is engineering trained - he would have been taught to control his variables from high school science - he knows as well as the team that was putting out the announcements that this 70% is a false representation of RAP's abilities.

All this talk to surprise and shock expressed by management at the low US results doesn't ring true to me because if I were analysing the raw data (background in bio-sciences) - I'd see clearly that I can't hope to get above 60% on the US trials. I'm not saying that all the reasons for the bad results (extra coughs/tv blaring...etc) are not valid reasons for the software faltering but even under idea conditions here in Oz, they couldn't pull off 75% so how can they expect any higher than 60% from the states?

I'll leave it there for now because this post is already too long and I'm sorry it's been so verbose but as someone who's spent time in the lab; this is shifty as hell. I've held off for a few days before I wrote this just to think about things and let last week's impact set in so that I'm not reacting emotionally to the bad news but this issue is just simple math and it just doesn't add up.

so what happened?

1 - management didn't understand one of the first tenets of scientific experimentation and mixed results together (hard to believe considering that medical/scientific professionals were involved in this trial)
2 - management was very aware that "sans RAP" results are very weak and so decided to combine the two samples sizes in order to boost their Aussie trial result to 70%

let me know what you guys think.