Ann: Becoming a substantial holder, page-14

I've said this several times before... the Australian studies weren't trials at all. They were studies designed to:
a) Gather symptoms and cough data for a bunch of patients, and the condition each patient was diagnosed with.
b) Train a classifier algorithm with the data and diagnoses.
c) Test the trained classifier (using leave-one-out) to measure the effectiveness of the training.

You can't use the training data to prove the training was effective. That's why we're initiating the new double-blind Australian trials.

I have a colleague at work who loves asking questions like, "what do we need to allow for that we haven't thought of?"

It seems like a stupid question. How can we think of what we haven't thought of?

But we can't possibly be the first company to ever try to collect sounds in a hospital or medical centre. Who else has attempted something like this? What snags did they hit? What did they do? Was there anything they luckily allowed for because it happened?

A quick search of ClinicalTrials.gov using cough sounds as the keywords turns up our study plus 12 others. Some can be ruled out as completely different: ultrasound images, neurological responses to sounds. But I can see 6 studies that can't be ruled out instantly. There might be more studies in archives or other medical resources. Perhaps some of those studies included sound recordings taken in hospitals. Perhaps some of those other organisations might be prepared to discuss issues they experienced or were able to predict with their trials.

If we could learn from other peoples' experiences it would be a whole lot better and cheaper than learning from our own mistakes!


PS: I hate working with the guy who asks about what we haven't thought of. He frustrates the heck out of me!