Sorry Redbar, lots of people agree with your Analysis, but it is wrongish as far as the maths goes.
Otherwise, ReasApps opportunity for even further improvement is there, the higher numbers of patients strengthens ResApps scientific case. The 93.7 from DROWZEL speaking as a scientist looks too precise (The extra 0.7 is just ridiculous, given the range of this measurement would have large), and should have an expected error range (or 95% CI range) as ResApp publishes. Very professional output from ResApp.
Again happy to be proved wrong.
To explain the maths (I am a statistician/physicist 20years + experience), everyone will switch off now.Both algorithms from each company will produce a number in the end for each patient, all models produce a single number.
If the number for a patient is above a threshold ->
positive detection.
If the number for a patient is below a threshold they are given a pass ->
nothing detected.
The
choice of threshold determines the trade-off between between false positives (ie. you tell the patient they have sleep apnoea event when don't.)
and true detection. If the threshold is low, you getting great Sensitivity (detect everything) but poor Specificity (ie. lots of false positives, people without sleep apnoea detected). Likewise if you choose a threshold that is too high, you get great Specificity but poor Sensitivity. The choice of threshold is entirely up to the manufacture and can be changed at will. There are an infinite number of Sensitivities and Specificity pairs of numbers.
But these pairs of numbers, follow what is called the ROC result for the test, below is a typical
ROC curve. ROCs do depend on the model, so the ROC curves maybe very different for DROWZLE and ResApp, we dont know neither is published. But with this
typical ROC curve, very middle of the road ROC that is showing a good result for method, I have placed both the Drowzle and ResApp result. In this case, a statistician would say both models
are performing similarly, the two companies have
only chosen different thresholds.
Sorry for getting geeky, but I would like to judge the company on truth not misunderstandings.
Strong believer in the future prospects of this company, have my cash in as a shareholder (Significant to me), hoping Sleep Apnoea will be the biggest product in the medium term. Interesting to know what ResMed is thinking.
