Thank you for sending me the link to this video.
I am an individual whom has an interest in my own health and organised the base blood test and the blood calculator through you.
I am pleased with your service and accept that it cannot be totally accurate and I am using it as a tool to point me towards the areas I should be looking at.
In terms of feedback on how things may be improved in the future:
1. The site could explain / point out the concepts of specificity and sensitivity
2. The predictions are very binary i.e you think that something is out of range but not the degree to which something is out of range - this may be a big ask.
3. I think it is potentially misleading to quote a probability that is greater than the specificity %.
In the video you mentioned that the model is being updated. Will the predictions for tests that have been run be automatically updated and if so how will users know their results have changed.
Thanks very much for your feedback. We graciously accept all feedback and everything is read and carefully considered. Your suggestions have been shared with the team.
When we update the models, we will send out an email letting you know that there has been an update and what predictions might have been affected on your report. The next time you access your report link, it will update automatically.
I hope this helps. Please let us know if you have any other questions!
That was indeed useful. I did have some confirmed data that was both low when predicted to be so and high when predicted as low. The logic is is helpful (looking into methyl folate now and I am on 500% folic acid). Awaiting my 23andmeresults which would also explain MTHFR...if applicable to me.
Question: Certain race and subpopulations are known to have distinct traits. My doc insists, for example, that people of African origin (Afro-Caribbean etc.) like me, have higher CO2 and lower iron levels. (Some posit there are actually valid physiological reasons for these findings). Have you considered this? Considered it as a modifier/tracker in the algorithm?
Yes, we know that this is incredibly important for a number of markers, including how certain markers are predictive of disease risk in some races much more than in others. At the moment, we are in the process of adding in notifiers for people to give references on how certain markers mean different things in different races. In the future, these will be built into the optimal ranges and score calculators. As for the predictions, the data we have includes a pretty representative sample of the US population (including race diversity), so any normal variation in markers will be taken into account. However, we may also add race as a modifier in the future if needed.
Since meeting Chris and Tommy over here in the UK last month at the BASEM conference I've become increasingly excited about the blood chemistry calculator (perhaps my engineering background as something to do with this!) - I reckon I eyeball between 20 - 30 sets of these standard markers 5 days per week, 45 weeks of the year for the past 13 years = approximately 88,000 sets of these 38; and I can pick up a few patterns but I'm still pretty rubbish. So I am very excited about using this in my private coaching work. Interestingly I suspect that the times when the calculator delivers a false positive - as you say in the video; "this person looks as if they have x but they don't" - is going to be super useful because you then have to go away and explain that result and find a root cause. I suspect these events are going to turn out to be some of the most clinically useful ones and most challenging/interesting ones to solve! Thanks guys for all your hard work!! Michael
This video is really helpful to explain how the algorithm is working. A video like this would be very helpful to be included when you are sent the link to your report as watching this has made things a lot clearer. I had a few things with probability of 100% can that be possible if the test does not have 100% sensitivity and specificity?
Good question. You do need to keep the sensitivity and specificity factors in mind. If the calculator forecasts something with 100% probability and then you go to check the sensitivity/specificity you'll find that there is, in most cases, some chance of false positive (specificity). So a specificity of, say, 96 indicates there's a 96% chance that the 100% is accurate, and a 4% chance it is not accurate.