Not really. Normal labs that aren't aiming to be revolutionary ultra-profitable mega corporations will be fine.

Having a slightly more accurate result is useless, because there is far more biological variation in a person (from disease, the vagaries of physiology, compliance, pharmacokinetics, diet, the time of day, mental state, the season etc etc) than technical variation in a lab test. You would be optimising the wrong thing.

The only example I can think of where someone would probably be harmed by this is if they were undergoing some kind of a procedure with a bleeding risk, and the Theranos test said their clotting was fine when it wasn't, and they had a significant haemorrhage during the procedure. Someone will probably launch a class action lawsuit anyway.

I think that the Theranos board will realise that they need to get rid of Holmes in the next 6 - 12 months. Even if the tech actually works, having her there has become a liability.

Not practical for most patients in the real world. Insurers won't pay for redundant lab tests. Getting blood drawn multiple times is painful and logistically difficult. For patients with clotting problems just getting blood drawn presents some risk. And there often won't be 3 independent labs in the patient's area.

You could reduce your accuracy by doing three tests of the same blood sample (of course, exposing yourself to the risk that the blood was taken incorrectly, was stored in an incorrect vessel, was transported incorrectly, etc...).

In terms of practicality/Insurance - this would mostly be a function of how absolutely certain we are that the tests are accurate, and how damaging incorrect results would be. If we are 99.99% certain the values are correct, and incorrectly results wouldn't result in irreparable harm - then yes, not worth the investment in redundant lab tests.

But, if there is, say, a 5% chance I might die on an incorrect test result that has a 1% chance of happening - and that test is performed 12 times/year - then you can pretty quickly calculate the net-present-value of my life versus cost of test. (I.E. If I think my Life is worth $10mm, and the test normally costs $500 each time, or $6000/year, and I need to have this test done for the next 10 years, (assuming the value of my life doesn't drop over time, to make this simpler) - then each time I get the test there is a 1% chance of failure * 5% chance of death = 0.05% chance of dying as a result of each test.

The odds of me dying then, in the next 120 test = 100%-(chance-of-not-dying-120-times-in-a-row) = 1-.9995 ^ 120) = 5.825% chance of dying in this scenario using tests that cost $60,000. If I could reduce that to 1% (4.825% reduction) chance by spending an extra $120,000 (perform the test in two extra labs) - then my my investment of $120k would return 4.825%*10mm = $482,500.

You can work the numbers, assumptions, etc.. however appropriate - but for some combination of test costs, frequency, potential harm, and value assigned to life - there is quantifiable value that justifies performing the test multiple times (and, likewise, clearly circumstances in which, even if the test results are more likely to be inaccurate, not worth performing thee test).

Not having access to multiple labs is a real problem though.

That's not how healthcare economics works in the real world. No one can quantify the chance of death due to an incorrect test result precisely enough to calculate when it's cost effective to perform multiple redundant tests. The supporting data doesn't exist and no one is going to fund the research to figure it out. And what does 'incorrect' even mean? There is always some margin of error.

In real medical practice if a physician sees a lab result that is way out of range, or significantly different from the patient's previous result, or inconsistent with other aspects of the patient's condition, then she will order a repeat of the same test in order to confirm the results.

It's certainly how healthcare economics works. It determines what, if any tests should be run. It also determines how many parties/independent reviews should take place for high stakes tests such as tumor identification in carcinoma.

Other high-stakes scenarios in which the same tests are run multiple times are for high-economic cost events such as drug abuse, either in the workplace, or in sports.

You even note in your response, that, In real medical practice if a physician sees a lab result that is way out of range, or significantly different from the patient's previous result, or inconsistent with other aspects of the patient's condition, then she will order a repeat of the same test in order to confirm the results. - which is exactly what I'm suggesting people have done - only with a (presumably acceptable) time delay between when the first, and later tests are required. If the results of such tests were that clinical activity had to take place quickly, then you might not have the luxury of a delay, and in which case, multiple independent test vectors would be very useful - if they were economically feasible.

What people sometimes fail to consider - if the test was worth doing once - at, say, $600 - then why wouldn't it be worth doing (at presumably much higher confidence), three times at $100? And, if people agree, "OK, for $600 and $100 that makes sense" - they have to understand, those two numbers are completely arbitrary, and can only be understood in relationship to the value of the results. Whose to say there aren't a ton of tests today that are "$100" - that couldn't be done multiple times to be certain.

In fact, my friend was a lab technician in a private outfit. He quit when the boss said "no more running control tests before hospital tests. Its too expensive" So that meant they didn't even calibrate equipment/test for miscalibration any more! And the hospital is paying them for quality; they're delivering risk.

Not necessary. I worked in the medical device field for many years, writing software to control instruments that tested blood (serum).

The instruments can be trusted.

If a test is borderline, either the lab tech will manually retest it, or some instruments can be given business rules like "if result is on the threshold of good/bad, automatically rerun the test and provide both results to the physician."

We were well aware that the software we wrote affected real people's lives and acted accordingly. These idiots are only giving the rest of us a bad name.

If none of the three tests is calibrated and traceable to a standard you are just as well off throwing darts at a board and getting your numbers out of that. Take personal responsibility for your health and your bank account by throwing dice!

It's really funny how people have stopped hating on the FDA when Theranos is mentioned. In any other context the FDA is still worse than Satan and the pox.

If you're taking Warfarin, you have to get your INR checked every two to three weeks. It's really not practical to do that at three different labs every time. Also, this isn't some exotic edge case that only applies to a few people, but an extremely common medication. We're talking about huge numbers of patients here. If all their tests had to be triple-checked, the costs would be enormous.

See https://news.ycombinator.com/item?id=11243904 above as to why the cost issue is a risk assessment / return analysis. In certain scenarios in which potential harm is high from an incorrect diagnosis, it's worth the cost to perform the test completely independently.