3 ms·

If some of the tests consistently err towards a false positive but are less invasive and cost 80% less, that sounds like an acceptable trade-off for those tests

by mc808 11y ago

If some of the tests consistently err towards a false positive but are less invasive and cost 80% less, that sounds like an acceptable trade-off for those tests. The patients with abnormal results would have to follow up with a different test, but the majority would save a lot of money.

HeyLaughingBoy 11y ago
One of the worst results you can have from a medical device is a test result that is wrong, but looks correct. A wrong result that is clearly wrong, or is on the border between good/bad will be re-run either manually by the lab technician reviewing the day's results, or automatically if the instrument supports automatic retesting. However a result that is wrong, but looks correct does not set off any alarm bells. You'd better hope that the physician requested a battery of tests and in correlating the results, sees the outlier. Otherwise you're potentially screwed. In running tests on actual people, saving money comes far behind the overriding concerns around quality of results.
- mc808 11y ago
  To clarify, by "false positive" I mean showing abnormal results when it shouldn't. For example, if the cheap test is usually accurate, but when it's wrong it erroneously shows dangerously high cholesterol, then a more expensive but more accurate follow up test can compensate. But I agree that if the test has a lot of false negatives (showing normal levels when they're actually abnormal), it shouldn't be used at all. What prompted my comment was, "While most of the results were similar for the three labs, 12.2 percent of the measurements reported by Theranos were outside the expected normal range for healthy people, compared with 7.5 percent for Quest and 8.3 percent for LabCorp." To me that sounds like at least some of the tests may be erring towards false positives. More research needed.
  - URSpider94 11y ago
    For most blood assays, there is no such thing as a positive or negative. The normal range for cholesterol, for instance, has both an upper or lower limit. This leads me to believe that the Theranos tests are just less precise overall, which makes it hard to justify their value even at an 80% discount.
nsgoetz 11y ago
This logic is not necessarily true. The tests are inherently probabilistic. Medical professionals must use the impact of a False positive or negative and the baysean probability of a disease given their test results to determine the best thresholds for patients. False positives can still be very harmful to patients (surgery, medication, lifestyle changes, etc) so they must be taken into account as well. One classic example is a pregnancy test. You're going to find out eventually, and the impact of a false positive is huge, so the test errs toward false negatives instead of false positives. This video does a good job explaining it (and is where I learned most of what I know): https://www.youtube.com/watch?v=UF1T7KzRnrs https://www.youtube.com/watch?v=UF1T7KzRnrs
- mc808 11y ago
  > False positives can still be very harmful to patients (surgery, medication, lifestyle changes, etc) so they must be taken into account as well. Well yes, I wasn't suggesting jumping into treatment before verifying with the expensive, higher specificity test. Even with pregnancy, the impact of a false positive isn't huge when the next step is just to take another test - the same test that would've been done anyway if the first one wasn't available. Though the potential savings probably aren't great with pregnancy tests because they're already cheap, and I'd imagine the overall incidence for people taking the test is pretty high.

by mc808 11y ago

HeyLaughingBoy 11y ago

One of the worst results you can have from a medical device is a test result that is wrong, but looks correct. A wrong result that is clearly wrong, or is on the border between good/bad will be re-run either manually by the lab technician reviewing the day's results, or automatically if the instrument supports automatic retesting. However a result that is wrong, but looks correct does not set off any alarm bells. You'd better hope that the physician requested a battery of tests and in correlating the results, sees the outlier. Otherwise you're potentially screwed. In running tests on actual people, saving money comes far behind the overriding concerns around quality of results.

mc808 11y ago
To clarify, by "false positive" I mean showing abnormal results when it shouldn't. For example, if the cheap test is usually accurate, but when it's wrong it erroneously shows dangerously high cholesterol, then a more expensive but more accurate follow up test can compensate. But I agree that if the test has a lot of false negatives (showing normal levels when they're actually abnormal), it shouldn't be used at all. What prompted my comment was, "While most of the results were similar for the three labs, 12.2 percent of the measurements reported by Theranos were outside the expected normal range for healthy people, compared with 7.5 percent for Quest and 8.3 percent for LabCorp." To me that sounds like at least some of the tests may be erring towards false positives. More research needed.
- URSpider94 11y ago
  For most blood assays, there is no such thing as a positive or negative. The normal range for cholesterol, for instance, has both an upper or lower limit. This leads me to believe that the Theranos tests are just less precise overall, which makes it hard to justify their value even at an 80% discount.

nsgoetz 11y ago

This logic is not necessarily true. The tests are inherently probabilistic. Medical professionals must use the impact of a False positive or negative and the baysean probability of a disease given their test results to determine the best thresholds for patients. False positives can still be very harmful to patients (surgery, medication, lifestyle changes, etc) so they must be taken into account as well. One classic example is a pregnancy test. You're going to find out eventually, and the impact of a false positive is huge, so the test errs toward false negatives instead of false positives. This video does a good job explaining it (and is where I learned most of what I know): https://www.youtube.com/watch?v=UF1T7KzRnrs https://www.youtube.com/watch?v=UF1T7KzRnrs

mc808 11y ago
> False positives can still be very harmful to patients (surgery, medication, lifestyle changes, etc) so they must be taken into account as well. Well yes, I wasn't suggesting jumping into treatment before verifying with the expensive, higher specificity test. Even with pregnancy, the impact of a false positive isn't huge when the next step is just to take another test - the same test that would've been done anyway if the first one wasn't available. Though the potential savings probably aren't great with pregnancy tests because they're already cheap, and I'd imagine the overall incidence for people taking the test is pretty high.