Tuesday, July 2, 2013

A false positive nightmare

Once upon a time, I took an International Public Health class.  As part of this class, the professor was teaching us about false positives and false negatives (false positives = test results that say you have something when you do not, false negatives = test results that say you don't have something when you do), and he asked which one we'd prefer in an initial screening test for disease.  Most of the class said false positives...better to initially believe you have something and be told later you don't, right?  He agreed that we were likely right...in the US at least.  However, he explained, in other countries this may not be the case.  In some areas, even an initial suggestion that you had something like HIV could lead to some major fallout...spouse leaving, getting let go from your job, etc...that may not be easy to correct even once the final results were in. The problem is not always what a patient will do with information, but rather what others might do with the information.

I thought of this today when I read this story about a new mom in Pennsylvania who got her 3 day old baby taken away because she had eaten a bagel before going in to labor.  Yeah, you read that right.  The bagel happened to contain poppy seeds, and it turns out this caused her to test positive for opiates, which caused the hospital to report her, which caused her to have her daughter ripped out of her hands right after she got home.  Now, this story didn't make a tremendous amount of sense to me, so I read through the whole lawsuit (the hospital settled).  A few details that fill in some of the blanks:

  • This hospital has mandatory drug testing for all moms in labor.  This is actually not standard practice...my hospital for example only did this if there was cause.  No behavior on the part of the mother triggered this.
  • The cutoff used for the initial screening test is low...100 nanograms/uL.  In contrast the cutoff for say, Olympic athletes is 1000 nanograms/uL.  For federal drug testing, it's 2000 nanograms/uL for codeine, and 4000 for morphine.  The mother's levels were 300 nanograms/uL on the initial test, and 500 on the confirmatory test.  
  • The doctor who saw the mom and baby thought they were fine, so didn't even tell them about the test results.  She assumed they were a false negative.
  • The hospital reported these positive tests to state, whose policy states that two positive drug tests are all that's needed to take the child away.  They did no other investigation prior to removing the child.
Now based on the fact that the hospital and state social services have both paid money and changed their policy, I'm going to assume most of what's said above is true.  Given that, this is a scary real world case of people not understanding the ramifications of a false positive.  

Now truly, in the real world, is it better that a (known to be healthy) 3 day old baby spend two extra days in the care of a mother who uses opiates while an investigation can be done, or is it better that new parents have their baby taken away for several days for no reason?  The answer depends heavily on how often are they happening relative to one another.  Is it worth it if they happen at equal rates?  More false positives than false negatives?  More false negatives than false positives?

This is why it's so critical that people in many professions understand statistics.  As part of the lawsuit, it was explicitly mentioned that the training of the case worker failed to properly advise them that this could happen and to conduct themselves accordingly.  The judge who granted the ex parte petition also seemed to not know/not care about the false positive issue. 

Obviously we'd love to get the right information all the time, but the false positive/false negative debate is really about choosing which type of bad information you'd rather get.  This is a difficult choice, but the way to mitigate that is to remember that numbers are harder to misinterpret when you take them in the whole context, rather than just as stand alone facts.  In this case, the numbers are neutral...it's the standards set around them that cause the problems.


  1. Parent's nightmare, lawyers dream.

  2. It may also fuel paranoid subcultures. I say "may" because those may persist even without real examples to burn. But these don't help, certainly.