Amber, age 26, lay in the emergency room. She couldn't move her arms, her mind repeatedly blanked out for several minutes at a time, and she was extremely dizzy. Her stomach hurt so much that she hadn't had anything to eat or drink for many hours.

The doctor said, "I think you're just dehydrated."

They put an IV in her arm and hung a bag of dextrose solution - sugar water - to drip into her vein. Shortly after the bag of sugar water had emptied, a technician showed up to draw blood. He explained, "The doctor wants to see if you have diabetes."

Amber looked at him in confusion. At the moment, she was able to speak, and she said, "That's going to read high in sugar from the IV bag of dextrose I just got."

The technician replied, "It's OK. I'm drawing blood from your other arm."

Amber's mother was at her bedside. Her jaw dropped, and she protested, "The fluid that goes into that vein travels throughout the body!"

The technician shrugged and finished drawing Amber's blood.

Hours later, the doctor came in and said, "Your blood sugar level is high. I think you're diabetic." He sent her home with a glucose test meter and test strips, and told her to test her blood sugar level regularly.

Amber's primary care doctor later confirmed what Amber already knew: She is not diabetic. The problems that landed her in the emergency room had nothing to do with diabetes. She was sent home from the ER with no treatment for a potentially life-threatening brain infection that went undiagnosed.

What went wrong with her medical test? No one connected the dots. The doctor may have ordered the blood test before the IV line went in. The lab may have been backed up and so didn't get to Amber until after she had had roughly a quart of sugar water infused into her veins. Or perhaps the doctor ordered the blood test after ordering the dextrose drip while he was thinking about something else.

This busy doctor may have read the report out of context without thinking about what else was happening with Amber at the time. He might review dozens or hundreds of medical test results in one shift in the ER. He probably has a routine: quickly scan for the marks that show an abnormality. Prescribe treatment to deal with that problem. Move on to the next test result.

Research published in the journal Quality & Safety in Health Care showed that even in office visits, mistakes in medical tests are not unusual. An article in the New York Times called "Testing Mistakes at the Family Doctor," which summarizes that study, reported that the doctor sometimes orders the wrong test, the lab sometimes performs the test incorrectly, and the lab might even do the wrong test (even if the correct one was ordered). Thus, the results of the tests can be meaningless.

Amber's experience shows that even if the correct test is done, the doctor can easily draw the wrong conclusion and proceed down a path that doesn't help improve the patient's health.

One problem is that test results are often less reliable than you - and your doctor - might think they are. According to Michael Blastland and Andrew Dilnot in their book The Numbers Game, only about 8 percent of doctors in one study did the math right when asked to calculate how likely it was that a person whose test results indicated that they had a disease actually had it. The rest, they say, were "hopelessly confused."

Typically, perhaps only about 1 percent of the people given a screening test for a condition like diabetes actually have the disease. That's one person out of 100. The other 99 people are fine.

Assume that the test is 90 percent accurate. These numbers mean that 10 percent of the 99 people who are fine will be told that they are sick. That's 9.9 people. (I'll round that number to 10 people.) As a result, the one person who is sick and 10 who aren't will all be told that they have the disease - when only one person out of those 11 actually does.

Most doctors in this case assume that nine out of 10 people whose tests say that they have the disease actually have it.

What can you do to help ensure that your medical tests provide useful information?

  1. Ask for - and write down - the name of the test and what it is for (or have a friend or family member do it for you).  
  2. Ask what might skew the results so that they don't paint an accurate picture. For example, eating a high-fat food - say, cheesecake - shortly before having a cholesterol test will distort the results.  
  3. Follow the pre-test instructions carefully.  
  4. Check that all of the identifying information on the paperwork reporting the results appears as expected. For example, are your name and the name of the test correct?  
  5. If a test shows that you have a problem, don't simply take this answer at face value. Ask how this conclusion can be confirmed.