My patient needed to be delivered. She had just developed eclampsia, a potentially fatal disease that afflicts women in the second half of pregnancy. She had suffered a seizure and dangerously high blood pressure, and was at risk for far worse, including a stroke. No one knows why this condition arises, but delivery sure clears it up in a hurry.

So we gave medication to start labor, and the nurses placed a fetal heart monitor.

Worn like a belt, but higher on the abdomen, the ultrasound monitor would play a crucial role in the hours to come. It prints a read-out strip of the baby's heart rate, and the pattern would guide us in determining whether the delivery would be natural or through cesarean section.

As I suspected, the baby's heart-rate strip showed worrisome changes soon after labor began, and I knew it would get worse as labor progressed. We would fight through the night to have a natural delivery. But ultimately that single heart-rate test, which is surprisingly unreliable, would be a key factor in whether my patient would get a C-section or not.

Nearly all American mothers are monitored during labor, and bad fetal heart strips are an important cause of high cesarean section rates. A recent report detailed the dizzying increases: Almost one in three babies was delivered by cesarean in 2007, the most recent year for which data are available. That rate has grown by more than 50 percent in a decade.

I have performed hundreds of cesarean sections during residency, and many were the result of bad heart-rate strips.

A jagged pattern indicating increases in the heart rate reassures us that the baby's brain is awake and alert, and that labor could continue. But a flat line or decreases in the heart rate after contractions make us think the baby is not getting enough oxygen and pushes us to do a C-section - delivering the baby through incisions in the abdominal wall and the uterus.

For the worst readings, we believed every second counted and rushed the surgery: If the baby wasn't delivered one minute from the first incision into the skin, we had moved too slowly.

The complication we feared most was hypoxia, the baby not getting enough oxygen during labor. Going too long without adequate oxygen could result in a serious permanent injury, such as cerebral palsy, or even death.

No test is perfect. But almost every time we whisked a mother back to the operating room, and I cut through skin, fat, fascia, and finally the muscle of the uterus, expecting a blue, floppy baby, the child I delivered emerged pink, healthy, and a little bit angry.

Were we saving lives and averting disaster? Or were we performing unnecessary surgery?

Fetal heart-rate monitoring is a screening test. Good tests get several things right; they are cheap, detect a possible problem when there is still time to act, and minimize unnecessary follow-up tests.

The Pap smear is an excellent screening test: By examining a few cells brushed from the cervix - where the vagina opens into the uterus - doctors catch precancerous changes - or even early cancer - when it is easy to treat.

But fetal heart monitoring is an appallingly poor test. The test misses the majority of babies with cerebral palsy, the condition researchers hoped it would prevent. It causes increased rates of a painful and invasive surgery: cesarean section. Even worse, almost all women undergo continuous heart monitoring during labor, not just those at highest risk.

The odds of my patient's baby suffering from dangerous lack of oxygen were slim. A study in the New England Journal of Medicine found that only 1 of 500 babies with a bad strip had cerebral palsy. Moreover, it remained unclear if the condition had developed before labor, in which case cesarean couldn't prevent it.

A 2006 analysis by the British Cochrane Collaboration, evaluating all available research, found that fetal heart monitoring failed to reduce perinatal mortality - the risk of a baby's dying late in pregnancy, during birth, or shortly after birth - and increased cesarean section rates and forceps deliveries, compared with listening to a baby's heart rate intermittently.

As a medical student, I loved watching emergency cesarean sections. The baby's heart rate went down, doors swung open, residents rushed the patient down to the OR, and a frantic minute or two of surgery later, a screaming baby was out. The excitement pushed me to choose a career in obstetrics. I never questioned the need for the surgery.

Now, cesarean sections for bad tracings are one of the least satisfying parts of my job.

Steven Clark and Gary Hankins, two prominent obstetricians, voiced my frustration. "A test leading to an unnecessary major abdominal operation in more than 99.5 percent of cases should be regarded by the medical community as absurd at best," they wrote in the American Journal of Obstetrics and Gynecology. "Electronic fetal heart rate monitoring has probably done more harm than good."

Why do doctors cling to continuous fetal heart monitoring? An obstetrician will most likely point to the fear of being sued, but the complete answer is more complex. Our medical culture prizes technology and tests, even if they don't work and can cause harm.

"It's our bias that anything that can be quantified is an improvement," said H. Gilbert Welch, a professor at Dartmouth Medical School whose research focuses on harm caused by screening and over-diagnosis.

"I think we get in trouble when we start promising things to . . . well [patients]," Welch said in an interview. "It is not that hard to make them worse."

For three or four hours that night, I struggled with my patient's bad fetal heart strip. I wanted her to avoid a cesarean section. She had type 1 diabetes, and I expected her sugars to swing wildly after surgery, and her recovery to be slow.

To improve the strip, the nurses and I tried giving her oxygen, changing her position in the bed, even rubbing the baby's head through the cervix to wake it up.

Finally, at 3 a.m., I felt compelled to recommend cesarean. The strip continued to look bad, and my patient's labor progressed slowly.

We went to the operating room, and delivered the baby by cesarean. My patient's child greeted the world pink and well-oxygenated.

The test was wrong again.