Posted Sept 22, 2010
San Francisco — Stanford University researchers have developed a new method for predicting health problems in premature babies that they say is much more reliable than the standard Apgar score used to evaluate newborns’ health.
The new method, which the researchers have named the PhysiScore , rates infants’ health based on vital signs taken at birth and over the first three hours of life. In a study of 138 infants at Packard Children’s Hospital , the score was able to predict a baby’s chances of developing a serious illness with an accuracy of 91 percent to 98 percent.
The Apgar score has been favored for more than 50 years because it’s simple and fast — in the seconds after birth, doctors consider five health factors, such as how well an infant cries and how much he or she squirms about, and they assign a score.
But the Apgar test is more subjective. The PhysiScore relies on hard data, including a newborn’s weight, heart rate and respiratory weight. At the same time, the PhysiScore is fairly easy to determine because it takes data that’s already being collected and uses a computer algorithm to spit out a health score.
“At the three-hour point, out pops a prediction number,” said Dr. Anna Penn , a neonatologist at Packard Children’s Hospital and a co-author of the study, which was published this month the journal Science Translational Medicine . “What this will likely be is one piece of information used among several others. It’s something that can be valuable for decision-making.”
The Apgar score is useful to doctors for determining the broad treatment course of a baby born prematurely, or with other health problems. Babies born with a low Apgar score will likely be taken to a neonatal intensive care unit, or may even be transferred to a hospital able to provide more specialized care.
But in the Stanford study, the new method was far more accurate.
For example, the Apgar score could predict whether a baby would develop a dangerous infection with 74 percent accuracy; the PhysiScore could make the same prediction with 97 percent accuracy.
Part of the reason for the increased accuracy may be the new scoring system’s use of a computer algorithm to assess raw electronic data.
The PhysiScore uses vital signs like a newborn’s heart rate and oxygen saturation, as well as how those vital signs change over time. These are all numbers that have long been available to doctors, but it’s been impossible for medical staff to constantly process all of the information and evaluate every infant based on the data collected.
In other words, for years hospitals have been collecting useful medical information on infants, but not taking full advantage of it, said Daphne Koller , a computer science professor in the Stanford University School of Engineering, who helped design the new scoring system.
“Doctors look at the data on a monitor, and they look for certain cues like the respiratory is low, but largely that data is thrown away,” Koller said. “What we were struck by was the richness of information that’s already available and how much you can extract from it.”
In fact, researchers already are thinking about how similar scoring methods could be applied to other patients — people who have had heart attacks, for example, Penn said. With more hospital keeping electronic medical records, there is a large amount of data automatically collected on patients, and much of it isn’t being put to use, she said.
It remains to be seen whether the PhysiScore will be reliable when applied to larger groups of infants. The babies in the Stanford study were born premature — six weeks or more short of full-term, 40-week pregnancy — and weighed less than 4 pounds, 6.5 ounces.
The new method may not be as accurate when applied to babies born closer to 40 weeks of gestation. Or, the researchers said, there may be something about the babies cared for at Packard Children’s Hospital that sets them apart from other infants and would impact their PhysiScore. Researchers said they plan to conduct further studies of larger groups of infants in the coming months.
Dr. Chris Retajczyk , a neonatologist at California Pacific Medical Center, said he will be paying attention to further studies of the new score. He said he’s done research into other scoring systems to improve on the Apgar and found them often more reliable.
But they’re also overly complicated, he said. The appeal of Apgar is that it’s simple and quick — a score that shows up automatically on a medical monitor would be even better, he said.
Retajczyk said he’ll be especially interested in the PhysiScore if it’s able to accurately predict a wide variety of long-term health outcomes.
“In general, we know how to treat these kids and maximize our resources. The scores may make our ears perk up a bit, but it’s not going to be a major change to care,” Retajczyk said. “But if we can show a much more predictive value, like if there’s a 90 percent likelihood of (gastro-intestinal) problems, that would be really helpful.”
SCORING METHODS
The Apgar score is given to all newborns immediately after birth, and again five minutes later. The score is given on a scale of 1 to 10, with higher scores correlating to better health. Five factors are included in the score:
Activity and muscle tone: whether the baby is active and squirming.
Pulse: the baby’s heart rate.
Grimace response: how the baby reacts when stimulated.
Appearance: the baby’s skin color.
Respiration: the baby’s breathing and how well he or she cries.
The newly developed PhysiScore is a complex computer algorithm that collects vital signs from infants during the first three hours of life. The score is on a scale of 0 to 1, with higher scores indicating a greater likelihood of serious health problems. Ten factors are included in the equation:
Gestational age.
Birth weight.
Mean heart rate, heart rate at birth, and change in heart rate over time.
Mean respiratory rate, rate at birth, and change in rate over time.
Mean oxygen saturation and the total amount of time at low oxygen levels.