Stanford researchers have developed a new system to predict the health problems premature infants may face as they grow, one that is both more accurate and less invasive than currently used alternate methods.

Researchers have been working on the system, called PhysiScore, for the last two years and published an article detailing its development on Sept. 8 in “Science Translational Medicine.” According to Anand Rajani, a neonatal specialist and one of the main researchers and co-authors of the article, the idea for the PhysiScore came when Suchi Saria, a graduate student in computer science, and with computer science professor Daphne Koller and pediatrics professor Anna Penn, noticed specific patterns in premature infants who faced health complications that were not apparent in healthy premature infants.

“PhysiScore works by looking at three sets of signals that come from a monitor that is attached to every premature baby after they are born,” Rajani said.

Using an EKG machine, physicians continuously record these three signals: heart rate, respiratory rate and oxygen saturations. By running this data through a computer program, researchers found they could identify patterns unique to babies with health problems. They then developed a scoring system using a scale of one to 10 to determine the likelihood an infant will face health complications.

According to Koller, short-term variation in heart rate and respiratory rate proved to be the most effective determinants in predicting health complications. She explained that an infant’s level of responsiveness is a telling sign of future health, saying that “poor responders have a much flatter profile” when presented with stimuli such as the touch of a parent.

Prior to PhysiScore, physicians primarily determined the likelihood of developing illness by measuring the baby’s weight and relying on his or her gestational age. However, this method was based on large groups of infants and did not take into account specific circumstances.

“We knew at that time that machine learning was a powerful method and that being able to make a better, more individualized assessment of a given premature baby’s risk of being sick or well was needed, so we continued to develop PhysiScore,” Rajani said.

According to Koller, PhysiScore is the first method that is able to detect complications not present during the first three hours of life. PhysiScore is able to detect fragility and susceptibility to disease that other methods, including the Apgar method, are unable to determine. Computed five minutes after birth, the Apgar considers a combination of different factors, including skin color, pulse rate, reflexes, muscle tone and breathing. All other methods of determining premature infant health use data from the first 12 hours of life and require invasive testing.

“One of the unique characteristics of PhysiScore is its use of noninvasive measurements,” she said.

Rajani explained that since PhysiScore processes “continuous streams of data from electronic monitors over the first three hours of life…this gives [physicians] a far greater level of detail as to how the baby’s body is doing and thus gives [physicians] a much higher level of accuracy in making a determination about the baby’s future risk.”

PhysiScore has many potential uses in the health care field. An infant’s PhysiScore could help determine if he or she should be transferred to a higher level of care. In addition, the PhysiScore may help administrators determine nursing-to-patient ratio by predicting which infants need more care.

According to Rajani, the PhysiScore can also help compare the quality of neonatal intensive care units by “normalizing their outcomes based on some measure of patient acuity…given the accuracy of PhysiScore, it would be particularly useful for this purpose.”