Predictive Informatics Monitoring in Neonatal Intensive Care Units: Saving Premature Infants by Monitoring their Heartbeats and Respiration

Premature infants in Neonatal ICUs often deteriorate suddenly due to potentially catastrophic illnesses.  Bedside monitors pick up the electrocardiogram and other signals, and these might give clues about impending illnesses, but the signals usually fly by unnoticed.  To solve this problem of undetected impending changes, predictive informatics monitoring collects and integrates these data to give clinicians early warning of events, allowing earlier, more effective intervention. 

Inflammatory illness leads to changes in autonomic control of the cardiorespiratory system, and our group has previously shown that by monitoring changes in the rhythm of the heart, one can give early warning of bacterial infections.  This early warning system has reduced the death rate by such infections by 20%.

 Presently we are using similar methods to study respiration and oxygenation.  We have collected data representing all electronic signals from ~40 beds in a NICU for five years – a total of ~12 Terabytes of data (Pretty-Big-Data).  Among the present projects are to use this large dataset to:

1.  Generate statistics on cardiorespiratory characteristics of infants, including average heart rate, average heart rate variability, respiration rate and its variability, and average apnea burden, and characteristics of those apneas.

2.  Collect and study data on periodic apneas (also called periodic breathing).  Infants often go into a cycle of 7 seconds of breathing and 7 seconds of apnea.  This phenomenon has long been thought to be benign, but we have found extreme examples of periodic apneas that preceded sudden infant death.

3.  Collect and study extreme apneas.  We have found many examples in which a baby stopped breathing for 60 seconds or more.

4.  Develop physiological measures of the consequences of apneas. 

5.  Seek better warning signs of impending sepsis, necrotizing entercolitis, or intracranial hemorrhage.



In very long apneas, oxygen saturation and heart rate fall slowly.  We show the average heart rate and oxygen saturation for events of various durations.  The start of each apnea is t=0.

Adviser: John Delos