Toward the early diagnosis of neonatal sepsis and sepsis-like illness using novel heart rate analysis

Background and Objective. Abrupt clinical deterioration because of sepsis is a major cause of morbidity and mortality in neonates, and earlier diagnosis should improve therapy of this potentially catastrophic illness. In practice, clinical signs and laboratory data have not been perceived as sensitive or specific for early stages of sepsis. Because heart rate characteristics (HRC) are abnormal during fetal distress and neonatal illness, we hypothesized that abnormal HRC might precede the clinical diagnosis of neonatal sepsis, adding independent information to standard clinical parameters. Methods. In the neonatal intensive care unit at the University of Virginia, we prospectively studied infants admitted from August 1995 to April 1999 who were at risk for developing sepsis. Infants in the sepsis (culture-positive) and sepsis-like illness (culture-negative) groups had an abrupt clinical deterioration that raised clinical suspicion of infection and prompted physicians to obtain blood cultures and start antibiotic therapy. Infants without sepsis raised no clinical suspicion of illness and had no cultures obtained. We measured novel characteristics-moments and percentiles-of normalized heart rate (HR) time series for 5 days before and 3 days after sepsis, sepsis-like illness, or a random time in controls. We also calculated the Score for Neonatal Acute Physiology (SNAP) and the Neonatal Therapeutic Intervention Scoring System (NTISS) as clinical scores of the severity of illness. Results. There were 46 episodes of culture-positive sepsis in 40 patients and 27 episodes of culture-negative sepsis-like illness in 23 patients. We analyzed 29 control periods in 26 patients. Infants with sepsis and sepsis-like illness had lower birth weights and gestational ages and higher SNAP and NTISS scores than did infants without sepsis. The most important new finding was that the infants in the sepsis and sepsis-like illness groups had increasingly abnormal HRC for up to 24 hours preceding their abrupt clinical deterioration. The abnormal HRC were reduced baseline variability and short-lived decelerations in HR. These abnormalities led to significant changes in HRC measures, for example, the third moment (skewness: .59 +/- .10 for sepsis and .51 +/- .12 for sepsis-like illness, compared with -.10 +/- .13 for control over the 6 hours before abrupt deterioration). Culture-positive and culture-negative patients had similar HRC and clinical scores, including a significant rise in SNAP in the 24 hours before the event. Multivariable logistic regression analysis showed that HRC and clinical scores independently added information in distinguishing infants with sepsis and sepsis-like illness from control patients in the 24 hours before abrupt deterioration. Conclusions. Newborn infants who had abrupt clinical deterioration as a result of sepsis and sepsis-like illness had abnormal HRC and SNAP that worsened over 24 hours before the clinical suspicion of sepsis. A strategy for monitoring these parameters in infants at risk for sepsis and sepsis-like illness might lead to earlier diagnosis and more effective therapy.