Abnormal Heart Rate Characteristics Preceding Neonatal Sepsis and Sepsis-Like Illness

Abstract

Late-onset neonatal sepsis is a significant cause of morbidity and mortality, and early detection could prove beneficial. Previously, we found that abnormal heart rate characteristics (HRC) of reduced variability and transient decelerations occurred early in the course of neonatal sepsis and sepsis-like illness in infants in a single neonatal intensive care unit (NICU). We hypothesized that this finding can be generalized to other NICUs. We prospectively collected clinical data and continuously measured RR intervals in all infants in two NICUs who stayed for Ͼ7 d. We defined episodes of sepsis and sepsis-like illness as acute clinical deteriorations that prompted physicians to obtain blood cultures and start antibiotics. A predictive statistical model yielding an HRC index was developed on a derivation cohort of 316 neonates in the University of Virginia NICU and then applied to the validation cohort of 317 neonates in the Wake Forest University NICU. In the derivation cohort, there were 155 episodes of sepsis and sepsis-like illness in 101 infants, and in the validation cohort, there were 118 episodes in 93 infants. In the validation cohort, the HRC index 1) showed highly significant association with impending sepsis and sepsis-like illness (receiver operator characteristic area 0.75, p Ͻ 0.001) and 2) added significantly to the demographic information of birth weight, gestational age, and days of postnatal age in predicting sepsis and sepsis-like illness Abbreviations BW, birth weight GA, gestational age HRC, heart rate characteristics HRV, heart rate variability NICU, neonatal intensive care unit ROC, receiver operating characteristic SampEn, sample entropy UVA, University of Virginia WFU, Wake Forest University Late-onset nosocomial neonatal sepsis occurs after 3 days of life and is a major health problem in high-risk newborn infants in the neonatal intensive care unit (NICU). It occurs in 25% of very low birth weight (Ͻ1500 g) infants and leads to a more than doubling of mortality and a 50% increase in hospital stay (1). Early detection is very difficult using clinical signs and currently available laboratory tests (2, 3). Clinicians have found neither to be reliable (4, 5), and the diagnostic challenge is manifest in the Centers for Disease Control and Prevention's definition that allows diagnosis of neonatal "clinical sepsis" with either a negative blood culture or no blood culture at all (6). These high-stakes uncertainties lead to two shortcomings of the current practice. First, infants with sepsis are often detected only when seriously ill, increasing morbidity and mortality and lowering the chance for prompt, complete recovery with antibiotic therapy. Second, physicians have a very low threshold for suspecting sepsis, and as many as 10 -20 infants are treated unnecessarily for every infant who truly is septic (3-5). The National Institute of Child Health and Human 0031-3998/03/5306-0920 PEDIATRIC RESEARCH Vol. 53, No. 6, 2003 Copyright © 2003 Printed in U.S.A. ABSTRACT920

Autoregressive Hidden Markov Models for the Early Detection of Neonatal Sepsis

“…For the episodebased analysis, we use precision-recall (PR) curves 3 [29]. We report the average precision (AP) [30] and the maximum F-score 4 . Both the second-by-second and the episode-based analyses are projections of the inferences onto different metrics and can reveal different performance aspects.…”

“…Nevertheless, it is acknowledged that the test can have a poor accuracy [1], [4], [9]. First, small sample volumes and antibiotic therapy can give false-negative results [1].…”

“…Because of the dangers of delaying treatment, antibiotic therapy is usually started at the same time as taking the blood sample. However, applying low thresholds in suspecting sepsis results in a high number of patients being treated unnecessarily for each true case [4]. Thus, if achievable, a reliable early detection of sepsis based on monitoring data would be of great value.…”

“…The HRC are then fed to a logistic regression classifier. A larger dataset was employed for demonstrating that HRC add predictive information to a classifier using only demographic features to discriminate sepsis and sepsis-like illness patients from controls [4]. More precisely, they showed an increase in AUC (area under ROC curve) from 0.72 to 0.77 on a test set.…”

Lipopolysaccharide prolongs action potential duration in HL-1 mouse cardiomyocytes

“…Projecting in vitro observations to heart function in situ, these findings account at least in part for reduced heart rate variability (8,30,31,38) and depressed cardiac function that accompanies sepsis (36,44). This may be particularly relevant to neonatal sepsis where mortality is often preceded by reduced heart rate variability and transient decelerations (10,11).…”