Multiple vital signs analysis algorithm detects systemic inflammatory response in premature infants with late onset sepsis and necrotizing enterocolitis

May 07, 2019


Mithal LB , Yogev R, Palac HL, Kaminsky D, Gur I, Mestan KK. Vital signs analysis algorithm detects inflammatory response in premature infants with late onset sepsis and necrotizing enterocolitis. Early Hum Dev. 2018 Feb; 117:83-89. doi: 10.1016/j.earlhumdev.2018.01.008. Epub 2018 Jan 23. PMID: 29351876


Nikolina Zdraveska, MD, PhD
Assistant Professor of Pediatrics
Department for Neonatology, University Children Hospital Skopje, Macedonia


Prevention, treatment


(P) In a cohort of preterm infants (23-32 GA) (I) is the performance of the revised RALIS algorithm through multiple vital signs analysis useful in detection of systemic inflammatory responses (C) earlier than clinical suspicion in (O) LOS and NEC (T) in the first month of life?


  • Design: observational, retrospective
  • Allocation: observational study – not applicable
  • Blinding: N/A
  • Follow-up period: Upon discharge
  • Setting: A single academic referral center at Prentice Women’s Hospital of Northwestern Medicine in Chicago, IL, USA affiliated with Northwestern University Feinberg School of Medicine
  • Patients: All premature infants <33 GA admitted to NICU between 2008 and 2012
    • Inclusion criteria
      • Complete vital sign data from birth to 28 postnatal day available in the electronic medical record
    • Exclusion criteria
      • Infants transferred out of NICU or death within first month of life
      • Congenital malformations
      • Infants treated with oscillator or jet ventilation due to the inability to interpret correctly the respiratory rates
    • Intervention: RALIS presents a mathematical algoritm that icorporates vital sign data: heart rate, respiratory rate, temperature, desaturation events and bradycardia events and generates a score based on significant vital sign changes from the individual patients’s baseline. Vital sign data were electronicaly entered per routine NICU protocol by experienced nurses. The RALIS score ranged on a 0-10 scale, with ≥5 as cuf-off for acute inflamatory response, generating an alert when lasting at least 6 consecutive hours.
  • Outcomes:
    • Primary outcome:
      • LOS was based on physician initiated sepsis evaluation associated with positive culture (blood, CSF or urine for bacteria or fungi) and antibiotic treatement requirement, occuring at >72 postnatal hours.
      • “Expanded” LOS was defined as positive endotracehal tube cultures, chest radiographic infiltrate, and antibiotic course treatment.
      • Culture negative LOS was defined in patients with negative culture but clinical suspicion of sepsis treated with antibiotic course.
      • NEC was defined with clinical or radiological evidence of NEC, treated with antibiotics and enteral feeding interruption.
      • The control group included the patients with negative blood cultures and no antibiotic treatment during their hospitalization (the patients received antibiotics only in the first 48h to rule-out early onset sepsis).
    • Secondary outcomes: Association of RALIS and sepsis episodes by comparison of the hours between RALIS time alert and the time of clinical suspition for sepsis (defined when sample was sent for culture)
  • Analysis and Sample Size: sample size determination was based on LOS as the primary outcome and hours between RALIS alert and culture as the primary predictor. The mean time difference used was 59±(67) hours in the LOS group and 22±(64) in false positive culture group, according to the previous studies and sufficient statistical power of the cohort was acheived (alpha 0.05). Sensitivity, specificity, PPV and NPV of RALIS was calculated for 1) expanded LOS, 2) LOS, 3) NEC and 4) LOS plus NEC. Exploratory cross validation analysis was implemented for accessing the validity of RALIS alert in LOS prediction using logistic regression modeling. Cross validation analysis was performed to access the robustness of the RALIS method.


There were 161 episodes in 155 patients distributed as follows: 41 cases expanded LOS (blood, CSF, urine, ETT), 31 LOS (blood, urine), 14 culture-negative sepsis, 13 false-positive cultures and 93 controls. The majority of microbiologic isolates were gram negative bacteria and CONS. Gender, multiple gestations and prolonged rupture of membranes were not significantly different amongst groups. Controls had higher birth weight and lower incidence of BPD than other groups. Chorioamnionitis was more prevalent in the expanded LOS and LOS groups.

RALIS associated events for each groups and the calculated sensitivity, specificity, positive predictive value and negative predictive value were as follows:

Variable Control Expanded LOS                   (blood, CSF, urine, ETT) LOS (blood, urine) NEC LOS (blood, urine) and/or NEC Culture negative LOS False positive culture
N 93 41 31 9 38 14 13
RALIS alert associated with culture sent, n (%) 19 (21) 33(81) 25(81) 8(89) 32(84) 11(79) 8(62)
Sensitivity, %


Positive Predictive Value

Negative Predictive Value

N/A 84

















The mean time of the RALIS alert and culture sent was significantly earlier in the LOS group (-43.1±79.4hours; p=0.012, 95% CI: -75.8 to -10.3 hours), as well as in the LOS and/or NEC outcome (-33.0±79.3 hours; p=0.025, 95% CI:-61.5 to-4.4 hours).


The results of the study showed that RALIS multiple vital signs algorithm can early detect systemic inflammatory response in LOS and NEC with mean alert 33 hours before clinical suspicion. Moreover, RALIS algorithm can be more useful in reassuring clinician of absence of LOS and NEC due to the high negative predictive value. Authors conclude that RALIS is a promising tool which can aid in medical decisions in combination with other clinical and laboratory assessment.


Neonatal sepsis substantially contributes to neonatal mortality and morbidity and presents a major public health challenge worldwide.(1) Despite the decline in the incidence of late onset sepsis (LOS) due to infection control measurements, it still disproportionately affects very low birth weight (VLBW) infants causing significant morbidity, lifelong neurologic impairments and prolonged hospitalization.(2,3)

The diagnosis of sepsis remains difficult due to its non-specific clinical presentation which could be attributed to many other non-infectious etiologies, as well as the lack of reliable laboratory biomarkers. Because of the potential dire consequences, clinicians have a low threshold for LOS treatment leading to antibiotic overuse in a large proportion of VLBW infants and its adverse effects such as: increased antibiotic resistance, fungal infections, microbiota modification, necrotizing enterocolitis (NEC) and death.(4) NEC, defined as bowel infection, bacterial penetration and tissue destruction, presents with non-specific signs at early stage, very often overlapping with sepsis.(5) Both neonatal sepsis and NEC are associated with systemic inflammatory responses.

The development of a reliable and specific method to detect an impending infection or systemic inflammatory response prior to the clinical deterioration has been of particular research interest in recent years.(6,7) Previous studies have shown that RALIS software with multiple vital signs monitoring was an effective tool in detecting sepsis 2.5 days earlier than the onset of the clinical suspicion of infection.(8)

The current study aimed to evaluate the performance of the revised RALIS algorithm in the detection of a systemic inflammatory response in LOS and NEC in preterm infants (23-32 GA). Vital sign data incorporated into the revised RALIS algorithm included: heart rate, respiratory rate, temperature, desaturation events and bradycardia events. A score was generated (range: 0-10) and the clinical team was alerted if ≥5 for 6 consecutive hours. Premature infants with complete electronic vital sign data were included and classified according to the primary outcome into LOS, expanded LOS, NEC, culture-negative LOS or controls.

This study consistently proved the effectiveness of the RALIS algorithm in detecting LOS and NEC with a mean alert of 33 hours before clinical suspicion, particularly emphasizing its utility in excluding LOS with a high negative predictive value of 93%. Although sensitive, the positive diagnostic value in identifying LOS and NEC remained limited, showing a moderate PPV value (67%) similar to that in previous reports (8). The authors note the study’s limitations, mainly regarding the sample size and its retrospective nature.

Thus, this study indicates the particular usefulness of the RALIS algorithm in identifying clinician-concern ahead of clinical-concern without consequence, which may more confidently aid the physician’s decision-making with regard to withholding or discontinuing antibiotic treatment.

This study adds additional data regarding the benefit of multiple sign monitoring algorithms in the prediction of neonatal infections; however, future prospective studies are warranted to investigate its association with existing or novel laboratory biomarkers, as well as its utility beyond the first month of life.


  1. Qazi SA, Stoll BJ. Neonatal sepsis: a major global public health challenge. Pediatr Infect Dis J 2009; 28:S1–2
  2. Bizzarro MJ. Shabanova V, Baltimore RS, Dembry LM, Ehrenkranz RA, Gallagher PG. Neonatal Sepsis 2004-2013: The Rise and Fall of Coagulase-Negative Staphylococci. J Pediatr 2015; 166:1193-9
  3. Adams-Chapman I, Stoll BJ. Neonatal infection and long-term neurodevelopmental outcome in the preterm infant. Curr Opin Infect Dis 2006; 19:290-297
  4. Stoll BJ, Hansen N, Fanaroff AA, Wright LL, Carlo WA, Ehrenkranz RA et al. Late-onset sepsis in very low birth weight neonates: the experience of the NICHD Neonatal Research Network. Pediatrics2002; 110:285-91
  5. Srinivasan PS, Brandler MD, D’Souza A. Necrotizing enterocolitis. Clin Perinatol 2008; 35:251-72
  6. Sullivan BA, Grice SM, Lake DE, Moorman JR, Fairchild KD. Infection and other clinical correlates of abnormal heart rate characteristics in preterm infants. J Pediatr 2014; 164:775-80
  7. Gur I, Riskin A, Markel G, Bader D, Nave Y, Barzilay B, Eyal FG, Eisenkraft A.  Pilot study of a new mathematical algorithm for early detection of late-onset sepsis in very low-birth-weight infants. Am J Perinatol. 2015; 32:321-30
  8. Mithal LB, Yogev R, Palac H, Gur T, Mestan KK. Computerized vital sign analysis and late onset infections in extremely low gestational age infants. J Perinat Med 2016; 44:491-7