Heart rate and oxygen saturation patterns in very low birth weight infants with early onset sepsis and histologic chorioamnionitis

Abstract

BACKGROUND:

Chorioamnionitis and early onset sepsis (EOS) in very low birth weight (VLBW,< 1500 g) infants may cause a systemic inflammatory response reflected in patterns of heart rate (HR) and oxygenation measured by pulse oximetry (SpO₂). Identification of these patterns might inform decisions about duration of antibiotic therapy after birth.

OBJECTIVE:

Compare early HR and SpO₂ patterns in VLBW infants with or without early onset sepsis (EOS) or histologic chorioamnionitis (HC).

STUDY DESIGN:

Retrospective study of placental pathology and HR and SpO₂ in the first 72 h from birth in relation to EOS status for inborn VLBW NICU patients 2012–2019.

RESULT:

Among 362 VLBW infants with HR and SpO₂ data available, clinical, or culture-positive EOS occurred in 91/362 (25%) and HC in 81/355 (22%). In univariate analysis, EOS was associated with higher mean HR, lower mean SpO₂, and less negative skewness of HR in the first 3 days after birth. HC was associated with higher standard deviation and skewness of HR but no difference in SpO₂. In multivariable modeling, significant risk factors for EOS were mean HR, gestational age, HC, mean SpO₂, and skewness of SpO₂.

CONCLUSION:

HR and SpO₂ patterns differ shortly after birth in VLBW infants exposed to HC or with EOS, likely reflecting a systemic inflammatory response.

Keywords

Chorioamnionitis early onset sepsis heart rate oxygenation premature infants very low birth weight vital signs

1 Introduction

Early-onset sepsis (EOS) causes significant neonatal morbidity and mortality, particularly for preterm infants. The incidence of EOS remains low among all live births (0.5–0.8 cases/1000), but is highest among very low birth weight (VLBW) infants (15–19 cases/1000) [1]. Due to the common presence of antenatal risk factors, the non-specific signs of illness, and the potentially lethal consequences of untreated sepsis, the majority of VLBW infants receive empiric antibiotics after birth, and 30–40% are treated for clinical sepsis with prolonged empiric antibiotics despite a negative blood culture [2, 3].

The systemic inflammatory response associated with sepsis may be reflected in vital sign instability [4]. EOS is usually associated with an ascending amniotic infection leading to chorioamnionitis, and an associated fetal inflammatory response may impact HR patterns of the fetus and the newly born infant [5 –7]. Abnormal heart rate characteristics have been well described in cases of late-onset neonatal sepsis by our group and others [8 –10]. In the current work we sought to determine whether VLBW infants with exposure to histologic chorioamnionitis (HC) or with culture-positive or clinical EOS have distinct HR patterns from infants without these inflammatory conditions. Additionally, we studied patterns of systemic oxygenation measured by pulse oximetry (SpO₂), since the systemic inflammatory response may impact lung function and control of breathing. Our primary hypothesis was that infants with EOS would have different HR and SpO₂ patterns in the first 3 days after birth compared to those with sepsis ruled out or those with no blood culture or antibiotics. Our secondary hypothesis was that VLBWs exposed to HC, with or without EOS, would have different HR and SpO₂ patterns compared to those with no significant placental inflammation on histopathology.

2 Methods

2.1 Patient population and study design

We performed a retrospective cohort study at the University of Virginia Neonatal Intensive Care Unit, an academic Level IV NICU. The Institutional Review Board approved the study with waiver of consent. All inborn VLBW (≤1500 grams birthweight) infants < 32 weeks’ gestation admitted from 2012–2019 with at least 4 hours of continuous vital sign data available in the first 24 hours after delivery were included.

2.2 Data collection and definitions

We collected heart rate (HR) and oxygen saturation (SpO₂) data every 2 seconds from the bedside monitor for all infants, using the BedMaster system (Hillrom, Chicago, IL; formerly Excel Medical, Jupiter, FL). SpO₂ was measured with the default averaging time of 8 seconds, using Masimo technology. Clinical data, blood culture results, and histologic placental pathology findings were extracted from electronic health records. Early-Onset Sepsis (EOS) was defined as positive blood culture within 72 hours of birth not considered a contaminant and treated with at least 5 days of antibiotics (EOS) or negative blood culture within 72 hours of birth treated with at least 5 days of antibiotics due to clinical risks and signs of sepsis. The designation of no EOS included cases of sepsis ruled out (negative blood culture within 72 hours of birth treated with < 3 days of antibiotics) or sepsis not suspected (no blood culture or antibiotics within 3 days of birth). Histologic chorioamnionitis was determined by histopathological examination of the placenta, which is routine for very preterm births at our center. We defined HC as grade 2, stage 2–3 maternal and/or fetal inflammatory response (chorioamnionitis or funisitis) according to Amsterdam criteria [11].

2.3 Vital sign data analysis

From the raw data, sampled every 2 seconds, we calculated the mean, standard deviation, skewness, and kurtosis of HR and SpO₂ hourly in the first 72 hours after birth. These calculations include the four mathematical moments that describe the shape of a histogram of the data in each one-hour window. To determine whether these HR and SpO₂ features were different between infants grouped by EOS or HC, we used linear mixed effects models of available hourly measurements from 6–72 hours after birth. The response variables (HR and SpO₂ features) were standardized using the rankit transformation of the percentile of each infant at the same number of hours since birth. The fixed effects were EOS or HC and hours since birth. The random effects were hours since birth clustered by infant.

2.4 Multivariable modeling

We used multivariable ridge regression. We optimized the ridge penalty term using 10-fold cross-validated area under the receiver operating characteristic curve (cvAUC) split by patients, and adjusted for repeated measures using Huber-White standard errors. We evaluated model performance using the 10-fold cvAUC and calibration plot, and assessed variable importance by chi-squared statistics and decreased cvAUC after feature permutation.

3 Results

In the 8-year period of study, 362 VLBW infants had at least 4 hours of HR and SpO₂ data available in the first day after birth. Of these, 9 (2%) had blood culture-positive sepsis, 82 (23%) clinical sepsis, 187 (52%) sepsis ruled out, and 84 (23%) sepsis not suspected within 72 hours of birth. Organisms isolated in the nine cases of culture-positive sepsis included two each of Escherichia coli and Hemophilus influenza and one each of Group B Streptococcus, Enterococcus faecalis, and Staphylococcus epidermidis. Due to the small number of cases of positive blood culture sepsis and the overlapping treatment approach, we analyzed cases of clinical and culture-positive sepsis grouped as EOS and cases of sepsis ruled out or not suspected as no EOS. Table 1 shows the patient characteristics for these two groups. Sex, race, and ethnicity were similar between groups but mean GA and BW were lower for infants with EOS. Infants with EOS were more likely to require intubation for surfactant or mechanical ventilation in the first 72 hours after birth than those without (75% with EOS vs 39% without EOS).

Table 1
Characteristics of infants with and without confirmed or clinical EOS

Overall (N = 362) EOS (N = 91) No EOS (N = 271)

Gestational Age* (weeks) 28.4 (2.4) 27.3 (2.2) 28.8 (2.3)

Birthweight (grams)* 1059 (282) 964 (280) 1091 (277)

Male sex 169 (46.7%) 43 (47.3%) 126 (46.5%)

Race

Black 80 (22.1%) 15 (16.5%) 65 (24.0%)

White 255 (70.4%) 70 (76.9%) 185 (68.3%)

Other or Unknown 27 (7.5%) 6 (6.6%) 21 (7.75%)

Hispanic ethnicity 30 (8.3%) 7 (7.7%) 23 (8.5%)

Histologic chorioamnionitis* 81 (22.3%) 32 (35.2%) 49 (18.0%)

Cesarean delivery 274 (75.7%) 69 (75.8%) 205 (75.6%)

Died 19 (5.3%) 4 (4.4%) 15 (5.54%)

	Overall (N = 362)	EOS (N = 91)	No EOS (N = 271)
Gestational Age* (weeks)	28.4 (2.4)	27.3 (2.2)	28.8 (2.3)
Birthweight (grams)*	1059 (282)	964 (280)	1091 (277)
Male sex	169 (46.7%)	43 (47.3%)	126 (46.5%)
Race
Black	80 (22.1%)	15 (16.5%)	65 (24.0%)
White	255 (70.4%)	70 (76.9%)	185 (68.3%)
Other or Unknown	27 (7.5%)	6 (6.6%)	21 (7.75%)
Hispanic ethnicity	30 (8.3%)	7 (7.7%)	23 (8.5%)
Histologic chorioamnionitis*	81 (22.3%)	32 (35.2%)	49 (18.0%)
Cesarean delivery	274 (75.7%)	69 (75.8%)	205 (75.6%)
Died	19 (5.3%)	4 (4.4%)	15 (5.54%)

Variables are summarized as mean (standard deviation) or number (%). EOS was defined as a blood culture for signs and symptoms of sepsis within 72 hours of birth and treated for at least five days with antibiotics, whether or not the culture was positive. *p < 0.05 by Wilcoxon Rank sum test or Chi-squared test.

Figure 1 shows HR and SpO₂ features calculated hourly over 72 hours from birth, grouped by sepsis category. Mean HR was higher and mean SpO₂ lower for those with EOS compared to those without. Standard deviation and kurtosis of HR and SpO₂ were similar between groups. Skewness of HR (which represents asymmetry of a histogram of HR values) was not significantly different between the two sepsis groups, but skewness of SpO₂ was higher (less negative) for infants with versus those without EOS.

Fig. 1

HR and SpO₂ characteristics, by EOS category. Mean, standard deviation (SD), skewness and kurtosis of HR and SpO₂ calculated hourly over the first 72 hours after birth. Infants with clinical or culture-positive early-onset sepsis (EOS) had higher mean HR (p < 0.001), lower HR standard deviation (p = 0.008), and no significant difference of HR skewness and HR kurtosis. Infants with EOS had lower mean SpO₂ (p < 0.001), higher SpO₂ skewness (p < 0.001), lower SpO₂ kurtosis (p < 0.001) and no significant difference of SpO₂ standard deviation.

Placental pathology was available for 355 (98%) infants. HC was present in 22% overall and more common in infants with versus those without EOS (35% vs 18%, p < 0.05). Figure 2 shows hourly HR and SpO₂ features in the first 72 hours after birth for infants with and without exposure to HC. Mean HR and SpO₂ were similar, but HR standard deviation and skewness were higher for those infants with HC exposure (Fig. 2).

Fig. 2

HR (top) and SpO₂ (bottom) metrics for infants exposed to histologic chorioamnionitis (red) compared to no chorioamnionitis (blue). The lines are hourly average of metrics and the ribbon indicates approximated 95% CI. The vertical dashed line indicates 6 hours since birth.

We used ridge regression to analyze the multivariable relationship of vital sign patterns and clinical variables (GA, BW, HC) with the outcome of EOS. The overall model had a cross-validated AUC of 0.723 (95% CI, 0.719–0.726). Using chi-squared tests, GA, HR mean, SpO₂ mean, and SpO₂ skewness were significant risk factors for EOS in the model. Figure 3 illustrates feature importance in the model by decrease in cvAUC after removing each variable through permutation.

Fig. 3

Physiologic and clinical risk markers of EOS in a multivariable prediction model. The figure shows variables in descending order of the strength of their association with EOS, determined using variable permutation. This method involves calculating the cross-validated AUC of the full model, then re-running it with one variable left out and calculating the decrease in cross-validated AUC from the full model and repeating this step with each variable. The vertical line indicates the AUC of the overall model, which was 0.723 (95% CI, 0.719–0.726). Red dots indicate statistically significant predictors, determined by chi-squared tests (p < 0.05).

4 Discussion

A systemic inflammatory response can manifest as alterations in vital sign patterns due to changes in the autonomic control of heart rate and breathing. In a large cohort of VLBW infants with continuous HR and SpO2 data and placental pathology available, we found abnormal HR and SpO2 patterns associated with exposure to histologic chorioamnionitis and confirmed or clinical EOS. Compared to infants without EOS, infants with EOS had higher mean HR, lower mean SpO₂, and higher skewness of SpO₂. Infants exposed to HC had higher HR variability and HR skewness compared to those without. In multivariable analysis accounting for GA, BW, and exposure to HC, mean HR and SpO2 were associated with EOS. Understanding physiologic markers of illness in this population could help to guide antibiotics and monitoring.

Our results confirm the findings of other studies showing that premature infants with EOS have higher mean HR, even after adjusting for differences across gestational ages [12, 13]. Confirmed EOS with bacteremia was rare in our cohort and occurs rarely in similar cohorts despite a nearly 10-fold higher incidence among preterm infants than term infants. Clinical sepsis, or sepsis-like inflammatory illness, was common in our cohort. This diagnosis is subjective and therefore variable across NICUs [14]. We grouped cases of EOS and clinical sepsis since these infants have similar presentations, risk factors, and levels of illness severity. We speculate that the mechanism of increased HR in sepsis or sepsis-like illness is multifactorial, involving activation of the sympathetic nervous system, hypovolemia due to capillary leak from pathogen toxins and cytokines, or decreased cardiac contractility with a compensatory increase in HR to maintain cardiac output [15]. The difference in mean HR in infants with EOS compared to those without sepsis was about 10 beats per minute (∼150 versus ∼140 bpm), and most infants with EOS were not tachycardic [16].

We also found that oxygenation, as measured by SpO₂, was different in the first 72 hours after birth in infants with versus without EOS. We used calculated SpO₂ features that describe the shape of a histogram of the data and found that SpO₂ skewness was associated with EOS after accounting for the other features and clinical variables. We note that skewness of SpO₂ is inversely related to mean SpO₂, in that infants with higher SpO₂ are more likely to have small desaturations shifting the SpO₂ histogram to the left, or more negative skewness, whereas infants with lower SpO₂ have a more balanced histogram with increases and decreases in SpO₂ and a higher (less negative) skewness value.

The mechanism of lower mean SpO₂ in infants with clinical or culture-positive EOS could be lung inflammation or pulmonary edema as part of a systemic inflammatory response. For infants not on mechanical ventilation, central apnea may occur which could lower the mean SpO₂ [17]. All VLBW infants are prone to apnea and those with sepsis may have increased apnea due to release of endogenous prostaglandins or inflammatory cytokines that impact control of breathing [18, 19]. Pulmonary hypertension during sepsis might also contribute to lower SpO₂ though preterm infants are less susceptible to this than term infants [20]. Similar to the differences in HR, we note that SpO₂ means, while statistically different between groups, fell within the typical target range for preterm infants (∼92% for infants with EOS versus ∼94% for those without EOS). Furthermore, infants with EOS were more premature and therefore more likely to require mechanical ventilation or surfactant administration. Despite using the same SpO2 targets and alarms for all infants, differences in respiratory support and lung disease between groups may have impacted the frequency and severity of intermittent hypoxia and the ability to titrate FiO2 to achieve SpO2 values in the goal range.

Histologic chorioamnionitis was present in over half of cases of culture-positive EOS and about a third of cases of clinical sepsis. We chose to focus on histologic rather than clinical chorioamnionitis since clinical signs such as elevated maternal temperature may occur in absence of infection [21]. Placental inflammation may also occur in absence of a bacterial infection [22] but nonetheless reflects an intrauterine inflammatory process that might impact vital signs after birth. While we did not find a significant difference in mean HR of infants with or without intrauterine HC exposure, we found higher SD and skewness of HR. The increased SD HR was unexpected since reduced HR variability is a signature of late-onset sepsis in premature infants [23]. Furthermore, HR decelerations are a signature of late-onset sepsis, while the association we found between a more positive HR skewness and EOS indicates more HR accelerations. Sympathetic nervous system activation causes accelerations of HR, so these contrasting results may indicate that the systemic inflammatory response to EOS triggers a predominantly sympathetic autonomic nervous system response while late-onset sepsis triggers a predominantly parasympathetic response [24]. Further work is needed to understand this novel finding.

Developing a prediction model for EOS in very preterm infants that informs decisions about testing and therapies is fraught with difficulty and was not our intent in performing these analyses. The EOS calculator [25] developed for risk assessment of term and late preterm infants is predicated on a very low pre-test probability (< 1:1000 term versus > 1:100 very preterm infants have EOS), and on the fact that most at-risk term infants have no signs of sepsis at birth. In the EOS calculator, the presence of respiratory distress indicates clinical illness associated with higher risk of EOS and therefore the tool recommends obtaining a blood culture and initiating antibiotics [25]. Since nearly all very preterm infants have some degree of respiratory distress, decisions about antibiotics must be made based on perinatal risk factors [26] and clinical signs beyond those expected for a typical very preterm infant. In most cases, even signs of significant illness, such as the need for high-level respiratory support or hypotension, are due to prematurity and not bacterial infection. In spite of this, convincing practitioners to discontinue antibiotics on very preterm infants with non-specific clinical signs that are likely not attributable to sepsis is not a simple proposition.

The strength of our study is the large number of VLBW preterm infants for whom we had both placental histopathology and continuous HR and SpO₂ data archived, but we note several limitations. First, the low rate of culture-positive sepsis and the overlap of clinical presentation with culture-negative sepsis led us to combine these groups. We were unable to ascertain the reason for extending antibiotic duration for culture-negative sepsis but in most cases, it included some combination of major antenatal risk factors, severe illness, abnormal laboratory testing results, and, perhaps, concerning vital sign trends. Presence or absence of histologic chorioamnionitis would not have factored into the decision about whether or not to continue antibiotics since pathology results at our institution are typically not available until beyond day five. Another limitation of our study is that we were unable to account for the many clinical variables such as medications and level of respiratory support that may impact HR and SpO₂ and influence decisions about duration of antibiotic therapy for VLBW infants. Finally, we did not analyze all vital signs that may be impacted by a systemic inflammatory response, but restricted the work to HR and SpO₂ since they are continuously monitored in all VLBW NICU patients. Despite these limitations, our results add to what is known on the pathophysiology of VLBW preterm infants exposed to inflammation in the perinatal period.

5 Conclusions

VLBW preterm infants with suspected or confirmed EOS have higher HR and lower SpO₂ in the first few days after birth compared to those without sepsis. Exposure to HC did not significantly impact mean HR, but was associated with higher standard deviation and skewness of HR. These vital sign differences likely reflect a systemic inflammatory response and may be associated with adverse outcomes.

Footnotes

Acknowledgments

This work was supported by the following awards granted by the National Institute of Health: HD 097254 (PI B. Sullivan) and HD0972071 (PI Fairchild).

Conflict of interest

None declared.

Disclosure statement

None.

Human subjects

This study was approved by the University of Virginia IRB with waiver of informed consent.

References

Flannery

, Edwards

, Puopolo

, Horbar

. Early-Onset Sepsis Among Very Preterm Infants. Pediatrics. 2021;7.

Greenberg

, Chowdhury

, Hansen

, Smith

, Stoll

, Sánchez

, et al. Prolonged duration of early antibiotic therapy in extremely premature infants. Pediatr Res. 2019;85(7):994–1000.

Puopolo

, Benitz

, Zaoutis

, Committee on fetus and newborn, committee on infectious diseases. Management of neonates born at ≥35 0/7 weeks’ gestation with suspected or proven early-onset bacterial sepsis. Pediatrics. 2018;142(6).

Sullivan

, Fairchild

. Vital signs as physiomarkers of neonatal sepsis. Pediatr Res. 2022;91(2):273–82.

Kim

, Romero

, Chaemsaithong

, Chaiyasit

, Yoon

, Kim

. Acute chorioamnionitis and funisitis: Definition, pathologic features, and clinical significance. Am J Obstet Gynecol. 2015;213(4 Suppl):S29–52.

Jung

, Romero

, Yeo

, Diaz-Primera

, Marin-Concha

, Para

, et al. The fetal inflammatory response syndrome: The origins of a concept, pathophysiology, diagnosis, and obstetrical implications. Semin Fetal Neonatal Med. 2020;25(4):101146.

Budal

, Bentsen

MHL

, Kessler

, Ebbing

, Lindemann

, Haugen

, et al. Histologic chorioamnionitis in extremely preterm births, microbiological findings and infant outcome. J Matern Fetal Neonatal Med. 2023;36(1):2196599.

Kausch

, Brandberg

, Qiu

, Panda

, Binai

, Isler

, et al. Cardiorespiratory signature of neonatal sepsis: Development and validation of prediction models in 3 NICUs. Pediatr Res. 2023;93(7):1913–21.

Cabrera-Quiros

, Kommers

, Wolvers

, Oosterwijk

, Arents

, van der Sluijs-Bens

, et al. prediction of late-onset sepsis in preterm infants using monitoring signals and machine learning. Crit Care Explor. 2021;3(1):e0302.

10.

Honoré

, Forsberg

, Adolphson

, Chatterjee

, Jost

, Herlenius

. Vital sign-based detection of sepsis in neonates using machine learning. Acta Paediatr. 2023;112(4):686–96.

11.

Khong

, Mooney

, Ariel

, Balmus

NCM

, Boyd

, Brundler

M-A

, et al. Sampling and definitions of placental lesions: Amsterdam placental workshop group consensus statement. Arch Pathol Lab Med. 2016;140(7):698–713.

12.

Graves

, Rhodes

. Tachycardia as a sign of early onset neonatal sepsis. Pediatr Infect Dis. 1984;3(5):404–6.

13.

Nguyen

, Vandenbroucke

, Hernández

, Pham

, Beuchée

, Pladys

. Early-onset neonatal sepsis is associated with a high heart rate during automatically selected stationary periods. Acta Paediatr. 2017;106(5):749–54.

14.

Cantey

, Prusakov

. A proposed framework for the clinical management of neonatal “culture-negative” sepsis. J Pediatr. 2022;244:203–11.

15.

Wynn

, Wong

. Pathophysiology of neonatal sepsis. Fetal and neonatal physiology. Elsevier; 2017. pp. 1536-1552.e10.

16.

Alonzo

, Nagraj

, Zschaebitz

, Lake

, Moorman

, Spaeder

. Heart rate ranges in premature neonates using high resolution physiologic data. J Perinatol. 2018;38(9):1242–5.

17.

Fairchild

, Lake

. cross-correlation of heart rate and oxygen saturation in very low birthweight infants: Association with apnea and adverse events. Am J Perinatol. 2018;35(5):463–9.

18.

Siljehav

, Hofstetter

, Leifsdottir

, Herlenius

. Prostaglandin E2 mediates cardiorespiratory disturbances during infection in neonates. J Pediatr. 2015;167(6):1207–13.e3.

19.

Balan

, Kc

, Hoxha

, Mayer

, Wilson

, Martin

. Vagal afferents modulate cytokine-mediated respiratory control at the neonatal medulla oblongata. Respir Physiol Neurobiol. 2011;178(3):458–64.

20.

Lai

M-Y

, Chu

S-M

, Lakshminrusimha

, Lin

H-C

. Beyond the inhaled nitric oxide in persistent pulmonary hypertension of the newborn. Pediatr Neonatol. 2018;59(1):15–23.

21.

, Zheng

, Zhu

, Chai

. A retrospective study of risk factors for early-onset neonatal sepsis with intrapartum maternal fever. PeerJ. 2022;10:e13834.

22.

Sebire

. Implications of placental pathology for disease mechanisms; methods, issues and future approaches. Placenta. 2017;52:122–6.

23.

Sullivan

, Grice

, Lake

, Moorman

, Fairchild

. Infection and other clinical correlates of abnormal heart rate characteristics in preterm infants. J Pediatr. 2014;164(4):775–80.

24.

Fairchild

, O’Shea

. Heart rate characteristics: Physiomarkers for detection of late-onset neonatal sepsis. Clin Perinatol. 2010;37(3):581–98.

25.

Kuzniewicz

, Puopolo

, Fischer

, Walsh

, Li

, Newman

, et al. A quantitative, risk-based approach to the management of neonatal early-onset sepsis. JAMA Pediatr. 2017;171(4):365–71.

26.

Puopolo

, Mukhopadhyay

, Hansen

, Cotten

, Stoll

, Sanchez

, et al. Identification of extremely premature infants at low risk for early-onset sepsis. Pediatrics. 2017;140(5).