Late-Onset Sepsis in Very Low Birth Weight Infants

This population-based observational study was based on analysis of data prospectively collected by the Israel National VLBW (BW ≤1500 g) Infant Database on all live infants born between January 1995 and December 2019. All 28 neonatal departments in Israel ( Appendix) participated in the data collection as previously described.^13,14  The data collected included: demographic details, antenatal and perinatal history, postdelivery status, neonatal diagnoses, medical and surgical treatments including episodes of sepsis, infant’s age when late-onset sepsis occurred, causative pathogen for each late-onset sepsis episode, and outcome at discharge. Data were collected on all infants until death or discharge home. A prestructured form was completed for each infant, checked for logic errors, and if necessary, returned to the participating center for clarification. Infants were usually treated from birth to discharge at a single center, regardless of unit size. Interhospital transfers were followed by the database coordinator until final discharge home. Birth hospital and patient identification remain confidential by consensus agreement of all participating centers. All departments used an operating manual and standard definitions based on those of the Vermont-Oxford Network.¹⁵  Inclusion of all VLBW infants was confirmed by cross-checking with the Israel National Birth Registry.

The Israel National VLBW database includes more than 99% of all VLBW infants born in Israel. From 1995 through 2019, the database included records of 38 050 VLBW infants. Figure 1 describes the study cohort after exclusion of infants with congenital anomalies, death before 72 hours of age, and births <23 weeks gestation. The final study population comprised 31 612 VLBW born ≥23 weeks, of whom 7423 (23.5%) had 1 or more episodes of late-onset sepsis.

Late-onset sepsis required a blood culture positive for a causative pathogen obtained 72 hours or more after birth and clinical sepsis symptoms.^13,15  For CONS, the definition of late-onset sepsis before 2016, required 1 positive blood culture, 5 treatment days, and clinical features of sepsis^13,15,16  and from 2016, required 2 positive blood cultures and clinical features of sepsis. In analyzing sepsis episodes, causative pathogens were classified into 9 groups as follows: CONS, Staphylococcus aureus, other gram-positive cocci, Klebsiella species, Escherichia Coli (E. coli), Pseudomonas species, other gram-negative bacteria, all fungi, and mixed organisms. Pathogens considered contaminants included: Micrococcci, Lactobacilli, Diphteroids, Bacillus species, and Bacteroid species and were not included as sepsis episodes. Repeat blood cultures within 7 days that were positive for the same pathogen were considered as a single sepsis episode.

Definitions of pregnancy complications have been previously reported in detail.¹³  The GA in completed weeks was defined as the best estimate of GA based on last menstrual period, obstetric history and examination, prenatal ultrasound, or early postnatal physical examination. Small for gestational age (SGA) was defined as BW below the 10th percentile for GA according to the gender specific growth charts of Kramer et al.¹⁷  We did not use the Israeli growth charts (Dolberg et al¹⁸ ), because of the very small sample at the lower GA's and inaccurate estimation of SGA in the smallest infants.¹⁹  Delivery room resuscitation refers to intensive resuscitation measures including: intubation and/or cardiac compression and/or epinephrine administration.

Bivariate analyses were used to compare characteristics of infants with or without late-onset sepsis. The association between late-onset sepsis and perinatal and neonatal factors was tested using the χ-square test for categorical variables and the 2-sample t test for continuous variables. The mutual effect of variables associated with the existence of late-onset sepsis was assessed using multivariable logistic regression analyses, applying the Generalized Estimating Equations method to control for correlations in infants from multiple births. Results of the multivariable model are presented as adjusted odds ratios (OR) with 95% confidence intervals (CI).

The study period was divided into 4 epochs: 1995 to 2000, 2001 to 2006, 2007 to 2012 and 2013 to 2019. The first epoch served as a baseline to which the subsequent 3 epochs were compared. The length of hospital stay from birth until discharge home or death was determined for each infant and the pathogen specific late-onset sepsis rate for each epoch was calculated per 10 000 hospital admission days. Time trends were analyzed using the multivariable Cox regression models with consideration for repeated late-onset sepsis episodes and reported as hazard ratio’s (HR) with 95% CI’s. Proportionality hazard assumptions of the Cox models were met. For correction of the proportionality assumption violation, the interactions with the survival time were added and the HR with 95% CI were estimated at the median survival time. In view of the potential impact of the change in definition of CONS late-onset sepsis in 2016, we undertook a subgroup analysis excluding 3720 infants with episodes of CONS late-onset sepsis only and all 5041 episodes of CONS late-onset sepsis. In addition, a sensitivity analysis was performed for the whole study population of 31 612 infants for overall rates of late-onset sepsis, excluding all CONS sepsis episodes. All statistical analyses were 2-tailed and P values below .05 were considered statistically significant. Statistical analyses were performed using the SAS statistical software Version 9.4 (SAS Institute, Inc, Cary, NC). Forest plot was created using GraphPad Prism 9 (GraphPad Software, San Diego, CA).

This study was conducted in accordance with prevailing ethical principles and approved by the Human Research Committee of the Sheba Medical Center (SMC 9180-22).

Of 31 612 VLBW infants surviving to 72 hours, 7423 (23.5%) developed late-onset sepsis and 24 189 (76.5%) did not. Single episodes of late-onset sepsis were recorded in 5889 infants, 1151 infants had 2 late-onset sepsis episodes and 383 infants 3 late-onset sepsis episodes. Comparison of infants with and without late-onset sepsis is provided in Table 1. VLBW infants with late-onset sepsis were of a lower mean GA (27.7±2.6 w vs 29.8±2.7 w) and required more delivery room resuscitation (36.6% vs 16.6%) than those without late-onset sepsis. The percent of infants with late-onset sepsis declined from 29.5% in 1995 to 2000 to 13.0% in 2013 to 2019 (P < .0001). The changes over time in late-onset sepsis rates in different GA and BW groups are presented in Table 2. Mortality rates during the initial hospitalization decreased from 9.9% to 8.8%, 8.0%, and 7.2% in the 4 epochs, respectively.

Multivariable analysis of perinatal factors associated with late-onset sepsis among VLBW infants is shown in Fig 2. In comparison with 1995 to 2000, the risk of late-onset sepsis was similar in 2001 to 2006. In contrast, the 2007 to 2012 epoch was associated with a significantly decreased risk of late-onset sepsis (OR 0.71, 95% CI, 0.66–0.77) and the 2013 to 2019 epoch with a markedly lower risk of late-onset sepsis (OR 0.32, 95% CI, 0.29–0.35) in comparison with 1995 to 2000. Lower GA (23–27 w vs 31 w or greater) was strongly associated with increased risk of late-onset sepsis (OR 8.90; 95% CI, 7.85–10.09).

A total of 9340 episodes of late-onset sepsis were recorded; CONS (n = 5041, 54.0%), Staphylococcus aureus (n = 327, 3.5%), other gram-positive cocci (n = 476, 5.1%), Klebsiella sp. (n = 1116, 11.9%), E. coli (n = 383, 4.1%), Pseudomonas sp. (n = 304, 3.3%), other gram-negative bacteria (n = 720, 7.7%), fungi (n = 803, 8.6%), and mixed organisms (n = 170, 1.8%). Temporal changes in the late-onset sepsis rates per 10 000 hospitalization days, by pathogen groups, are presented in Fig 3. Gram-negative and fungal late-onset sepsis rates decreased consistently in all epochs, whereas gram-positive late-onset sepsis rates decreased only in the last epoch. Temporal changes in the late-onset sepsis rates by specific pathogens are presented in Fig 4. late-onset sepsis rates declined for Klebsiella, Pseudomonas, and Staph aureus late-onset sepsis, whereas E. coli late-onset sepsis remained unchanged. The marked decrease in the rate of CONS sepsis in the most recent epoch may reflect the change in diagnostic criteria.

Table 3 shows the effect of epoch on rates of late-onset sepsis by pathogen groups after adjustment for significant demographic and perinatal variables. For all late-onset sepsis episodes, the HR decreased in the final epoch to 0.40 (95% CI, 0.37–0.43) as compared with 1995 through 2000. For gram-positive pathogen late-onset sepsis, the HR decreased significantly only in 2013 through 2019 to 0.47 (95% CI, 0.37–0.59). The HR for gram-negative pathogen late-onset sepsis decreased throughout all epochs to 0.54 (95% CI, 0.48–0.61) in 2013 through 2019. The most dramatic decrease in HR was achieved for fungal late-onset sepsis to 0.17 (95% CI, 0.12–0.22) in 2013 through 2019.

A subgroup analysis was performed, excluding 3720 infants with CONS late-onset sepsis only and all 5098 episodes of CONS late-onset sepsis. This analysis included 27 892 infants of whom 3703 (13.3%) had a total of 4299 episodes of late-onset sepsis. For all late-onset sepsis episodes, the adjusted HR decreased in the final epoch to 0.41 (95% CI, 0.37–0.46) as compared with 1995 through 2000. The HR in the final epoch compared with the initial epoch decreased for gram-positive late-onset sepsis excluding CONS, for gram-negative late-onset sepsis and for fungal late-onset sepsis to 0.42 (95% CI, 0.33–0.54), 0.49 (95% CI, 0.44–0.55), and 0.15 (95% CI, 0.11–0.20), respectively. In a sensitivity analysis undertaken for the whole study population of 31 612 infants, the adjusted HR for overall rates of late-onset sepsis, excluding CONS sepsis, decreased in 2001 through 2006 to 0.89 (95% CI, 0.82–0.96), in 2007 through 2012 to 0.61 (95% CI, 0.56–0.67), and in the final epoch to 0.45 (95% CI, 0.41–0.50) as compared with 1995-2000.

In a national Israeli cohort of VLBW infants, 23.5% had 1 or more episodes of late-onset sepsis. The rate of late-onset sepsis declined over the study period and decreased from 29.5% to 13.0% of the VLBW infant cohort in the final 2013 to 2019 epoch. Late-onset sepsis was most strongly associated with a lower GA. Although the pathogen specific rate of most pathogens declined with time, considerable heterogeneity exists in the magnitude of the observed change.

Many NICU practices may potentially influence late-onset sepsis rates among VLBW infants. Quality improvement measures are often implemented in the NICU, however the evidence that these measures actually achieve improvement is not always present. This is especially true when prevention of late-onset sepsis is the goal because of the multifactorial etiology of late-onset sepsis. The current study examined late-onset sepsis trends over a 25-year period in a national VLBW cohort. We have shown that in the final study epoch (2013–2019), late-onset sepsis rates declined to 25.6 of 10 000 hospital days, representing a greater than 50% decrease in late-onset sepsis rates over time. This decrease is even more impressive when considering the fact that during the study period mortality rates of the most immature infants, who are at the highest risk for late-onset sepsis, decreased. It is evident that during the first 2 epochs, no effective intervention was achieved as late-onset sepsis rates slightly increased. During the third epoch, late-onset sepsis rates decreased, and during the final epoch a more dramatic decrease in late-onset sepsis rates was achieved. In 2016, in Israel, a national prevention program was initiated aimed at reducing late-onset sepsis.²⁰  Multiple measures were implemented, including staff education, improvement of infection control, aseptic technique, and care of central lines and many other measures, however, which specific measures may have resulted in the dramatic decrease in late-onset sepsis rate in the final study epoch is unknown. The Israel national infection prevention program used the Central for Disease Control definitions for late-onset sepsis, which differ somewhat from previously applied definitions. The main change was a stricter definition for CONS late-onset sepsis. This change may have resulted in an overestimation in the magnitude of the decrease of CONS late-onset sepsis, however, it did not influence the rates of late-onset sepsis caused by other pathogens.

Although previous studies have shown conflicting results regarding changes in late-onset sepsis rates, many recent reports support a trend toward a decrease in late-onset sepsis rates. Greenberg et al²¹  have shown in a study on extremely premature infants from the National Institute of Child Health and Human Development Neonatal Research Network, that the incidence of late-onset sepsis decreased over time from 41% to 34% of infants. Boel et al²²  in a study from the UK and Kim et al²³  in a Korean study have both similarly shown trends toward decreased late-onset sepsis in extremely preterm infants. In VLBW infants, in a South American study, D' Apremont et al²⁴  reported a significant decrease in late-onset sepsis rates from 21.1% to 19.5%. The current study is comparable to the previous study regarding the incidence of late-onset sepsis, however it has a higher incidence of late-onset sepsis compared with a report from the German Neonatal Network.⁹  When evaluating the magnitude of the decrease in late-onset sepsis rate achieved, it is important to appreciate that when the baseline late-onset sepsis rate is higher, it is easier to achieve a larger decrease in the rate of late-onset sepsis compared with countries that have an initial low rate. This may be the reason that the German study failed to show a decrease in culture-confirmed late-onset sepsis.

Our study focused on perinatal factors associated with late-onset sepsis. As in previous studies,^8,12,13  GA was the factor most strongly associated with late-onset sepsis. Birth epoch was also associated with risk of late-onset sepsis, similar to the report by Greenberg et al.²¹  Infants requiring delivery room resuscitation are at an increased risk for late-onset sepsis that may be related to increased severity of the infants' initial condition or to an undiagnosed infection. Of note, Cesarean section delivery and antenatal corticosteroids had no significant effect on the risk of late-onset sepsis.

Few studies have examined temporal changes in pathogen specific late-onset sepsis rates in a population-based cohort of infants. Possibly the most important data from this study are changes in pathogen specific late-onset sepsis rates, which provides insight both to the effectiveness of prevention measures employed and to the empirical treatment required when sepsis is considered. During the study period we have shown a consistent decrease in the rate of gram-negative infections. There is no single identifiable measure that is responsible for this decrease, and it is likely related to an overall improvement in the care of VLBW infants and possibly to the increased use of breast milk. Despite the decrease in gram-negative infections, they have become relatively more important because other infections have decreased to a greater extent. Ran et al,²⁵  in a study from the Netherlands, have noted an increase in gram-negative infections among extremely low birth weight infants with late-onset sepsis and show that their rate increased from 0% to 27%, with an overrepresentation among infants with the lowest GA and BW. The increasing importance of gram-negative late-onset sepsis should prompt adequate coverage of these pathogens when choosing initial empirical treatment. Possibly the most striking change among pathogens causing late-onset sepsis is the near elimination of fungal infections. Most NICU's in Israel started antifungal prophylaxis during the third epoch and the remainder in the most recent epoch. It is highly likely that the dramatic decrease in fungal infections was caused by prophylactic antifungal therapy practiced by nearly all NICU’s, and to a lesser extent, to empirical antibiotic regimens, avoiding broad-spectrum antibiotics that are associated with increased fungal infections.²⁶  The very low fungal infection rate among VLBW infant in Israel may lead to discontinuation of antifungal prophylaxis in the future.

The current study is one of the largest to date to evaluate pathogen specific late-onset sepsis rates among VLBW infants. Our study population was a national cohort, exclusion criteria were minimal, and infants were followed when transferred between hospitals, avoiding selection bias. All NICUs in Israel participated and used standard definitions that have undergone minimal changes throughout the study period. A strict definition of late-onset sepsis was used, excluding infants without microbiologic confirmation.

Some study limitations should be acknowledged. We preferred a strict definition of late-onset sepsis, and this may have led to misclassification of a small number of infants in whom the blood culture was considered a contamination. The definition of late-onset sepsis caused by CONS was changed during the study period, preventing us from accurately estimating the magnitude of the decrease in late-onset sepsis caused by CONS in the final study epoch. In view of the potential impact of the change in definition of CONS late-onset sepsis in 2016, we undertook both a subgroup analysis excluding 3720 infants with episodes of CONS late-onset sepsis only and a sensitivity analysis for the whole study population excluding all CONS late-onset sepsis. These analyses however, revealed only minor changes in the study results, and the marked decrease in the HR for late-onset sepsis, especially in the last epoch occurred irrespective of the change in diagnostic criteria for CONS late-onset sepsis. Because of the study duration, some NICU practices have changed and, in particular, during the final epoch when a national program was implemented to reduce the rate of late-onset sepsis. The effect of antibiotic practices on late-onset sepsis could not be estimated because this data are not included in the database.

The strongest risk factor for late-onset sepsis was GA <27 w. Over a period of 25 years, the pathogen specific rates of late-onset sepsis among VLBW infants decreased approximately twofold for gram-positive and gram-negative bacterial infections and sixfold for fungal infections. Although late-onset sepsis rates and gram-negative sepsis rates have decreased, gram-negative bacteria still account for a significant proportion of late-onset sepsis.

Coordinating center: The Women and Children’s Health Research Unit, Gertner Institute for Epidemiology and Health Policy Research, Sheba Medical Center, Tel Hashomer.

Neonatal departments: Assaf Harofeh Medical Center, Rishon Le Zion; Assuta Hospital, Ashdod; Barzilay Medical Center, Ashkelon; Bikur Holim Hospital, Jerusalem; Bnei Zion Medical Centre, Haifa; Carmel Medical Center, Haifa; English (Scottish) Hospital, Nazareth; French Hospital, Nazareth; Hadassah University Hospital Ein-Karem, Jerusalem; Hadassah University Hospital Har Hazofim, Jerusalem; Haemek Medical Center, Afula; Hillel Yafe Medical Center, Hadera; Italian Hospital, Nazareth; Kaplan Hospital, Rehovot; Laniado Hospital, Netanya; Maayanei Hayeshua Hospital, Bnei-Brak; Meir Medical Center, Kefar Saba; Misgav Ladach Hospital, Jerusalem; Naharia Hospital, Naharia; Poria Hospital, Tiberias; Rambam Medical Center, Haifa; Rivka Ziv Hospital, Zefat; Schneider Children's Medical Center of Israel and Rabin Medical Center, Petach-Tikva; Shaare-Zedek Hospital, Jerusalem; Sheba Medical Center, Tel-Hashomer; Soroka Medical Center, Beer-Sheva; Sourasky Medical Center, Tel-Aviv; Wolfson Medical Center, Holon; Yoseftal Hospital, Eilat.

The authors acknowledge Dr Ilya Novikov for his advice and assistance with the statistical analysis of the data.

BW

birth weight

CI

confidence interval

CONS

coagulase negative Staphylococci

E. Coli

Escherichia Coli

EOS

early-onset sepsis

GA

gestational age

HR

hazard ratio

OR

odds ratios

VLBW

very low birth weight