Evaluation of the Neonatal Sequential Organ Failure Assessment and Mortality Risk in Preterm Infants With Necrotizing Enterocolitis

Abstract

Introduction:

The neonatal sequential organ failure assessment (nSOFA) score is a tool for calculating mortality risk of infants in the neonatal intensive care unit. The utility of the nSOFA in determining the risk of mortality or the association with surgical intervention among infants with necrotizing enterocolitis (NEC) has not been investigated.

Methods:

We performed a retrospective, cohort study of preterm (<37 weeks) infants with NEC Bell’s stage ≥IIA at six hospitals from 2008–2020. A nSOFA score (range 0–15) was assigned to each patient at nine time-points from 48 hours before or after clinical illness was suspected.

Results:

Of the 259 infants, nSOFA scores for infants who died (n=39) or had the composite outcome of surgery or death (n=114) were significantly higher (P<0.05) early in the NEC course compared to nSOFA scores for infants who survived medical NEC. Twelve hours after evaluation, the AUC was 0.87 (95% confidence interval (CI), 0.80–0.93) to discriminate for mortality and 0.84 (95% CI, 0.79–0.90) for surgery or death (P<0.001). A maximum nSOFA score of ≥4 at −6, 0, 6, or 12 hours following evaluation was associated with a 20-fold increase in mortality and 19-fold increase in surgery or death compared with a score of <4 (P<0.001).

Conclusion:

In this multi-center cohort, the nSOFA score was able to discriminate well for death as well as surgery or death among infants with NEC. The nSOFA is a clinical research tool that may be used in infants with NEC to improve classification by objective quantification of organ dysfunction.

Introduction

Necrotizing enterocolitis (NEC) is a devastating gastrointestinal disease affecting preterm infants with 15–30% mortality [1]. NEC survivors are at an increased risk for long-term morbidities including short bowel syndrome, stricture formation, and developmental delay [1]. The definition and stage of NEC integrate radiographic, laboratory, and qualitative illness severity criteria [2]. Many of these clinical staging tools feature ordinal designations, such as mild, moderate, and severe disease, applied to poorly defined clinical variables (e.g., lethargy, apnea, bradycardia, abdominal distension, and hypotension), as well as imprecise radiographic and laboratory criteria (e.g., intestinal dilation, ileus, thrombocytopenia, and respiratory/metabolic acidosis) [2]. Although clinical management for NEC has remained unchanged for decades, many novel potential therapeutics directed at prevention and treatment based on the underlying mechanisms of pathology are in development [3–5]. Clinical trials that test novel therapies, as well as their future implementation into clinical practice, would greatly benefit from objective, real-time, patient-specific classification of illness severity.

Illness severity scores are used in adult and pediatric intensive care units to inform clinical decisions and classify patients for research purposes. The sequential organ failure assessment (SOFA) score quantifies illness severity for sepsis, predicts short and long-term outcomes, evaluates biomarker strength, measures therapeutic effectiveness, and guides antibiotic de-escalation [6, 7]. The pediatric SOFA (pSOFA) score has shown utility for predicting mortality and defining sepsis in critically-ill children [8]. The neonatal sequential organ failure assessment (nSOFA) score is an adaptation of the SOFA and pSOFA applicable to the neonatal intensive care unit (NICU) population [9]. In a large, multi-center study, the nSOFA score demonstrated very good discrimination for mortality among very low birth weight (VLBW) infants <32 weeks gestation with late-onset infection (infection occurring at >72 hours of life) [10]. Another multi-center study of 20,152 infants validated the nSOFA as a prognostic tool for all-cause mortality [11]. Due to the unique nature of neonatal physiology, previously described NEC severity scores have not shown adequate predictive power to guide management and rely on subjective radiologic or physical exam criteria [12]. In this multi-center, retrospective cohort study, we measured the utility and generalizability of the nSOFA in the identification of preterm infants with NEC at the highest risk for death or the combined outcome of death or surgical intervention.

Methods

Study design

A retrospective, multi-center cohort design was used to study NEC at six large academic medical centers: Beth Israel Deaconess Medical Center, Boston Children’s Hospital, St. Louis Children’s Hospital, Texas Children’s Hospital, University of Florida Health Shands Children’s Hospital, and University of Virginia Children’s Hospital. Following site Institutional Review Board approvals that included a waiver of consent, data were collected using a standardized set of criteria. All events analyzed occurred between 2008 and 2020.

Inclusion and Exclusion Criteria

Preterm (<37 week) infants with a first-time diagnosis of modified Bell’s stage of IIA or higher NEC were studied [2]. NEC was defined as an abdominal x-ray with pneumatosis intestinalis, portal venous gas, or pneumoperitoneum, an abdominal ultrasound with pneumatosis, or a pathological diagnosis with intestinal necrosis at surgery or autopsy. Exclusion criteria were antibiotic administration in the 48 hours preceding NEC evaluation, congenital anomalies, incomplete medical records, or perforation at less than ten days of life.

Data collection

Evaluation for NEC (0 hours) was defined as the time at which a blood culture was drawn due to a change in clinical status associated with the diagnosis of NEC as detailed above. We evaluated time points beginning 48 hours prior to evaluation (−48 hours) and finished at the completion of antibiotic treatment for NEC or death while still receiving antibiotic treatment. The time point of −48 hours was chosen to determine the utility of the nSOFA score prior to the onset of known cases of NEC. We suspected that there would not be detectable physiologic changes related to the development of NEC prior to this time point. Surgical intervention was defined as peritoneal drain placement or laparotomy. NEC totalis was defined as necrosis of the majority of the large and small intestine. Demographic, laboratory, radiographic, and clinical data were collected. Data were collected on survival through discharge to a non-hospital setting for all surviving patients.

Neonatal sequential organ failure assessment (nSOFA)

The nSOFA score ranges from 0–15 and is calculated using the presence of organ dysfunction (eTable 1 in Supplement; online calculator available at https://peds.ufl.edu/apps/nsofa/default.aspx/). The nSOFA score and its component respiratory, hematologic, and cardiovascular scores were assigned at nine time points (−48 hours, −24, −12, −6, 0: evaluation time, +6, +12, +24, +48) to quantify organ dysfunction prior to illness, during treatment and death [10].

Analytical Methods

Demographic data were compared using Mann-Whitney and Chi-square tests. Comparison of changes in nSOFA scores and between groups were performed using a Kruskal-Wallis with Dunn’s multiple comparisons test. These data are represented in density plots (violin plots), which display the median and IQR of the data. In addition, these plots provide a graphic representation of the distribution of data points, as the horizontal width of a density curve is proportional to the number of patients with a specific nSOFA score. The area under the receiver operating characteristic curve (AUC) was calculated using the Wilson/Brown method. Odds ratios (OR) were calculated for specific nSOFA scores with comparisons made by Chi-squared tests.

Results

Cohort characteristics

We identified 259 infants that met our inclusion criteria. Of the 259 infants, 145 survived medical NEC (56%), 75 survived surgical NEC (29%), and 39 died (15%) (Table 1). There were 114 (44%) infants who were included in the composite outcome of death or surgical intervention. Neonates that survived surgical

Table 1.

Characteristics of cohort

	Survivors: Medical NEC	Survivors: Surgical NEC	P value a	Non-Survivors	P value b
Characteristic	n=145	N=75		n=39
Male sex, No. (%)	78 (54)	41 (55)	0.90d	22 (56)	0.77d
C-section delivery, No. (%)	97 (67)	61 (81)	0.02d	21 (54)	0.13d
Histologic chorioamnionitis, No. (%)e	30/130 (23)	18/69 (26)	0.64d	17/37 (46)	0.006d
Gestational age, mean, (SD), wk	29.1 (3.2)	27.2 (3.2)	<0.001c	25.9 (2.6)	<0.001c
Birth weight, mean (SD), g	1227 (538)	980 (463)	<0.001c	820 (345)	<0.001c
Age at evaluation, mean, (SD), d	22.4 (17.2)	21.0 (16.0)	0.70c	23.0 (15.3)	0.57c
PMA at evaluation, mean, (SD), wk	32.3 (2.9)	30.1 (3.1)	<0.001c	29.3 (2.7)	<0.001c
Positive blood culture at evaluation, No. (%)	19 (13)	25 (33)	<0.001d	16 (41)	<0.001d
Pathogen, No. (%)					0.21d
Gram negative	8 (42)	11 (44)	0.44d	7 (44)
Gram positive	6 (32)	5 (20)		2 (13)
CoNS	4 (21)	9 (36)		3 (19)
Fungal	0 (0)	0 (0)		1 (6)
Polymicrobial	1 (5)	0 (0)		3 (19)
Bell’s Stage III, No. (%)	51 (35)	75 (100)	<0.001d	39 (100)	<0.001d
NEC totalis, No. (%)	0 (0)	2 (3)	0.05d	19 (50)	<0.001d

Abbreviations: NEC, necrotizing enterocolitis; PMA, post menstrual age; CoNS, Coagulase negative staphylococcus.

^aComparison of Survivors: Medical NEC and Survivors: Surgical NEC.

^bComparison of Survivors: Medical NEC and Non-Survivors.

^cMann-Whitney test.

^dChi-squared test.

^eNumber divided by total known. Placental histology was unavailable in 23 cases.

NEC and/or died were of a significantly lower gestational age, smaller birth weight, a younger postmenstrual age (PMA), and more likely to have a positive blood culture at NEC diagnosis (P<0.001) than those who survived medical NEC (Table 1). No differences in sex, Gram stain class of pathogen isolated, or chronological age at the time of NEC were seen between the cohorts (Table 1).

The highest NEC incidence in our cohort occurred at 31.2 weeks PMA (eFigure 1, Supplement). Of the 108 infants who had surgical intervention, 33 died (31%). Six infants (2.3%) died without surgical intervention. Median time to death for non-survivors was 33 hours (interquartile range (IQR), 18–61.5) (eFigure 2) with a median nSOFA score of 11 at the time of death (IQR, 8–13) (data not shown).

nSOFA score utility for NEC mortality

Non-survivor nSOFA scores were significantly elevated relative to survivors (P<0.001) at each time point from the time of evaluation (0 hours) to + 48 hours (Figure 1A). There were no significant differences in the nSOFA for survivors vs. non-survivors prior to the evaluation for NEC. The AUC for the nSOFA to discriminate for mortality at evaluation was 0.75 (95% CI, 0.66–0.84), at +6 hours after evaluation was 0.83 (95% CI, 0.76–0.91), and at +12 hours was 0.87 (95% CI, 0.80–0.93; Figure 1B). Heat maps illustrate patient-specific kinetics of organ dysfunction through nSOFA scores for survivors and non-survivors (Figure 1C). Mortality associated with a maximum nSOFA score (nSOFA^max) ≤1 for measures at −6 hours, 0 hours, +6 hours, and +12 hours (n=128) was 1.5% versus 38% with a nSOFA^max of ≥4 (n=90), and 57% with a nSOFA^max ≥10 (n=30). A nSOFA^max of ≥4 during the −6 hour to +12 hour interval was associated with a 20-fold increase in mortality (OR, 19.9; 95% CI, 7.6–48.3, P<0.001) compared to a nSOFA^max of <4.

Figure 1. nSOFA scores among preterm NEC survivors compared to non-survivors.

(A) Median, quartiles, and probability density of nSOFA scores for survivors (n=220, white) and non-survivors (n=39, blue) are shown. Comparisons by Kruskal-Wallis with Dunn’s multiple comparisons test. ***P<0.001 (B) Receiver operating curve (ROC) for mortality based on nSOFA scores at 0, +6, and +12 hours relative to evaluation (Evaluation: Black circles; +6 hours: blue squares; +12 hours: orange triangles). All ROCs with P<0.001. (C) Heat map of nSOFA scores for individual survivors and non-survivors. The nSOFA score is displayed at each of the nine time points relative to sepsis evaluation. Data organized by nSOFA score compared to the time from NEC evaluation (T−48 to T+48). Blue indicates a higher nSOFA score. Black indicates a time point after death.

Comparison of the change in the nSOFA scores over time revealed that the largest change in nSOFA scores for non-survivors occurred in the peri-evaluation period in contrast to a relative absence of change in the scores of survivors (Figure 2A–C). There was a significant difference in the change in nSOFA scores between survivors (n=220) and non-survivors (n=39) at several intervals: pre-evaluation from T-48, T-24 and T-6 hours to 0 hours (Figure 2A), post-evaluation from 0 to +6 hours (Figure 2B), and peri-evaluation from −6 to +6, −12 to +12, −24 to +24 and −48 to +48 hours (Figure 2C). In addition, the median change in nSOFA scores among survivors from −6 to +12 hours was 0 (IQR, 0–1) compared to 5 (IQR, 3–10; P<0.001) for non-survivors (data not shown). An increase in nSOFA score of ≥4 for this interval was associated with an odds ratio of 17.8 for mortality (95% CI, 7.7–39.2, P<0.001).

Figure 2. Change in nSOFA scores during a NEC episode for survivors and non-survivors.

The change in nSOFA scores between the specified time intervals are shown for survivors (n=220, white) compared to non-survivors (n=39, blue). Median, quartiles, and probability density of are shown for the (A) pre-evaluation, (B) post-evaluation, and (C) peri-evaluation intervals. Comparisons by Kruskal-Wallis with Dunn’s multiple comparisons test. ***P<0.001, ****P<0.0001.

The individual components of the nSOFA score were subsequently compared. Organ-specific dysfunction in non-survivors compared to survivors manifested in the respiratory component at 0 hours (P< 0.0001, Figure 3A), followed by the cardiovascular and hematologic components at + 6 hrs (P<0.05, Figure 3B, C). Examination of four peri-evaluation time points, −6, 0, +6, and +12 hours revealed that non-survivors were more likely than survivors to have any increase in respiratory (OR, 10.1; 95% CI, 4.7–21.8), cardiovascular (OR, 67.3; 95% CI, 25.5–160.7), and hematologic (OR, 24.9; 95% CI, 10.8–56.4) nSOFA component scores (P<0.001).

Figure 3. nSOFA component scores among NEC survivors and non-survivors.

(A) Respiratory, (B) Cardiovascular, and (C) Hematologic components of the nSOFA score compared to the time from NEC evaluation (T−48 to T+48). Median, quartiles, and probability density of component nSOFA for survivors (n=220, white) and non-survivors (n=39, blue) are shown. Comparisons by Kruskal-Wallis with Dunn’s multiple comparisons test. *P< 0.05, ***P<0.001, ****P<0.0001.

The nSOFA has been validated for the calculation of the risk of death in preterm infants with late-onset sepsis [10]. We sought to determine if differences in the nSOFA that we observed between survivors and non-survivors were impacted by bacteremia. Among infants with NEC and a positive blood culture, survivors had a median nSOFA score of 0 at evaluation (IQR, 0–4) compared to 6 (IQR, 3–8.75) for non-survivors (P<0.001), and significant differences in the nSOFA between these groups were present at 0, +6, and +12 hours (P<0.01, Figure 4A). Infants with NEC and negative blood cultures who did not survive also had elevated nSOFA scores relative to survivors at all time points between 0 and + 48 hours (P<0.001, Figure 4C). The AUC for the nSOFA to discriminate for mortality was ≥0.8 at +6 and +12 hours (P<0.001, Figure 4B, D) regardless of blood culture results.

Figure 4. nSOFA scores among preterm survivors compared to non-survivors with NEC with or without a positive blood culture.

Median, quartiles, and probability density of nSOFA scores for survivors (white) and non-survivors (blue) with (A) positive blood culture (n=44, 16) and (C) negative blood cultures (n=177, 22). Comparisons by Kruskal-Wallis with Dunn’s multiple comparisons test. **P<0.01, ***P<0.001, ****P<0.0001. ROC for mortality based on nSOFA scores at 0, +6, and +12 hours relative to evaluation with (B) positive blood culture and (D) negative blood cultures. All ROCs with P<0.05.

Sex-restricted analyses showed nSOFA scores for non-survivor males were elevated starting at −24 hours and for non-survivor females starting at 0 hours (P<0.05, eFigure 3A, C). The AUC for the nSOFA score was ≥0.8 at +6 and +12 hours for both males and females (P<0.0001, eFigure 3B, D).

When infants were stratified based on birth weight, we found that the nSOFA score was significantly higher for non-survivors starting at 0 hours for both the extremely low birth weight (<1 kg, ELBW) and non-ELBW cohorts (P<0.01, eFigure 4A,C). The discriminatory capacity of the nSOFA improved from 0 hours to +12 hours with an AUC of 0.80 (95% CI, 0.71–0.90) for ELBW infants and an AUC of 0.96 (0.92–1) for non-ELBW infants at +12 hours (eFigure 4).

Site-specific analyses showed the AUC for mortality was >0.8 at +6 and +12 hours among institutions with 10 or more non-survivors during the study period (eTable 2).

nSOFA score utility for mortality or surgical intervention

Non-survivors and surgical NEC survivors had elevated nSOFA scores compared to medical NEC survivors at time points between −12 hours to + 48 hours (P<0.05, Figure 5A). The AUC for the nSOFA to discriminate for the combined outcome of death or surgical intervention at evaluation was 0.72 (95% CI, 0.65–0.78), +6 hours was 0.78 (95% CI, 0.72–0.84), and +12 hours was 0.84 (95% CI, 0.79–0.90; Figure 5B). A nSOFA score of ≥4 at the time of initial evaluation was associated with an increased risk of death or surgical intervention compared to scores <4 (OR, 7.0; 95% CI, 3.5–13.6). In addition, a nSOFA^max of ≥4 at the time points of −6, 0, +6, or +12 hours, compared to a nSOFA^max of <4 was highly associated with surgical intervention or death (OR, 18.7; 95% CI, 9.4–36.9, P<0.001) (Table 2).

Figure 5. nSOFA scores among preterm medical NEC survivors compared to infants that died or had surgical intervention.

(A) Median, quartiles, and probability density of nSOFA scores for medical NEC survivors (n=145, white) and infants with non-survivors and surgical NEC survivors (n=114, blue) are shown. Comparisons by Kruskal-Wallis with Dunn’s multiple comparisons test. *P<0.05, **P<0.01, ****P<0.0001. (B) ROC for the combined need for surgery and mortality based on nSOFA score at 0, +6, and +12 hours relative to sepsis evaluation. All ROCs with P<0.001.

Table 2.

Statistical characteristics of maximum nSOFA scores obtained between −6 and +12 hours for medical NEC survivors compared to infants that required surgery and/or died.

	Medical NEC	Surgery/Death	OR (95% CI)	Sensitivity	Specificity	PPV	NPV	LR
Maximum nSOFA	N=145	N=115
≥1	50	100	13.6 (7.1–25.8)	0.88	0.66	0.67	0.87	2.5
≥2	36	95	15.1 (8.0–27.6)	0.83	0.75	0.73	0.85	3.4
≥3	23	80	12.5 (6.8–22.9)	0.70	0.84	0.78	0.78	4.4
≥4	14	76	18.7 (9.4–36.9)	0.67	0.90	0.84	0.78	6.9
≥5	11	68	18.0 (8.9–37.4)	0.60	0.92	0.86	0.74	7.9
≥6	9	64	19.3 (8.9–41.8)	0.56	0.94	0.88	0.73	9.0
≥7	6	55	21.6 (8.9–49.3)	0.48	0.96	0.90	0.70	11.7
≥8	6	53	20.1 (8.3–46.0)	0.46	0.96	0.90	0.70	11.2
≥9	5	35	12.4 (4.8–30.0)	0.31	0.97	0.88	0.64	8.9
≥10	3	26	14.0 (4.3–44.7)	0.23	0.98	0.90	0.62	11.0
≥11	1	17	25.2 (4.3–267)	0.15	0.99	0.94	0.60	21.6

Abbreviations: NEC, necrotizing enterocolitis; OR: Odds ratio; PPV: Positive Predictive Value; NPV: Negative Predictive Value; LR: Likelihood ratio

Similar to the comparison of survivors and non-survivors (Figure 2), there were significant differences in the change in the nSOFA for medical NEC survivors compared to neonates who died or required surgery (eFigure 5). For medical NEC survivors, the median change in nSOFA scores between every time point examined was 0. Whereas, for infants that died or required surgery, the change in the nSOFA was significantly higher, particularly in the peri-evaluation period (eFigure 5C). For example, the median change in the nSOFA for infants that died or had surgical intervention was 2.5 between −6 and +12 hours (IQR, 0–6; P<0.0001) and was 4 between −48 and +48 hours (IQR, 1–7; P<0.0001) (eFigure 5C).

Discussion

This is the largest study to date of preterm infants with NEC for whom the kinetics of organ dysfunction has been serially quantified using a reproducible, objective, validated tool. Although the strength of our conclusions is limited by the retrospective nature of this study, we found that in a cohort of infants with NEC, the nSOFA effectively identified those that were at the highest risk for death or the combined outcome of death or surgical intervention.

The goal of implementing the nSOFA score for patients with NEC is to rapidly identify the highest risk patients within a window where a possible effective intervention could occur. We analyzed the AUC for the ROC curves for nSOFA scores during potentially actionable windows, i.e., at initial evaluation, + 6, or +12 hours, and found the highest discriminatory capacity at 12 hours after initial evaluation. The AUC was 0.87 for mortality (Figure 1) and 0.84 for surgery or death (Figure 5), and this good discriminatory capacity (AUC ≥ 0.78) was consistent for sub-analyses based on gender (eFigure 3), birth weight (eFigure 4), and academic center (eTable 2). In addition, the calculation of nSOFA^max between −6 to +12 hours can serve as a powerful dynamic tool to determine if an infant is at high risk for poor outcomes (Table 2). If an infant has an elevated nSOFA score at any point during that window, the likelihood of a poor outcome can be estimated based on our data. At +12 hours, 85% of infants who ultimately did not survive were still alive (eFigure 2). Thus, in the future, the nSOFA may allow clinicians to objectively identify the most high-risk infants with NEC during a time frame when enrollment into clinical trials that may alter illness progression and improve outcomes can be facilitated. Changes in clinical status in response to these novel treatments could also be quantified using the nSOFA, which will allow for higher quality and more reproducible findings. Incorporation of the nSOFA into clinical trials and subsequently into day-to-day practice will allow for administration of therapies to the precise population in which they had been previously studied.

The nSOFA was developed and validated as a tool to determine the risk of mortality in preterm neonates with late-onset sepsis [9, 10] as well as to determine the risk of death within the first 28 days from any cause among infants admitted to the NICU [11]. Our data indicate that the nSOFA was able to effectively capture the risk of mortality or the combined outcome of death or surgery among neonates with NEC, regardless of whether NEC was associated with bacteremia (Figure 4).

An improved understanding of the evolution of organ dysfunction in patients with NEC is critical to the design and implementation of successful therapeutic interventions. Despite studies providing greater comprehension of epidemiologic, preventative, and pathologic features of NEC, patient classification, including disease trajectory, based on measures of organ dysfunction has not been successfully integrated into practice. Modified Bell’s staging neither includes nor compares dynamic information related to outcomes [13, 2]. Other neonatal illness severity scores, including Score for Neonatal Acute Physiology with Perinatal Extension-II (SNAPPE-II), have not demonstrated the ability to predict surgical intervention or mortality in NEC and were not validated for longitudinal measures [14]. Heart Rate Characteristics (HRC) index shows promise as an early predictor of disease [15]. However, the difference in the HRC index between medical NEC (1.9 ± 1.7) and surgical NEC (3.3 ± 2.2), as well as the requirement of specialized equipment for HRC monitoring, may limit generalizability [15]. In this study, we found that the respiratory, cardiovascular, and hematologic components of the nSOFA were all significantly elevated in infants who died relative to survivors (Figure 3). Chronologically, the respiratory score was the first component that increased in non-survivors. It was significantly higher in non-survivors (P<0.0001) as early at 0 hours when the median score was 2 (IQR 0–6) for non-survivors and 0 (IQR 0–0) for survivors and remained significantly elevated for the duration of the study for infants who died (Figure 3). Cardiovascular compromise was detected next, with a significant elevation for non-survivors at +6 hours through + 48 hours (P<0.0001). Thrombocytopenia then developed with a small difference between survivors and non-survivors at + 6 hours (P <0.05) but a more significant difference developing at +12 and +24 hours (P<0.001). Thrombocytopenia has been previously associated with the development of bowel gangrene and increased mortality in infants with NEC [16]. Based on this evolution of the nSOFA components, a clinical picture emerges where clinical illness is first recognized in the setting of respiratory decompensation. This is followed by cardiovascular compromise requiring vasopressor support, and then thrombocytopenia results as bowel injury and systemic inflammation evolve.

Automation of the nSOFA score for infants with NEC would be an optimal scenario in clinical research as it would serve as the most rapid and streamlined approach for the identification of evolving organ dysfunction and patient classification. Integration of the nSOFA into the electronic health record (EHR) has already been successfully performed at multiple academic medical centers [9, 17]. Utilizing an automated scoring system, the EHR at each center incorporates laboratory and vital sign data into a nSOFA score every 15 minutes and alerts clinicians if a nSOFA score is elevated. This allows for detailed monitoring with analysis of trends in the nSOFA score for a particular infant. In this study, we found significant differences in the change in nSOFA for infants with NEC who had poor outcomes when examining time frames as short as 6 hours, such as between 0 and +6 hrs (Figure 2 and eFigure 5). Future prospective studies of infants with NEC in centers with the nSOFA incorporated into the EHR will need to be conducted to determine the necessary frequency of nSOFA calculation to allow for optimal risk stratification. In centers unable to integrate the nSOFA into an EHR, our data indicate that calculation of the score should begin with clinical suspicion and continue at least every 6 hours through 48 hours after initial evaluation. If a definitive diagnosis of NEC is made, rapid identification of an elevated nSOFA score or a significant increase in an infant’s score will allow for re-evaluation of the infant’s expected clinical trajectory and risk of death and/or surgery. In the future, this could be followed by enrollment in a clinical trial or provision of a targeted therapeutic. At present, we do not suggest the nSOFA should be used as a surgical decision-making tool, because the decision of whether an infant with NEC needs surgical intervention is complex and incorporates many clinical factors. Thus, the highest utility of the nSOFA for NEC is the identification of infants at highest risk for progressing to surgery or death under the current standard of care and, thus, those that may benefit from novel therapeutic interventions.

This study has limitations inherent to any retrospective analysis. Our goal was to measure the utility of the nSOFA score to determine which preterm infants with NEC were at the highest risk of death or the combined outcome of death or surgical intervention. The generalizability of our results is impacted by inevitable institution-specific differences, such as in standards of care, demographics, laboratory monitoring, and clinical thresholds for surgical intervention. Site-specific analyses demonstrated generalizability of the nSOFA score for NEC mortality at centers with more than five deaths from NEC during the study period. In addition, disease timing (PMA), mortality rate, bacteremia, and surgical intervention rate in our cohort were similar to other large cohorts of patients with NEC [18–20]. Fortunately, the nSOFA score is easily calculated and can be validated prospectively within a larger group of NICUs. In addition, it is likely that clinical trials using novel therapies for NEC will first be conducted at larger academic centers, which are more likely to have similar patient populations and outcomes to the centers included in this study. Another limitation to this study is that a diagnosis of NEC was not required at the time of the reference blood culture (0 hours) for inclusion into the study. A delay in arriving at a diagnosis of definitive NEC may reduce the window when calculating a nSOFA score may be beneficial given the rapid progression to death that we observed for some patients (eFigure 2). We suspect that with larger prospective trials examining the nSOFA score in infants with NEC and integration of scoring into the EHR, we will be able to identify these patients earlier in their disease course. These prospective studies will also inform whether an increasing nSOFA should prompt more frequent abdominal imaging or potentially measurement of novel biomarkers for NEC. The validity of the nSOFA score as a predictive tool will be determined in large-scale, prospective studies; however, our data reflect the potential of this scoring system to serve as a powerful adjunct in both research and clinical care.

Conclusion

In this study of preterm infants with definitive NEC, an elevated nSOFA score was associated with an increased risk of death or the combined outcome of death or surgery. This study represents an important step towards the goal of integration of a serially measured organ dysfunction metric to facilitate precise patient classification and risk stratification for NEC, which would be a significant advance for the field of neonatology.

Data Availability Statement

All data generated or analyzed during this study are included in this article [and/or] its supplementary material files. Further enquiries can be directed to the corresponding author.