Necrotizing enterocolitis (NEC) is a serious gastrointestinal disease in the neonatal intensive care units (NICUs) and is a major contributor to mortality in premature infants. Despite continuous improvements in medical technology, the morbidity and mortality rates are still as high as 7–13% and 10–40% (1, 2), and the survivors are likely to have long-term complications, such as short bowel syndrome, intestinal strictures or adhesions, cholestasis, and neurodevelopmental delays (3, 4). Moreover, stage II/III NEC tend to be more likely to have complications and have higher mortality and disability rates. Thus, it is crucial to establish a risk stratification assessment tool for NEC.

The etiology of NEC is thought to be multifactorial. However, the role of a single factor with NEC needs to be further elucidated. Studies have shown that NEC mainly occurs in 90–95% of infants <36 weeks of gestational age (GA) (5). A multi-center study of epidemiological data showed that the incidence of NEC is negatively correlated with birth weight (BW). The incidence of NEC is approximately 2.50% in low-birth-weight infants (<2,500 g), and the incidence is approximately 4.53% in very-low-birth-weight infants (<1,500 g) (6). In addition to the aforementioned GA and BW being recognized risk factors, some of the previous studies have also identified other independent risk factors, such as septicemia, patent ductus arteriosus (PDA), erythrocyte infusions, and formula feeding (7, 8) for the occurrence of NEC. However, traditional statistical methods can only determine the risk factors for the occurrence of the outcomes. These methods cannot predict the probability of the occurrence of NEC. This article analyzed the clinical data of NEC patients to find out the risk factors for NEC. Then, a prediction model was established based on these findings. Finally, an internal validation was performed based on this model by using the data in the verification group so that the model can be better used to predict the probability of NEC in other individuals in the future. The previous research on the establishment of prediction models for NEC have mainly focused on the establishment of a surgical risk prediction model and a death risk factor scoring system model for NEC (9, 10), but a risk stratification assessment model for NEC has not yet been established.

The nomogram is one of the most representative models of all the prediction models, and the nomogram is more accurate and has better performance characteristics than artificial neural networks (ANNs), classification and regression tree models (CARTs), look-up tables, and risk-group stratification models (11). For example, the nomogram Zhang et al. (12) developed can nicely and relatively accurately predict the probability of Bronchopulmonary Dysplasia (BPD) in infants with a BW <1,500 g, reducing the chances of exposure to risk factors could decrease the incidence of BPD. Therefore, the nomogram is really a simple and useful model to predict the risk of clinical disease. As a result, developing a predictive nomogram model of NEC will be promising for distinguishing Bell's stage II/III NEC early from Bell's Stage I NEC, and taking early specific interventions to reduce the probability of adverse outcomes.

We retrospectively analyzed infants with NEC hospitalized in Anhui Provincial Children's Hospital from January 2015 to January 2021. In total, 730 patients were included; a random sample of 530 newborns was evaluated to construct the nomogram, and the remaining 200 neonates were evaluated for the validation of the model. The flowchart of this study is illustrated in Figure 1. The diagnosis of NEC was based on the Bell and Walsh grading criteria, which includes Stage I, Stage II, and Stage III NEC (13). Infants with stages I and II were defined as suspected and defined NEC, respectively. And those with stage III were defined as having advanced NEC. In this study, stages ≥II NEC were considered a positive outcome. Patients with serious systemic diseases were excluded, such as immune deficiency disease, genetic metabolic disease, pulmonary hemorrhage, congenital intestinal malformation, congenital diaphragmatic hernia, malrotation of the intestine, meconium intestinal obstruction, and simple intestinal perforation. Patients with incomplete medical records were also excluded. This study was approved by the ethics review board of Anhui Provincial Children's Hospital (Nr: EYLL-2017-023).

In this retrospective cohort study, we calculated and validated a personalized prediction nomogram model that predicted the risk of Bell's stage II/III NEC, which is useful for clinicians when making treatment plans for infants with NEC. Finally, we identified six associated risk factors: GA, BW, PDA, septicemia, asphyxia, and hypoglycemia. There is no doubt that GA and BW should be included in the model, and both have been shown to be strongly associated with NEC in many studies [1, 2, 5, 6, and (21)]. Our research results also confirmed this idea. The possible factors involved in the development of NEC in infants with a lower GA are the immaturity of the gastrointestinal tract, immune defense, circulation regulation, digestive function, and barrier function (22).

Septicemia is recognized as a risk predictor for NEC. When sepsis occurs, a large number of bacteria can multiply in the blood, produce toxins, and cause damage in the intestine. However, the newborn's intestine is immature. Hence, exposure to the above factors can lead to the release of a large number of cytokines, causing inflammation in the body, which promotes necrosis of the intestinal mucosa, and damages the intestinal barrier function (23). It has been reported that the incidence of NEC in newborns with septicemia ranges from 34 to 57% (24–26). All patients suspected of sepsis should receive antibiotics basing guidelines, but long-term exposure to antibiotics may increase the occurrence of NEC (27), especially third-generation cephalosporins (28). Western countries prefer to use ampicillin and aminoglycoside antibiotics, so the use of antibiotics in neonates with sepsis requires reasonable choice.

In mice models, NEC can be induced by intermittent hypoxia and giving formula milk feeding (29). Previous clinical studies have shown that hypoxia can lead to NEC in infants (26). Our data support the results of studies that have described an association between NEC and asphyxia. This association may be due to the imbalance between oxidation and anti-oxidation in the body under low oxygen. Aydemir et al. (30) found that the levels of total antioxidant capacity and oxidative stress index correlated with the severity of NEC were significantly elevated. Therefore, newborns with asphyxia after birth need to need to be monitored for abdominal signs and treated in a timely manner to reduce the gastrointestinal burden, if necessary.

In our study, PDA was correlated with NEC, which was similar to the results of previous studies (7). This could be due to a reduced intestinal blood flow during catheter shunting. Prophylactic surgical ligation of the PDA was associated with significant reduction in the incidence of stage II or III NEC (31), but it is an invasive operation. Currently, people are paying increasing attention to medical therapy for PDA cases. A Cochrane systematic review concluded that ibuprofen and indomethacin were helpful for PDA closure, and could reduce the risk of NEC and transient renal insufficiency (32). However, there is currently no consensus on whether PDA in preterm infants requires treatment, how, and when these infants should be treated.

Hypoglycemia was also found to be a predictive factor in this study, and hypoglycemia is common in neonates. It may be caused by a variety of diseases, such as sepsis, hypermetabolic diseases, hyperinsulinemia, fatty acid oxidative disorders, gluconeogenic disorders, and glycogen deficiency (33). Studies have shown that hypoglycemia secondary to hypopituitarism increased the risk for NEC (34). In addition, 46 of the 118 patients with hypoglycemia in the training set also had sepsis in this study. However, there are few studies on the relationship between hypoglycemia and NEC, and further research is needed

In recent years, some researchers have linked erythrocyte infusions with the development of NEC in infants and have hypothesized that erythrocyte infusion could be a cause of NEC (35). The possible mechanisms of NEC induced by blood transfusion are as follows: (1) Blood transfusions cause hemodynamic instability and uneven distribution of intestinal blood supply, leading to ischemia, and necrosis of immature intestinal mucosa. (2) It may be similar to transfusion-related acute lung injury (TRALL), the host by primary disease induces vascular endothelial activation, chemokine release, and neutrophil activation, when biological immune response factors, such as Human Leukocyte Antigen (HLA) antibodies, bioactive substances, free hemoglobin, red blood cell debris, and inflammatory cytokines present in the blood promote a cascade reaction between neutrophils and the complement system causing secondary damage (36). (3) Reduced NO in stored blood leads to vasoconstriction and reduced intestinal blood flow and oxygenation (37). However, some studies have shown that there is no association between red blood cell transfusions and NEC (38), and some reports have even suggested that these transfusions could decrease the risk of NEC (39). Therefore, the relationship between erythrocyte infusions and the occurrence of NEC is not very clear. In our study, transfusion did not increase the risk of NEC. Moreover, formula feeding in this study was not included in the model as a risk factor for NEC, which could be related to the low breastfeeding rate in our hospital and the limitations of the sample size utilized. In the past 2 years, a milk donation bank was established in our hospital, and feeding method including breast milk, donor milk, and formula will be further evaluated in prospective studies.

A nomogram is a graph drawn by turning the regression coefficients into scores after appropriate mathematical operations, and its advantages are strongly visualized. And the degree of correlation between each variable and outcome could be directly reflected by the score. Clinicians could early identify stage II/III NEC early through this nomogram and calculate the probability of occurrence. If the probability is very high, the doctor needs to take appropriate treatment measures in time. In severe cases, it is even necessary to recommend surgery for infants, which helps avoid missing the optimal treatment time. But the nomogram looks complicated at first glance, and you may need to learn how to use it.

Several limitations exist in this study. Our clinical data originated from a single hospital, and the presence of selection bias cannot be ignored. Moreover, it may also be difficult to determine whether the study applies to patients in other countries, considering the racial differences (40), and further external verifications are needed. In future research, we plan to develop other clinical prediction models for NEC, such as models based on CART and ANNs. By comparing the advantages and disadvantages, we will select the best model for NEC prediction. Then, we will carry out further multicenter, prospective studies to validate the optimal model for generalization.

We have developed and internally validated a novel nomogram model for predicting the risk of Bell's stage II/III NEC compared to Bell's stage I NEC. The nomogram with the advantages of being easy to use and having an excellent calibration might help clinicians to make early individualized predictions of each patient's probability of Bell's stage II/IIINEC early and to take early therapeutic measures, thus reducing the severity of NEC. To our knowledge, we have established the first risk assessment tool for NEC.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

The studies involving human participants were reviewed and approved by Anhui Provincial Children's Hospital. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

LD and SS conceived and designed the study. SS collected and sorted out the data and was responsible for drafting the manuscript. JZ guided the data statistical analysis. JZ and YZ revised the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.