An Early Prediction of Neonatal Necrotizing Enterocolitis in High-Risk Newborns- Based on Two Medical Center Clinical Databases

An Early Prediction of Neonatal Necrotizing Enterocolitis in High-Risk Newborns- Based on Two Medical Center Clinical Databases | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article An Early Prediction of Neonatal Necrotizing Enterocolitis in High-Risk Newborns- Based on Two Medical Center Clinical Databases Yanling Mou, Jinhao Li, Jianjun Wang, Daiyue Yu, Huirong Yang, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4556691/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background : To improve the prognosis of necrotizing enterocolitis (NEC) in newborns, early identification and timely preventive interventions play an essential role. Based on the current situation, establishing a novel and simple prediction model is of great clinical significance. Methods : The clinical data of NEC neonates in Zhujiang Hospital of Southern Medical University from October 2010 to October 2022 were collected, and 429 non-NEC patients in the same period were selected as the control group by random sampling method. After that, all participants were randomly divided into training group (70%) and testing group (30%). Combining relevant clinical features and laboratory results, five machine learning (ML) algorithms and classical logistic regression models were established. To evaluate the performance of each model, the area under the receiver operating characteristic curve (ROC), accuracy, sensitivity, and specificity of various models were compared. 10-folds cross-validation was used to find the best hyperparameters for each model. Decision curve analysis (DCA) was further used to evaluate the performance of the established models for clinical applications, and create a column-line graph, ranking the feature importance in model by SHapely Additive exPlanation (SHAP). The column plots were calibrated using calibration curves. In addition, the established model was validated in time series analysis as well as in another medical center. Results : Six important features were finally included for modeling, including the Day (OR=1.15; 95% CI: 1.07-1.23; P =0.001), Gestational age (OR=0.77; 95% CI: 0.62-0.95; P =0.016), Eosinophil (EOS) (OR=3.76; 95% CI: 1.76-8.02; P =0.001), Hemoglobin (HB) (OR=0.98; 95% CI: 0.97-1.00; P =0.011), Platelet distribution width (PDW) (OR=1.21; 95% CI: 1.08-1.35; P =0.001) and High-sensitivity C-reactive protein (HSCRP) (OR=1.03; 95% CI: 1.01-1.06; P =0.007). While the logistic regression model achieved an AUC of 0.919, accuracy of 0.897, sensitivity of 0.832, F1-score of 0.778, and a Brier score of 0.0878 in the training group, the AUCs for the five machine learning models ranged from 0.774 to 0.972. Among these models, the LightGBM model performed the best, with an AUC of 0.960, accuracy of 0.894, sensitivity of 0.901, F1-score of 0.813, and a Brier score of 0.072. Conclusion : The LightGBM machine learning model can effectively identify neonatal patients at higher risk of NEC based on Day age, Gestational age, EOS, HB, PDW, and HSCRP levels. This model is useful for assisting in clinical decision-making. machine learning necrotizing enterocolitis diagnose prediction Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 What is Known The impact of necrotizing enterocolitis (NEC) in newborns has garnered ample attention from pediatricians, yet inexperience and the absence of advanced and sensitive diagnostic tools often result in a failure to timely control the disease. A variety of models have been developed to identify children with the modified Bell staging criteria and staged as II and III. What is New : Machine learning models that are trained on big data can assist in identifying patients with NEC at an earlier stage. Even though early NEC poses a risk of misdiagnosis, no child with a potential for a serious adverse prognosis should be ignored. Introduction Nowadays, necrotizing enterocolitis (NEC) is one of the primary causes of mortality and morbidity in neonatal intensive care units[ 1 ]. The overall incidence of NEC is approximately 1–5 per 100 live births[ 2 ], with a particularly high occurrence in preterm infants[ 3 ]. The clinical presentation of NEC is complex, such as abdominal distension, vomiting, and fecal occult blood. However, in many cases, the symptoms are atypical, and the condition progress rapidly. We primarily relied on doctors' clinical experience to make a diagnosis formerly. However, it is difficult for doctors in primary hospitals who were the main management of neonate in China to recognize NEC at an early stage. With the ever-increasing availability of novel markers[ 4 , 5 , 6 , 7 , 8 ], clinicians remain relentlessly on the hunt for more sensitive technological advancements, albeit diminishing their emphasis on fundamental patient data. Therefore, it is undoubtedly significant to develop a visual clinical prediction models and offer insights for early clinical diagnosis through model scoring. The modified Bell staging[ 9 ] is still the most commonly used clinical staging method currently. In addition, platelet count[ 10 ], lymphocyte count[ 11 ], and coagulation function indexes[ 12 ] were also considered as early indicators of NEC. Machine learning is a process of automatic learning from computerized data that enables computers to discern laws and patterns from data and make predictions and decisions based on them[ 13 ]. It is important to analyze and process the available data on a large scale using machine learning in order to develop a new application model for the early and accurate identification of NEC. Our study proposes the construction of an early prediction model for NEC using machine learning, aiming to validate the effectiveness of the prediction model and constructing a visualization model to enable early and accurate identification of NEC patients. This will provide a new approach for the early diagnosis of neonates, particularly preterm infants with NEC. Population and Methods Research population Neonates at Zhujiang hospital of Southern Medical University from October 2010 to October 2022 were selected. The inclusion criteria for the NEC group were as follows: (1) Diagnosis by two clinically experienced pediatricians in accordance with the modified Bell staging criteria[ 9 ] and staged as IA, IB, IIA, or IIB (which includes clinical data from the early stages of disease for the patients diagnosed as stage III according to criteria); (2) Newborns under 28 days old. The exclusion criteria: (1) Individuals with genetic metabolic disorders or congenital malformations; (2) Individuals with tumor diseases; (3) Use of drugs that may cause changes in blood routine; (4) Any coagulation abnormalities, such as intracranial hemorrhage, gastrointestinal hemorrhage, pulmonary hemorrhage, sepsis, or septicemia. (5) Lack of clinical or laboratory data. Non-NEC neonates who were admitted at the same time were randomly selected in proportion to serve as the control group. According to the classification management method for high-risk newborns[ 14 ], if one of the following inclusion criteria exists, we consider the infant to be in a high-risk state: (1) Premature infants. (2) Low birth weight infants. (3) Small for gestational age infants. (4) Hyperbilirubinemia. (5) Water electrolyte and metabolic disorders. (6) Retinopathy of premature infants. (7) Other conditions such as hypothermia, apnea, omphalitis, head hematoma, and no other accompanying conditions exclusion criteria are the same as NEC group. Study Indicators The clinical characteristics of the patients included gender, Day age, Gestational age, Birthweight, Maternal age, Apgar score at 1 minute, mode of delivery, and the presence of patent ductus arteriosus[ 15 ]. Birthweight was averaged from three measurements by two trained nurses, and arterial duct failure was diagnosed by a cardiac ultrasound expert. The control group underwent the first routine blood test after admission, while the latest blood routine results of NEC group before diagnosis as the disease group. The collected information included the white blood cell count (WBC), absolute neutrophil value (NEU), absolute lymphocyte value (LYM), absolute monocyte value (MONO), absolute eosinophil value (EOS), absolute basophil value (BASO), red blood cell count (RBC), hemoglobin (HB), platelet count (PLT), mean platelet volume (MPV), platelet pressure product (PCT), platelet volume distribution width (PDW), and High-sensitivity C-reactive protein (HSCRP) levels. Analytical Methods Missing data were filled using the R package "mice", employing multiple imputation with the number of imputations set to 5. Subsequently, the dataset was randomly divided into training and test sets at a 7:3 ratio, and stratified sampling was performed using the R package "caret". Logistic one-way analysis of variance (LASSO) was conducted on the training set using the R package "stats". Subsequently, the metrics that showed statistical significance ( P < 0.05) were subjected to further analysis using LASSO with the R package "glmnet". Support Vector Machine (SVM) analysis was conducted using the R package "e1071" to determine the optimal number of features, based on the observed decrease in the mean square error plot. The corresponding number of features were then included in order of importance. Additionally, Random Forest (RF) analysis was carried out using the R package "RandomForest", with a criterion of MeanDecreaseAccuracy > 10. Subsequently, LASSO analyses were conducted using the Draw Venn Diagram website ( https://bioinformatics.psb.ugent.be/ ) to identify the intersection of the important variables obtained from screening the three models: LASSO, SVM, and RF. Further logistic multifactor analysis was performed on the intersection features ( P < 0.05). The most significant features in the intersection set were utilized to build four SVM models with linear, polynomial, radial, and sigmoid using the R package "e1071". Additionally, a K-nearest-neighbor model was constructed using the R package "caret", which selects the optimal K-value through grid search. Furthermore, a Bayes model was built using the R package "e1071", and a LightGBM model was constructed using the R package "LightGBM". Compare the area under the ROC curve, accuracy, sensitivity, and specificity of each model to determine the best model. Finally, the R package "runway" was utilized to construct the calibration curves for the respective models, and the R package "dcurves" was employed to create the clinical decision curves for evaluating their clinical predictive value. (Fig. 1 ) Results Clinical characteristics The study population consisted of 143 NEC patients and the control group comprised a randomized sample of 429 contemporaneous non-NEC neonates. The study population was divided into a training set and a testing set at a 7:3 ratio. One-way logistic regression analysis was performed on the training set, revealing statistically differences between the two groups in terms of day age( P < 0.001), Gestational age ( P < 0.001), Birth weight ( P < 0.001), Apgar score at 1 minute ( P < 0.001) and delivery mode ( P = 0.031). However, there were no significant differences in the mother's age ( P = 0.810), the presence of arterial ductus arteriosus ( P = 0.081) and sex ( P = 0.202). (Table 1 ) Table 1 Baseline and clinical characteristics of the study cohort. Normal (N = 301) NEC (N = 101) OR (univariable) Sex Female 126 (41.9%) 35 (34.7%) Male 175 (58.1%) 66 (65.3%) 1.36 (0.85–2.17, p = 0.202) Day 2.0 ± 4.2 8.1 ± 7.2 1.20 (1.14–1.27, p < 0.001) Gestational age 38.4 ± 2.2 35.9 ± 4.1 0.71 (0.66–0.78, p < 0.001) Birthweight 3.0 ± 0.6 2.3 ± 0.9 0.28 (0.20–0.40, p < 0.001) MomYear 30.7 ± 4.5 30.5 ± 5.7 0.99 (0.95–1.04, p = 0.810) Apgar score at 1 minute 9.4 ± 1.4 8.7 ± 2.0 0.78 (0.68–0.89, p < 0.001) Delivery mode No 183 (60.8%) 49 (48.5%) Cesarean 118 (39.2%) 52 (51.5%) 1.65 (1.05–2.59, p = 0.031) PDA No 228 (75.7%) 85 (84.2%) Yes 73 (24.3%) 16 (15.8%) 0.59 (0.32–1.07, p = 0.081) Laboratory tests There was no significant difference in MONO ( P = 0.351) between the NEC and non-NEC groups. However, two groups exhibited statistically differences in WBC ( P < 0.001), NEU ( P < 0.001), LYM ( P = 0.032), EOS ( P = 0.001), BASO ( P < 0.001), RBC ( P < 0.001), HB ( P < 0.001), PLT ( P < 0.001), MPV ( P < 0.001), PCT ( P < 0.001), PDW ( P < 0.001), and HSCRP ( P < 0.001).(Table 2 ) Table 2 Serum markers of the study cohort. Normal (N = 301) NEC (N = 101) OR (univariable) WBC 14.8 ± 6.1 11.5 ± 6.7 0.90 (0.87–0.95, p < 0.001) NEU 9.6 ± 5.5 6.5 ± 4.6 0.88 (0.83–0.92, p < 0.001) LYM 3.5 ± 1.4 3.1 ± 2.0 0.85 (0.73–0.99, p = 0.032) MONO 1.4 ± 0.7 1.5 ± 1.2 1.13 (0.88–1.45, p = 0.351) EOS 0.3 ± 0.3 0.6 ± 0.9 2.28 (1.39–3.73, p = 0.001) BASO 0.1 ± 0.1 0.1 ± 0.1 0.00 (0.00–0.00, p < 0.001) RBC 4.6 ± 0.7 3.9 ± 0.8 0.26 (0.18–0.37, p < 0.001) HB 161.7 ± 24.0 131.7 ± 27.7 0.96 (0.95–0.97, p < 0.001) PLT 278.9 ± 93.6 227.5 ± 150.2 1.00 (0.99-1.00, p < 0.001) MPV 10.2 ± 1.0 10.9 ± 1.5 1.66 (1.36–2.02, p < 0.001) PCT 0.3 ± 0.1 0.2 ± 0.1 0.00 (0.00-0.05, p < 0.001) PDW 11.6 ± 2.5 15.3 ± 3.7 1.43 (1.32–1.56, p < 0.001) HSCRP 3.2 ± 8.1 28.8 ± 53.1 1.06 (1.04–1.09, p < 0.001) Screening the characteristic variables for constructing the model LASSO mitigated the impact of multicollinearity on regression results by setting the coefficients of related independent variables to 0 based on the correlation between independent variables. A total of 12 most relevant indicators and feature variables were selected. RF was used to build multiple decision trees by randomly selecting samples and features from the dataset and combining their results to make predictions and a total of 13 feature variables were screened. SVM was used to construct a model by maximizing the distance between data points in the space between categories and planes, aiming to find the optimal boundary. To maximize the sum of the distances from the data points of each category in the space to the classification boundary, also known as the support vector, a total of 9 feature variables were selected. The intersection was then taken to identify the important feature variables in all three models, including Day age, Gestational age, Birth weight, LYM, EOS, HB, PDW and HSCRP. (Fig. 2 ) And then, the six feature variables were analyzed using multifactor logistic regression. The results indicated that the intersection of Day age (OR = 1.15; 95% CI: 1.07–1.23; P = 0.001), Gestational Age (OR = 0.77; 95% CI: 0.62–0.95; P = 0.016), EOS (OR = 3.76; 95% CI: 1.76–8.02; P = 0.001), HB (OR = 0.98; 95% CI: 0.97-1.00; P = 0.001), PDW (OR = 1.21; 95% CI: 1.08–1.35; P = 0.001) and HSCRP (OR = 1.03; 95% CI: 1.01–1.05; P = 0.006) were identified as risk factors (Table 3 ) . Table 3 Multivariate analysis of common variables Variables Normal (N = 301) NEC (N = 101) OR (univariable) OR (multivariable) Day 2.0 ± 4.2 8.1 ± 7.2 1.20 (1.14–1.27, p < 0.001) 1.15 (1.07–1.23, P = 0.001) Gestational age 38.4 ± 2.2 35.0 ± 4.1 0.71 (0.65–0.77, p < 0.001) 0.77 (0.62–0.95, P = 0.016) Birthweight 3.0 ± 0.6 2.3 ± 0.9 0.26 (0.18–0.37, p < 0.001) 0.75 (0.31–1.83, P = 0.533) LYM 3.5 ± 1.4 3.1 ± 2.0 0.85 (0.73–0.99, p = 0.032) 0.81 (0.66-1.00, P = 0.053) EOS 0.3 ± 0.3 0.6 ± 0.9 2.28 (1.39–3.73, p = 0.001) 3.76 (1.76–8.02, P = 0.001) HB 161.7 ± 24 131.7 ± 27.7 0.96 (0.95–0.97, p < 0.001) 0.98 (0.97-1.00, P = 0.011) PDW 11.6 ± 2.5 15.3 ± 3.7 1.43 (1.32–1.56, p < 0.001) 1.21 (1.08–1.35, P = 0.001) HSCRP 3.2 ± 8.1 28.8 ± 53.1 1.06 (1.04–1.09, p < 0.001) 1.03 (1.01–1.06, P = 0.007) Machine Model Construction (1) NEC prediction model based on Lasso variable screening LASSO regression was employed to screen and downscale the six variables that showed statistical differences in Table 3 , Day age, Gestational age, EOS, HB, PDW, and HSCRP. As the log (λ) increased, the mean standard error also increased, and the normalization coefficients of the two candidate variables were compressed to varying degrees until they all reached zero. During the construction of the model, we found that the LASSO model possesses the most prominent stability, even though it is a non-optimal model in terms of numerical evaluation. (2) Modeling based on simple Bayesian screening variables The Naive Bayes model assumed that the features were independent, which simplifies the plain Bayes algorithm but can sometimes sacrifice a certain level of classification accuracy. The simple Bayesian model performed poorly in our selection, as evidenced by unstable AUC values in the training set and validation set, and relatively low sensitivity, specificity, and accuracy. (3) Modeling based on screening variables using SVM Based on the six feature variables in Table 2 , four kernel functions were used in the SVM model to build the model. The polynomial, linear, radial, and sigmoid AUCs in the validation set were 0.94, 0.92, 0.97 and 0.81, respectively. The polynomial performed the best with a specificity of 0.94 and 0.91 between the two groups, but the AUC performance is unstable at 0.94 and 0.87. On the other hand, the radial function has stable and the highest AUC performance between the two groups, and we choose it as the best function in SVM to build the model. (4) Modeling based on KNN screening variables K Nearest Neighbors (KNN) is a classification algorithm in supervised learning, which is well-suited for the binary nature of our data. In our study, the optimal K value was selected based on the data characteristics and then cross-validation methods were used to build the KNN model. The KNN generally performs well, yielding better prediction results, but shows significant variation between the two groups. (5) Modeling based on LightGBM variable screening Light Gradient Boosting Machine (LightGBM) is a method for implementing Gradient Boosting Decision Trees. As the model we finally selected after thorough consideration, this model has the best values among the evaluation indicators. In addition, even though we only used the training set data for modeling, we achieved more stable validation results in the validation set with higher AUC, accuracy, and sensitivity. (6) Modeling based on logistic variable screening In order to further evaluate the prediction performance of machine learning models, we built a traditional logistic regression statistical model with six selected variables. Most of the machine learning models have higher prediction efficacy than traditional statistical models, which may be related to the characteristics of machine learning itself. On the other hand, the AUC of the traditional regression model is close to 0.9 in both the training set and the test set, which to some extent also indicates the reliability of these metrics. Model Evaluation The predictive performance of the five models was evaluated using AUC, accuracy, precision, sensitivity specificity F1-Score and Brier Score. A 10-fold cross-validation (CV) was applied to avoid overfitting. The AUC of the Radial of SVM and KNN models was 0.956 and 0.972 in the training set, but in the test set, their AUC values declined significantly to 0.948 and 0.881, respectively. While the predictive performance of LightGBM based on AUC was stable, with AUC values of 0.960 and 0.969, respectively(Table 4 ). Furthermore, LightGBM demonstrates higher accuracy, precision, sensitivity, specificity and F1-Score of 0.935, 0.830, 0.929, 0.938 and 0.876, respectively, compared to the stability of SVM, LASSO, Bayesian, and KNN in the test set. In addition to the sigmoid of SVM, all five machine learning models exhibit better prediction performance than the traditional logistic regression model (Table 5 ).(Fig. 3 ) Table 4 Evaluation results of the model in the training set Model AUC Specificity Sensitivity Accuracy Precision F1 Brier Score Polynomial 0.9332 0.9269 0.8614 0.9104 0.7982 0.8286 0.0963 Linear 0.9165 0.8538 0.8614 0.8557 0.6641 0.7500 0.0871 Radial 0.9560 0.9169 0.8713 0.9055 0.7788 0.8224 0.0631 Sigmoid 0.7742 0.8372 0.6931 0.8010 0.5882 0.6364 0.1442 Lasso 0.9172 0.8870 0.8614 0.8806 0.7190 0.7838 0.0998 Bayes 0.9105 0.8339 0.8614 0.8408 0.6350 0.7311 0.1198 KNN 0.9722 0.8571 1.0000 0.8930 0.7014 0.8245 0.0569 LightGBM 0.9596 0.8937 0.9010 0.8955 0.7398 0.8125 0.0715 Logistic 0.9190 0.8970 0.8317 0.8806 0.8806 0.7778 0.0878 Table 5 Evaluation results of the model in the testing set Model AUC Specificity Sensitivity Accuracy Precision F1 Brier Score Polynomial 0.8644 0.8438 0.8333 0.8412 0.6364 0.7216 0.1280 Linear 0.9040 0.8359 0.8571 0.8412 0.6316 0.7273 0.0987 Radial 0.9481 0.8359 0.9286 0.8588 0.6500 0.7647 0.0858 Sigmoid 0.8443 0.8281 0.8095 0.8235 0.6071 0.6939 0.1261 Lasso 0.9003 0.8516 0.8333 0.8471 0.6481 0.7292 0.1167 Bayes 0.8797 0.8438 0.8095 0.8353 0.6296 0.7083 0.1603 Knn 0.8810 0.8516 0.8333 0.8471 0.6481 0.7292 0.0922 LightGBM 0.9688 0.9375 0.9286 0.9353 0.8298 0.8764 0.0586 Logistic 0.9046 0.8906 0.7857 0.8647 0.7021 0.7416 0.0998 Best machine model for visualization Based on the model evaluation metrics, we ultimately selected LightGBM as the top-performing machine learning model. Using SHAP for visualization, the result showed that both the day age and gestational age play a significant role in the model, with the highest contribution value. When the gestational age is smaller, the contribution value to the prediction of NEC is larger. Additionally, when the day age is 0, the contribution value is negative, indicating that NEC is less likely to occur at this time. In addition, HB levels are significantly negatively correlated with NEC, with lower HB levels being more inclined towards NEC and EOS having the other hand, parameters associated with inflammation, such as HSCRP and PDW are more inclined towards NEC with higher levels.(Fig. 4 ) Validation analysis included Long-Term Time Series Forecasting and External validation Clinical information was collected from patients diagnosed and treated as NEC stage IA, IB, IIA, and IIB by doctors of the same seniority, according to the modified BELL staging system in our hospital from October 2022 to January 2024 and Zhongshan Boai hospital. The established model was applied to the clinical features and laboratory examination information we collected. The AUC, accuracy, sensitivity and specificity of Long-Term Time Series Forecasting were 0.926, 0.933, 0.808 and 0.974, respectively. In external validation data, the AUC, accuracy, sensitivity and specificity were 0.901, 0.872, 0.776 and 0.903. (Fig. 5 ) Discussion NEC is a serious gastrointestinal disease in neonates, particularly in preterm infants, classified into three stages according to Bell's Modified Staging Criteria. The symptoms of stage I and II include unstable body temperature, apnea, decreased heart rate, abdominal symptoms such as increased gastric residue, mild abdominal distension, fecal occult blood, and normal or mild intestinal obstruction visible on X-ray. In addition, there may be a loss of bowel sounds, abdominal tenderness, intestinal obstruction, intestinal wall pneumatosis in stage II. Timely intervention can effectively prevent disease progression. Once the disease reaches stage III, the mortality rate was obvious increased[ 16 ]. Therefore, it is crucial to identify NEC as early as possible. In our study, we combined clinical characteristics with the most commonly used clinical indicators for establishing a model for predicting early NEC. Interestingly, compared to the NEC group, one-way logistic regression analysis revealed that the control group exhibited higher leukocyte, neutrophil, and lymphocyte counts. The plausible explanations for this are as follows: firstly, control subjects comprised high-risk infants with a baseline level of inflammation, whereas the NEC group's blood samples were collected during the early, less severe stages of inflammation, preceding the diagnosis. Secondly, the marginal increase in leukocytes could be attributed to the entry of Escherichia coli, endotoxin, and inflammatory factors into the bloodstream, prompting a state of acute stress and leading to a significant adhesion of activated leukocytes, primarily neutrophils, to the vascular walls, thereby resulting in a seemingly minor elevation in circulating leukocyte counts[ 19 , 20 ]. Our study was found that six indicators, including day age, gestational age, HB, PDW, HSCRP, and EOS were important in both traditional and machine learning models. LightGBM model indicates that the day age is the most important factor, often working in conjunction with gestational age, in the occurrence of NEC. It is well established that preterm infants have an underdeveloped gut, lower microbial abundance in the gut compared to term infants, and deficient immune function[ 20 , 21 , 22 ]. This leads to an increased risk of infections, with an incidence of NEC ranging from 2%-13% in preterm infants and a mortality rate as high as 20%-30%[ 23 ]. In our study, both gestational age and day age can be used as important variables in assessing the risk of NEC, consistent with previous studies[ 24 ]. Blood counts and CRP are extremely common in clinical tests. Elevated levels of HSCRP can indicate early-stage inflammation, while persistently high CRP levels can reliably indicate the development of intestinal infections. Michelle Keane et al.[ 25 ] found that elevating CRP was not only correlated with the severity of NEC but also predicted whether surgery was necessary. The same observation was made in an international survey on the management of NEC[ 26 ]. Alterations of PLT in neonate were associated with disease severity and prognosis[ 27 ]. However, in our study, PLT did not exhibit a significant difference in the initial stages of NEC. Conversely, PDW, an indicator of platelet activation, demonstrated a notable difference and served as a high-risk predictor in the model. It is probably because that in early NEC, PLT is not significantly depleted but responds to the inflammatory process by activating platelets[ 28 , 29 ]. Eosinophils, as a type of immune cell, can exacerbate intestinal inflammation by releasing specific proteins through degranulation[ 30 ]. Eosinophils was considered by Christensen RD as a white blood cell (WBC)-related biomarker for the onset of NEC[ 31 , 32 ]. Roberts et al.[ 33 ] found that persistent blood eosinophils is a predictor of complications during the course of NEC. In the best-constructed model, lower HB levels tend to correspond to more severe staging of NEC. Anemia is a prevalent comorbidity in preterm infants[ 34 ], and NEC, as an intestinal ischemic disease[ 35 ], is closely associated with anemia throughout the disease's progression. The presence of intestinal necrosis, bloody stools, and bloody peritoneal fluid leads to decrease hemoglobin level. Furthermore, the induced inflammatory response affects the production of hemoglobin and the patient's survival[ 36 ]. In a case-control study[ 37 ], the researchers compared splanchnic tissue oxygen saturation, extraction, and variability of splanchnic oxygen saturation in NEC patients with different hemoglobin levels, demonstrated that low hemoglobin levels were associated with intestinal injury. This may accelerate the development of NEC. In the predictive model we selected as the best, low levels of hemoglobin were a significant point of difference between early NEC and normal controls. In terms of model constructed, logistic regression models were also effective in in predicting disease turnover in high-risk patients. In fact, there are some models that focus on distinguishing whether NEC occurs. Julia M. Pantalone et al.[ 38 ] analyzed complete blood counts (CBCs) around illness onset in NEC cases compared with controls, and developed LDA, SVM, RF, and Logistic regression models. RF had the best performance with an AUC of 0.877, sensitivity 0.8, and specificity 0.793 for the comparison of S-NEC vs. Controls. Although RF was the best performer among all models in the comparison of S-NEC vs. M-NEC, its AUC value was still below 0.8 and had low sensitivity among all models. For intestinal perforation due to severe NEC, Irles et al.[ 39 ] developed a back-propagation artificial neural network (ANN) based on critical variables such as neonatal platelet and neutrophil levels, birthweight, sex, gestational age and maternal age. Back-propagation ANN developed two estimation models for predicting intestinal perforation caused by NEC. The application of machine learning models for early prediction of NEC has not been widely adopted, and there are currently no machine learning models capable of predicting early NEC risk based on basic data. In this study, the most frequently utilized test metrics, along with general clinical characteristics, were employed to develop prevalent machine learning models for predicting the beginning of NEC. When comparing machine learning algorithms, LightGBM and the radial basis function kernel of SVM both demonstrated excellent predictive performance for NEC in the training set and test set. By comparison, the LightGBM model proved to be the most effective in integrating each assessment metric. The model identified several risk factors for the development of NEC in neonates, including day age, gestational age, EOS, HB, PDW and HSCRP, with day age being the most critical predictor of NEC. Accurate risk factor assessment facilitates the identification of high-risk patients. In our study, we found that gestational age combined with day age at onset was highly predictive of NEC, similar to previous research[ 40 ]. The smaller the gestational age, the greater the SHAP contribution value. This indicates that preterm infants are highly susceptible to NEC around seven days after birth. Therefore, close attention should be given at this time. Our study used ML algorithms to predict NEC in early neonatal life. However, there are some limitations. It was a two-center study and the amount of sample was limited, validation of large databases is needed. In future, we will continue to refine and validate the existing machine learning model using large-scale, multicenter population data. Conclusion The LightGBM model, which considers day age, gestational age, EOS, HB, PDW, and HSCRP, demonstrates good performance in predicting NEC in the early stage. Clinicians can use this model to identify highrisk newborns requiring focused attention, enhance observation of their clinical manifestations, clarify diagnoses through imaging examinations . Abbreviations BASO basophil DCA Decision curve analysis EOS Eosinophil HB Hemoglobin HSCRP High-sensitivity C-reactive protein LASSO Least Absolute Shrinkage and Selection Operator LightGBM Light Gradient Boosting Machine LYM lymphocyte ML machine learning MONO monocyte MPV mean platelet volume NEC necrotizing enterocolitis NEU neutrophil PDW platelet distribution width PLT platelet PCT plateletcrit RBC red blood cell RF Random Forest ROC receiver operating characteristic curve SHAP SHapely Additive exPlanation SVM Support Vector Machine WBC white blood cell Declarations Ethics approval and consent to participate Research have been performed in accordance with the Declaration of Helsinki. The participants involved in this study were reviewed and approved by Ethics Committee of Zhujiang Hospital of Southern Medical University (2024-KY-123-01), ensuring that all ethical standards and requirements have been met. Informed consent was obtained from all subjects and/or their legal guardians for this study. Consent for publication Written informed consent was obtained from the parents. Competing interests 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. Funding This study was supported by the Guangdong Basic and Applied Basic Research Foundation of China (2019A1515011086) and Clinical research Fund of Guangdong Medical Association(A202302026). Author Contribution All authors contributed to the study conception and design. Yanling Mou prepared the manuscrip and collected data. Jinhao Li summarized and analyzed data. Jianjun Wang, Daiyue Yu and Huirong Yang conceived and designed the experiments. Xi Zhang, Rongying Tan and Djibril Adam Mahama made contributions to acquisition of data and interpretation of data. Liucheng Yang and Kai Wu guided the project and revised the manuscript. All authors read and approved the final manuscript. Acknowledgements Not applicable Data Availability Due to strict ethical and legal guidelines that govern the handling of patient data, the datasets generated and analysed during the current study are not publicly available but are available from the corresponding author on reasonable request. References Cao Y, Jiang S, Sun J, Hei M, Wang L, Zhang H, Ma X, Wu H, Li X, Sun H, Zhou W, Shi Y, Wang Y, Gu X, Yang T, Lu Y, Du L, Chen C, Lee SK, Zhou W, et al. Assessment of Neonatal Intensive Care Unit Practices, Morbidity, and Mortality Among Very Preterm Infants in China. JAMA Netw open. 2021;4(8):e2118904. https://doi.org/10.1001/jamanetworkopen.2021.18904 . Thompson AM, Bizzarro MJ. Necrotizing enterocolitis in newborns: pathogenesis, prevention and management. Drugs. 2008;68(9):1227–38. https://doi.org/10.2165/00003495-200868090-00004 . Yee WH, Soraisham AS, Shah VS, Aziz K, Yoon W, Lee SK, Canadian Neonatal Network. Incidence and timing of presentation of necrotizing enterocolitis in preterm infants. Pediatrics. 2012;129(2):e298–304. https://doi.org/10.1542/peds.2011-2022 . Picaud JC, De Magistris A, Mussap M, Corbu S, Dessì A, Noto A, Fanos V, Cesare Marincola F. Urine NMR Metabolomics Profile of Preterm Infants With Necrotizing Enterocolitis Over the First Two Months of Life: A Pilot Longitudinal Case-Control Study. Front Mol Biosci. 2021;8:680159. https://doi.org/10.3389/fmolb.2021.680159 . Vitali R, Terrin G, Palone F, Laudadio I, Cucchiara S, Boscarino G, Di Chiara M, Stronati L. Fecal High-Mobility Group Box 1 as a Marker of Early Stage of Necrotizing Enterocolitis in Preterm Neonates. Front Pead. 2021;9:672131. https://doi.org/10.3389/fped.2021.672131 . Liu XC, Li LQ, Ling KR, Guo L, Hu XY, Li C. Fecal HBD-2 and Claudin-3 may be potential biomarkers to predict the deterioration of necrotizing enterocolitis: A prospective study. Front Pead. 2022;10:1062798. https://doi.org/10.3389/fped.2022.1062798 . Chen G, Lv X, Tang W. Fecal calprotectin as a non-invasive marker for the prediction of post-necrotizing enterocolitis stricture. Pediatr Surg Int. 2023;39(1):250. https://doi.org/10.1007/s00383-023-05534-7 . Moschino L, Verlato G, Duci M, Cavicchiolo ME, Guiducci S, Stocchero M, Giordano G, Fascetti Leon F, Baraldi E. The Metabolome and the Gut Microbiota for the Prediction of Necrotizing Enterocolitis and Spontaneous Intestinal Perforation. Syst Rev Nutrients. 2022;14(18):3859. https://doi.org/10.3390/nu14183859 . Bell MJ, Ternberg JL, Feigin RD, Keating JP, Marshall R, Barton L, Brotherton T. Neonatal necrotizing enterocolitis. Therapeutic decisions based upon clinical staging. Ann Surg. 1978;187(1):1–7. https://doi.org/10.1097/00000658-197801000-00001 . Herzlich J, Mandel D, Marom R, Mendelsohn R, Eshel Fuhrer A, Mangel L. Blood Glucose, Lactate and Platelet Count in Infants with Spontaneous Intestinal Perforation versus Necrotizing Enterocolitis-A Pilot Study. Child (Basel Switzerland). 2023;10(6):1028. https://doi.org/10.3390/children10061028 . Zheng J, Wang H. Association Between Platelet-to-Lymphocyte Ratio and Preterm Necrotizing Enterocolitis. Front Pead. 2021;9:686880. https://doi.org/10.3389/fped.2021.686880 . Feng W, Hou J, Die X, Sun J, Guo Z, Liu W, Wang Y. Application of coagulation parameters at the time of necrotizing enterocolitis diagnosis in surgical intervention and prognosis. BMC Pediatr. 2022;22(1):259. https://doi.org/10.1186/s12887-022-03333-y . Krishnan R, Rajpurkar P, Topol EJ. Self-supervised learning in medicine and healthcare. Nat biomedical Eng. 2022;6(12):1346–52. https://doi.org/10.1038/s41551-022-00914-1 . Subspecialty Group of Neonatology, the Society of Pediatrics, Chinese Medical Association, &, Editorial, Board, Chinese Journal of Pediatrics. Zhonghua er ke za zhi = Chinese. J Pediatr. 2023;61(10):869–73. https://doi.org/10.3760/cma.j.cn112140-20230630-00434 . Kordasz M, Racine M, Szavay P, Lehner M, Krebs T, Luckert C, Hau EM, Berger S, Kessler U. Risk factors for mortality in preterm infants with necrotizing enterocolitis: a retrospective multicenter analysis. Eur J Pediatrics. 2022;181(3):933–9. https://doi.org/10.1007/s00431-021-04266-x . Meister AL, Doheny KK, Travagli RA, Maywood NJ.), 245(2), 85–95. https://doi.org/10.1177/1535370219891971 . Humberg A, Fortmann I, Siller B, Kopp MV, Herting E, Göpel W, Härtel C, German Neonatal Network, German Center for Lung Research and Priming Immunity at the beginning of life (PRIMAL) Consortium. Preterm birth and sustained inflammation: consequences for the neonate. Semin Immunopathol. 2020;42(4):451–68. https://doi.org/10.1007/s00281-020-00803-2 . Ernst E, Hammerschmidt DE, Bagge U, Matrai A, Dormandy JA. Leukocytes and the risk of ischemic diseases. JAMA. 1987;257(17):2318–24. Campbell EL, Kao DJ, Colgan SP. Neutrophils and the inflammatory tissue microenvironment in the mucosa. Immunol Rev. 2016;273(1):112–20. https://doi.org/10.1111/imr.12456 . Healy DB, Ryan CA, Ross RP, Stanton C, Dempsey EM. Clinical implications of preterm infant gut microbiome development. Nat Microbiol. 2022;7(1):22–33. https://doi.org/10.1038/s41564-021-01025-4 . Kaplina A, Kononova S, Zaikova E, Pervunina T, Petrova N, Sitkin S. Necrotizing Enterocolitis: The Role of Hypoxia, Gut Microbiome, and Microbial Metabolites. Int J Mol Sci. 2023;24(3):2471. https://doi.org/10.3390/ijms24032471 . Korpela K, Blakstad EW, Moltu SJ, Strømmen K, Nakstad B, Rønnestad AE, Brække K, Iversen PO, Drevon CA, de Vos W. Intestinal microbiota development and gestational age in preterm neonates. Sci Rep. 2018;8(1):2453. https://doi.org/10.1038/s41598-018-20827-x . Jones IH, Hall NJ. Contemporary Outcomes for Infants with Necrotizing Enterocolitis-A Systematic Review. J Pediatr. 2020;220:86–e923. https://doi.org/10.1016/j.jpeds.2019.11.011 . Samuels N, van de Graaf RA, de Jonge RCJ, Reiss IKM, Vermeulen MJ. Risk factors for necrotizing enterocolitis in neonates: a systematic review of prognostic studies. BMC Pediatr. 2017;17(1):105. https://doi.org/10.1186/s12887-017-0847-3 . Keane M, Fallon R, Riordan A, Shaw B. (2015) Markedly raised levels of C-reactive protein are associated with culture-proven sepsis or necrotising enterocolitis in extremely preterm neonates. Acta paediatrica (Oslo, Norway: 1992), 104(7), e289–e293. https://doi.org/10.1111/apa.12978 . Zani A, Eaton S, Puri P, Rintala R, Lukac M, Bagolan P, Kuebler JF, Hoellwarth ME,Wijnen R, Tovar J, Pierro A, & EUPSA Network (2015) International survey on the management of necrotizing enterocolitis. European journal of pediatric surgery: official journal of Austrian Association of Pediatric Surgery … et al] = Zeitschrift fur Kinderchirurgie,25(1), 27–33. https://doi.org/10.1055/s-0034-1387942. Go H, Ohto H, Nollet KE, Takano S, Kashiwabara N, Chishiki M, Maeda H, Imamura T, Kawasaki Y, Momoi N, Hosoya M. Using Platelet Parameters to Anticipate Morbidity and Mortality Among Preterm Neonates: A Retrospective Study. Front Pead. 2020;8:90. https://doi.org/10.3389/fped.2020.00090 . Koupenova M, Clancy L, Corkrey HA, Freedman JE. Circulating Platelets as Mediators of Immunity, Inflammation, and Thrombosis. Circul Res. 2018;122(2):337–51. https://doi.org/10.1161/CIRCRESAHA.117.310795 . van der Meijden PEJ, Heemskerk JWM. Platelet biology and functions: new concepts and clinical perspectives. Nat Rev Cardiol. 2019;16(3):166–79. https://doi.org/10.1038/s41569-018-0110-0 . Gurtner A, Borrelli C, Gonzalez-Perez I, Bach K, Acar IE, Núñez NG, Crepaz D, Handler K, Vu VP, Lafzi A, Stirm K, Raju D, Gschwend J, Basler K, Schneider C, Slack E, Valenta T, Becher B, Krebs P, Moor AE, … Arnold I C (2023) Active eosinophils regulate host defence and immune responses in colitis. Nature, 615(7950), 151–157. https://doi.org/10.1038/s41586-022-05628-7. Christensen RD, Lambert DK, Gordon PV, Baer VL, Gerday E, Henry E. Neonates presenting with bloody stools and eosinophilia can progress to two different types of necrotizing enterocolitis. J perinatology: official J Calif Perinat Association. 2012;32(11):874–9. https://doi.org/10.1038/jp.2011.163 . Wahidi LS, Sherman J, Miller MM, Zaghouani H, Sherman MP. Early Persistent Blood Eosinophilia in Necrotizing Enterocolitis Is a Predictor of Late Complications. Neonatology. 2015;108(2):137–42. https://doi.org/10.1159/000431305 . Zdravic D, Yougbare I, Vadasz B, Li C, Marshall AH, Chen P, Kjeldsen-Kragh J, Ni H. Fetal and neonatal alloimmune thrombocytopenia. Semin Fetal Neonatal Med. 2016;21(1):19–27. https://doi.org/10.1016/j.siny.2015.12.004 . Patel RM, Knezevic A, Shenvi N, Hinkes M, Keene S, Roback JD, Easley KA, Josephson CD. Association of Red Blood Cell Transfusion, Anemia, and Necrotizing Enterocolitis in Very Low-Birth-Weight Infants. JAMA. 2016;315(9):889–97. https://doi.org/10.1001/jama.2016.1204 . Roberts AG, Younge N, Greenberg RG. Neonatal Necrotizing Enterocolitis: An Update on Pathophysiology, Treatment, and Prevention. Paediatr Drugs. 2024;26(3):259–75. https://doi.org/10.1007/s40272-024-00626-w . Ganz T. Anemia of Inflammation. N Engl J Med. 2019;381(12):1148–57. https://doi.org/10.1056/NEJMra1804281 . Kalteren WS, Bos AF, van Oeveren W, Hulscher JBF, Kooi EMW. Neonatal anemia relates to intestinal injury in preterm infants. Pediatr Res. 2022;91(6):1452–8. https://doi.org/10.1038/s41390-021-01903-x . Pantalone JM, Liu S, Olaloye OO, Prochaska EC, Yanowitz T, Riley MM, Buland JR, Brozanski BS, Good M, Konnikova L. Gestational Age-Specific Complete Blood Count Signatures in Necrotizing Enterocolitis. Front Pead. 2021;9:604899. https://doi.org/10.3389/fped.2021.604899 . Irles C, González-Pérez G, Carrera Muiños S, Michel Macias C, Sánchez Gómez C, Martínez-Zepeda A, Cordero González G, Laresgoiti Servitje E. Estimation of Neonatal Intestinal Perforation Associated with Necrotizing Enterocolitis by Machine Learning Reveals New Key Factors. Int J Environ Res Public Health. 2018;15(11):2509. https://doi.org/10.3390/ijerph15112509 . Sharma R, Hudak ML, Tepas JJ 3rd, Wludyka PS, Marvin WJ, Bradshaw JA, Pieper P. Impact of gestational age on the clinical presentation and surgical outcome of necrotizing enterocolitis. J perinatology: official J Calif Perinat Association. 2006;26(6):342–7. https://doi.org/10.1038/sj.jp.7211510 . Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4556691","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":314544915,"identity":"5198d4f8-8db7-4271-b28f-bce7d0413331","order_by":0,"name":"Yanling Mou","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yanling","middleName":"","lastName":"Mou","suffix":""},{"id":314544916,"identity":"4c7c0dc4-8833-43f9-a07a-62c6171b7a46","order_by":1,"name":"Jinhao Li","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jinhao","middleName":"","lastName":"Li","suffix":""},{"id":314544917,"identity":"f8986955-2775-4db1-a7c2-2fdecc842d53","order_by":2,"name":"Jianjun Wang","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jianjun","middleName":"","lastName":"Wang","suffix":""},{"id":314544918,"identity":"38e1fe39-92fb-4cc5-b24a-a96886f616ef","order_by":3,"name":"Daiyue Yu","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Daiyue","middleName":"","lastName":"Yu","suffix":""},{"id":314544919,"identity":"68f011e6-0526-45bc-bb99-c59a183a1ff7","order_by":4,"name":"Huirong Yang","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Huirong","middleName":"","lastName":"Yang","suffix":""},{"id":314544920,"identity":"1b4827f8-19f9-4e3c-b1dd-0a0b42d5f7d4","order_by":5,"name":"Xi Zhang","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xi","middleName":"","lastName":"Zhang","suffix":""},{"id":314544922,"identity":"4aa587c9-5952-412b-9133-6b55e79ccd0a","order_by":6,"name":"Rongying Tan","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Rongying","middleName":"","lastName":"Tan","suffix":""},{"id":314544924,"identity":"eec8e2c9-ffb5-4d67-a062-d78352efb214","order_by":7,"name":"Djibril Adam Mahamat","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Djibril","middleName":"Adam","lastName":"Mahamat","suffix":""},{"id":314544927,"identity":"ce7c0e05-dbd8-4612-9c47-a0acf8698bf6","order_by":8,"name":"Liucheng Yang","email":"","orcid":"","institution":"Southern Medical University","correspondingAuthor":false,"prefix":"","firstName":"Liucheng","middleName":"","lastName":"Yang","suffix":""},{"id":314544928,"identity":"c4b9a582-36c9-482e-99ad-9f9c076e3cc4","order_by":9,"name":"Kai Wu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAArUlEQVRIiWNgGAWjYFACNgaGhAoIU4IELWdI1sLYRooWgxtpiR8ezquLNjjAfPA2D4NdHjFaDkskbjucu+EAW7I1D0NyMUEtZjfSG4BaDgC18JhJ8zAcSGwgQkvzj8Q5dUAt/N+I1ZJ2TCKxgRlkCxtxWuzPPEuzSDh2OHfmYTZjyzkGyYS1SLanGd/8UVOX23e8+eGNNxV2hLUgADOIMCBe/SgYBaNgFIwCPAAAKCw+EvU6dKkAAAAASUVORK5CYII=","orcid":"","institution":"Southern Medical University","correspondingAuthor":true,"prefix":"","firstName":"Kai","middleName":"","lastName":"Wu","suffix":""}],"badges":[],"createdAt":"2024-06-10 08:23:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4556691/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4556691/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":59597907,"identity":"51bcdf6e-2039-47fd-b1cd-6271de4ab6f7","added_by":"auto","created_at":"2024-07-03 16:16:25","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2000621,"visible":true,"origin":"","legend":"\u003cp\u003eResearch Design\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4556691/v1/358ccd093d0b70bb2b2be442.jpeg"},{"id":59597908,"identity":"ed346b70-071e-4928-bdd8-5b26ab1468f0","added_by":"auto","created_at":"2024-07-03 16:16:25","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2862887,"visible":true,"origin":"","legend":"\u003cp\u003eScreening of Important Clinical Characteristics Variables (A) LASSO Regression Regularization Parameter Plot, Choosing lambda.1-s.e. implies a simpler and more interpretable model, and choosing lambda.min results in a model with optimal performance, lambda.min=lambda.1-s.e.=12 in our study (B) LASSO Coefficient Distribution Plot that The 12 clinical features with coefficients not 0 were filtered (C) Histogram of the importance of the variables based on the mean decrease in correctness of Random Forest, where day age and PDW contributed the most to the model (D) Combining the mean decrease in correctness of Random Forest and the mean decrease in the Gini index, a dot plot of the importance of each variable was obtained (E) Based on the Support Vector Machines, 13 variables were selected to be included in the next step of the analysis ( (F) Importance plot of traits based on support vector machine\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4556691/v1/ff7361220658f81b3ad51313.jpeg"},{"id":59597909,"identity":"9afa66dd-c236-4ec3-95cb-57b6fcfed158","added_by":"auto","created_at":"2024-07-03 16:16:25","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1766554,"visible":true,"origin":"","legend":"\u003cp\u003eA. AUC values of multiple machine learning models in the training set B. AUC values of multiple machine learning models in the testing set\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4556691/v1/653d5e396726cc5b5ccb5122.jpeg"},{"id":59597911,"identity":"24e9f22c-cfbe-4dd3-bf90-ee8f7e30759a","added_by":"auto","created_at":"2024-07-03 16:16:25","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2684710,"visible":true,"origin":"","legend":"\u003cp\u003eVisualization and model evaluation of the best model LightGBM selected for the study A. Dependency plot based on SHAP values. B. Histogram of the importance of the variables for the best model LightGBM. C. LightGBM model calibration plot, the confidence intervals fully contain the dashed line indicating a good model fit. D. Plot of clinical decision making curves of the LightGBM model in the training set, the horizontal coordinate is the threshold probability, vertical coordinate is net benefit, treat None line represents all negative, treat All represents all positive, LightGBM model curve is all above treat All diagonal line, indicating that it can have clinical benefit in practically all cases.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4556691/v1/2ba01930356ae12e920029f8.jpeg"},{"id":59598600,"identity":"3c3675cd-4d56-4720-a669-3207a41c1a66","added_by":"auto","created_at":"2024-07-03 16:24:25","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1230574,"visible":true,"origin":"","legend":"\u003cp\u003eValidation of the best model A. The performance of the established best model in the time series with an AUC value of 0.92 indicates that it is applicable to future clinical decisions in our center B. The performance of the LightGBM model in the clinical center with an AUC of 0.872 shows that the model has good extrapolation properties\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4556691/v1/b35650103318f293fcabd403.jpeg"},{"id":72686554,"identity":"47469b76-2bd0-46d6-a619-6756af376af0","added_by":"auto","created_at":"2024-12-31 08:23:59","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4611982,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4556691/v1/dfaa3999-fb57-4c2b-9233-444c46b44fb0.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"An Early Prediction of Neonatal Necrotizing Enterocolitis in High-Risk Newborns- Based on Two Medical Center Clinical Databases","fulltext":[{"header":"What is Known","content":"\u003cp\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eThe impact of necrotizing enterocolitis (NEC) in newborns has garnered ample attention from pediatricians, yet inexperience and the absence of advanced and sensitive diagnostic tools often result in a failure to timely control the disease.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eA variety of models have been developed to identify children with the modified Bell staging criteria and staged as II and III.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eWhat is New\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eMachine learning models that are trained on big data can assist in identifying patients with NEC at an earlier stage.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eEven though early NEC poses a risk of misdiagnosis, no child with a potential for a serious adverse prognosis should be ignored.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eNowadays, necrotizing enterocolitis (NEC) is one of the primary causes of mortality and morbidity in neonatal intensive care units[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. The overall incidence of NEC is approximately 1\u0026ndash;5 per 100 live births[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], with a particularly high occurrence in preterm infants[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The clinical presentation of NEC is complex, such as abdominal distension, vomiting, and fecal occult blood. However, in many cases, the symptoms are atypical, and the condition progress rapidly. We primarily relied on doctors' clinical experience to make a diagnosis formerly. However, it is difficult for doctors in primary hospitals who were the main management of neonate in China to recognize NEC at an early stage. With the ever-increasing availability of novel markers[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], clinicians remain relentlessly on the hunt for more sensitive technological advancements, albeit diminishing their emphasis on fundamental patient data. Therefore, it is undoubtedly significant to develop a visual clinical prediction models and offer insights for early clinical diagnosis through model scoring.\u003c/p\u003e \u003cp\u003eThe modified Bell staging[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] is still the most commonly used clinical staging method currently. In addition, platelet count[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e], lymphocyte count[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e], and coagulation function indexes[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] were also considered as early indicators of NEC. Machine learning is a process of automatic learning from computerized data that enables computers to discern laws and patterns from data and make predictions and decisions based on them[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. It is important to analyze and process the available data on a large scale using machine learning in order to develop a new application model for the early and accurate identification of NEC.\u003c/p\u003e \u003cp\u003eOur study proposes the construction of an early prediction model for NEC using machine learning, aiming to validate the effectiveness of the prediction model and constructing a visualization model to enable early and accurate identification of NEC patients. This will provide a new approach for the early diagnosis of neonates, particularly preterm infants with NEC.\u003c/p\u003e"},{"header":"Population and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eResearch population\u003c/h2\u003e \u003cp\u003eNeonates at Zhujiang hospital of Southern Medical University from October 2010 to October 2022 were selected. The inclusion criteria for the NEC group were as follows: (1) Diagnosis by two clinically experienced pediatricians in accordance with the modified Bell staging criteria[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] and staged as IA, IB, IIA, or IIB (which includes clinical data from the early stages of disease for the patients diagnosed as stage III according to criteria); (2) Newborns under 28 days old. The exclusion criteria: (1) Individuals with genetic metabolic disorders or congenital malformations; (2) Individuals with tumor diseases; (3) Use of drugs that may cause changes in blood routine; (4) Any coagulation abnormalities, such as intracranial hemorrhage, gastrointestinal hemorrhage, pulmonary hemorrhage, sepsis, or septicemia. (5) Lack of clinical or laboratory data. Non-NEC neonates who were admitted at the same time were randomly selected in proportion to serve as the control group. According to the classification management method for high-risk newborns[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], if one of the following inclusion criteria exists, we consider the infant to be in a high-risk state: (1) Premature infants. (2) Low birth weight infants. (3) Small for gestational age infants. (4) Hyperbilirubinemia. (5) Water electrolyte and metabolic disorders. (6) Retinopathy of premature infants. (7) Other conditions such as hypothermia, apnea, omphalitis, head hematoma, and no other accompanying conditions exclusion criteria are the same as NEC group.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eStudy Indicators\u003c/h2\u003e \u003cp\u003eThe clinical characteristics of the patients included gender, Day age, Gestational age, Birthweight, Maternal age, Apgar score at 1 minute, mode of delivery, and the presence of patent ductus arteriosus[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Birthweight was averaged from three measurements by two trained nurses, and arterial duct failure was diagnosed by a cardiac ultrasound expert.\u003c/p\u003e \u003cp\u003eThe control group underwent the first routine blood test after admission, while the latest blood routine results of NEC group before diagnosis as the disease group. The collected information included the white blood cell count (WBC), absolute neutrophil value (NEU), absolute lymphocyte value (LYM), absolute monocyte value (MONO), absolute eosinophil value (EOS), absolute basophil value (BASO), red blood cell count (RBC), hemoglobin (HB), platelet count (PLT), mean platelet volume (MPV), platelet pressure product (PCT), platelet volume distribution width (PDW), and High-sensitivity C-reactive protein (HSCRP) levels.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eAnalytical Methods\u003c/h2\u003e \u003cp\u003eMissing data were filled using the R package \"mice\", employing multiple imputation with the number of imputations set to 5. Subsequently, the dataset was randomly divided into training and test sets at a 7:3 ratio, and stratified sampling was performed using the R package \"caret\".\u003c/p\u003e \u003cp\u003eLogistic one-way analysis of variance (LASSO) was conducted on the training set using the R package \"stats\". Subsequently, the metrics that showed statistical significance (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) were subjected to further analysis using LASSO with the R package \"glmnet\". Support Vector Machine (SVM) analysis was conducted using the R package \"e1071\" to determine the optimal number of features, based on the observed decrease in the mean square error plot. The corresponding number of features were then included in order of importance. Additionally, Random Forest (RF) analysis was carried out using the R package \"RandomForest\", with a criterion of MeanDecreaseAccuracy\u0026thinsp;\u0026gt;\u0026thinsp;10. Subsequently, LASSO analyses were conducted using the Draw Venn Diagram website (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://bioinformatics.psb.ugent.be/\u003c/span\u003e\u003cspan address=\"https://bioinformatics.psb.ugent.be/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) to identify the intersection of the important variables obtained from screening the three models: LASSO, SVM, and RF. Further logistic multifactor analysis was performed on the intersection features (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003eThe most significant features in the intersection set were utilized to build four SVM models with linear, polynomial, radial, and sigmoid using the R package \"e1071\". Additionally, a K-nearest-neighbor model was constructed using the R package \"caret\", which selects the optimal K-value through grid search. Furthermore, a Bayes model was built using the R package \"e1071\", and a LightGBM model was constructed using the R package \"LightGBM\". Compare the area under the ROC curve, accuracy, sensitivity, and specificity of each model to determine the best model. Finally, the R package \"runway\" was utilized to construct the calibration curves for the respective models, and the R package \"dcurves\" was employed to create the clinical decision curves for evaluating their clinical predictive value. (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eClinical characteristics\u003c/h2\u003e \u003cp\u003eThe study population consisted of 143 NEC patients and the control group comprised a randomized sample of 429 contemporaneous non-NEC neonates. The study population was divided into a training set and a testing set at a 7:3 ratio. One-way logistic regression analysis was performed on the training set, revealing statistically differences between the two groups in terms of day age(\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Gestational age (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Birth weight (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), Apgar score at 1 minute (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and delivery mode (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.031). However, there were no significant differences in the mother's age (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.810), the presence of arterial ductus arteriosus (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.081) and sex (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.202). (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBaseline and clinical characteristics of the study cohort.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNormal (N\u0026thinsp;=\u0026thinsp;301)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNEC (N\u0026thinsp;=\u0026thinsp;101)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOR (univariable)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e126 (41.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e35 (34.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e175 (58.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e66 (65.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.36 (0.85\u0026ndash;2.17, p\u0026thinsp;=\u0026thinsp;0.202)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDay\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.0\u0026thinsp;\u0026plusmn;\u0026thinsp;4.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.1\u0026thinsp;\u0026plusmn;\u0026thinsp;7.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.20 (1.14\u0026ndash;1.27, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGestational age\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e38.4\u0026thinsp;\u0026plusmn;\u0026thinsp;2.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e35.9\u0026thinsp;\u0026plusmn;\u0026thinsp;4.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.71 (0.66\u0026ndash;0.78, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBirthweight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.0\u0026thinsp;\u0026plusmn;\u0026thinsp;0.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.3\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.28 (0.20\u0026ndash;0.40, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMomYear\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e30.7\u0026thinsp;\u0026plusmn;\u0026thinsp;4.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30.5\u0026thinsp;\u0026plusmn;\u0026thinsp;5.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.99 (0.95\u0026ndash;1.04, p\u0026thinsp;=\u0026thinsp;0.810)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eApgar score at 1 minute\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9.4\u0026thinsp;\u0026plusmn;\u0026thinsp;1.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.7\u0026thinsp;\u0026plusmn;\u0026thinsp;2.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.78 (0.68\u0026ndash;0.89, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDelivery mode\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e183 (60.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e49 (48.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCesarean\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e118 (39.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e52 (51.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.65 (1.05\u0026ndash;2.59, p\u0026thinsp;=\u0026thinsp;0.031)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e228 (75.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e85 (84.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e73 (24.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e16 (15.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.59 (0.32\u0026ndash;1.07, p\u0026thinsp;=\u0026thinsp;0.081)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eLaboratory tests\u003c/h2\u003e \u003cp\u003eThere was no significant difference in MONO (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.351) between the NEC and non-NEC groups. However, two groups exhibited statistically differences in WBC (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), NEU (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), LYM (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.032), EOS (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001), BASO (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), RBC (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), HB (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), PLT (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), MPV (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), PCT (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), PDW (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and HSCRP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001).(Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSerum markers of the study cohort.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNormal (N\u0026thinsp;=\u0026thinsp;301)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNEC (N\u0026thinsp;=\u0026thinsp;101)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eOR (univariable)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e14.8\u0026thinsp;\u0026plusmn;\u0026thinsp;6.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e11.5\u0026thinsp;\u0026plusmn;\u0026thinsp;6.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.90 (0.87\u0026ndash;0.95, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNEU\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e9.6\u0026thinsp;\u0026plusmn;\u0026thinsp;5.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e6.5\u0026thinsp;\u0026plusmn;\u0026thinsp;4.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.88 (0.83\u0026ndash;0.92, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLYM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e3.5\u0026thinsp;\u0026plusmn;\u0026thinsp;1.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e3.1\u0026thinsp;\u0026plusmn;\u0026thinsp;2.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.85 (0.73\u0026ndash;0.99, p\u0026thinsp;=\u0026thinsp;0.032)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMONO\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e1.4\u0026thinsp;\u0026plusmn;\u0026thinsp;0.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e1.5\u0026thinsp;\u0026plusmn;\u0026thinsp;1.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.13 (0.88\u0026ndash;1.45, p\u0026thinsp;=\u0026thinsp;0.351)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEOS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e0.3\u0026thinsp;\u0026plusmn;\u0026thinsp;0.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e0.6\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.28 (1.39\u0026ndash;3.73, p\u0026thinsp;=\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBASO\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e0.1\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e0.1\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.00 (0.00\u0026ndash;0.00, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e4.6\u0026thinsp;\u0026plusmn;\u0026thinsp;0.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e3.9\u0026thinsp;\u0026plusmn;\u0026thinsp;0.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.26 (0.18\u0026ndash;0.37, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e161.7\u0026thinsp;\u0026plusmn;\u0026thinsp;24.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e131.7\u0026thinsp;\u0026plusmn;\u0026thinsp;27.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.96 (0.95\u0026ndash;0.97, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePLT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e278.9\u0026thinsp;\u0026plusmn;\u0026thinsp;93.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e227.5\u0026thinsp;\u0026plusmn;\u0026thinsp;150.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.00 (0.99-1.00, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMPV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e10.2\u0026thinsp;\u0026plusmn;\u0026thinsp;1.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e10.9\u0026thinsp;\u0026plusmn;\u0026thinsp;1.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.66 (1.36\u0026ndash;2.02, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePCT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e0.3\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e0.2\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.00 (0.00-0.05, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePDW\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e11.6\u0026thinsp;\u0026plusmn;\u0026thinsp;2.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e15.3\u0026thinsp;\u0026plusmn;\u0026thinsp;3.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.43 (1.32\u0026ndash;1.56, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHSCRP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e3.2\u0026thinsp;\u0026plusmn;\u0026thinsp;8.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e28.8\u0026thinsp;\u0026plusmn;\u0026thinsp;53.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.06 (1.04\u0026ndash;1.09, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eScreening the characteristic variables for constructing the model\u003c/h2\u003e \u003cp\u003eLASSO mitigated the impact of multicollinearity on regression results by setting the coefficients of related independent variables to 0 based on the correlation between independent variables. A total of 12 most relevant indicators and feature variables were selected. RF was used to build multiple decision trees by randomly selecting samples and features from the dataset and combining their results to make predictions and a total of 13 feature variables were screened. SVM was used to construct a model by maximizing the distance between data points in the space between categories and planes, aiming to find the optimal boundary. To maximize the sum of the distances from the data points of each category in the space to the classification boundary, also known as the support vector, a total of 9 feature variables were selected. The intersection was then taken to identify the important feature variables in all three models, including Day age, Gestational age, Birth weight, LYM, EOS, HB, PDW and HSCRP. (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAnd then, the six feature variables were analyzed using multifactor logistic regression. The results indicated that the intersection of Day age (OR\u0026thinsp;=\u0026thinsp;1.15; 95% CI: 1.07\u0026ndash;1.23; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001), Gestational Age (OR\u0026thinsp;=\u0026thinsp;0.77; 95% CI: 0.62\u0026ndash;0.95; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.016), EOS (OR\u0026thinsp;=\u0026thinsp;3.76; 95% CI: 1.76\u0026ndash;8.02; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001), HB (OR\u0026thinsp;=\u0026thinsp;0.98; 95% CI: 0.97-1.00; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001), PDW (OR\u0026thinsp;=\u0026thinsp;1.21; 95% CI: 1.08\u0026ndash;1.35; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001) and HSCRP (OR\u0026thinsp;=\u0026thinsp;1.03; 95% CI: 1.01\u0026ndash;1.05; \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006) were identified as risk factors (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) .\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMultivariate analysis of common variables\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNormal\u003c/p\u003e \u003cp\u003e(N\u0026thinsp;=\u0026thinsp;301)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNEC\u003c/p\u003e \u003cp\u003e(N\u0026thinsp;=\u0026thinsp;101)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eOR (univariable)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOR (multivariable)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDay\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e2.0\u0026thinsp;\u0026plusmn;\u0026thinsp;4.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e8.1\u0026thinsp;\u0026plusmn;\u0026thinsp;7.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.20 (1.14\u0026ndash;1.27, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.15 (1.07\u0026ndash;1.23, P\u0026thinsp;=\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGestational age\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e38.4\u0026thinsp;\u0026plusmn;\u0026thinsp;2.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e35.0\u0026thinsp;\u0026plusmn;\u0026thinsp;4.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.71 (0.65\u0026ndash;0.77, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.77 (0.62\u0026ndash;0.95, P\u0026thinsp;=\u0026thinsp;0.016)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBirthweight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e3.0\u0026thinsp;\u0026plusmn;\u0026thinsp;0.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e2.3\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.26 (0.18\u0026ndash;0.37, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.75 (0.31\u0026ndash;1.83, P\u0026thinsp;=\u0026thinsp;0.533)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLYM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e3.5\u0026thinsp;\u0026plusmn;\u0026thinsp;1.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e3.1\u0026thinsp;\u0026plusmn;\u0026thinsp;2.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.85 (0.73\u0026ndash;0.99, p\u0026thinsp;=\u0026thinsp;0.032)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.81 (0.66-1.00, P\u0026thinsp;=\u0026thinsp;0.053)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEOS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e0.3\u0026thinsp;\u0026plusmn;\u0026thinsp;0.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e0.6\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.28 (1.39\u0026ndash;3.73, p\u0026thinsp;=\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.76 (1.76\u0026ndash;8.02, P\u0026thinsp;=\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e161.7\u0026thinsp;\u0026plusmn;\u0026thinsp;24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e131.7\u0026thinsp;\u0026plusmn;\u0026thinsp;27.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.96 (0.95\u0026ndash;0.97, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.98 (0.97-1.00, P\u0026thinsp;=\u0026thinsp;0.011)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePDW\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e11.6\u0026thinsp;\u0026plusmn;\u0026thinsp;2.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e15.3\u0026thinsp;\u0026plusmn;\u0026thinsp;3.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.43 (1.32\u0026ndash;1.56, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.21 (1.08\u0026ndash;1.35, P\u0026thinsp;=\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHSCRP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e3.2\u0026thinsp;\u0026plusmn;\u0026thinsp;8.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e28.8\u0026thinsp;\u0026plusmn;\u0026thinsp;53.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.06 (1.04\u0026ndash;1.09, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.03 (1.01\u0026ndash;1.06, P\u0026thinsp;=\u0026thinsp;0.007)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eMachine Model Construction\u003c/h2\u003e \u003cp\u003e(1) \u003cb\u003eNEC prediction model based on Lasso variable screening\u003c/b\u003e\u003c/p\u003e \u003cp\u003eLASSO regression was employed to screen and downscale the six variables that showed statistical differences in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, Day age, Gestational age, EOS, HB, PDW, and HSCRP. As the log (λ) increased, the mean standard error also increased, and the normalization coefficients of the two candidate variables were compressed to varying degrees until they all reached zero. During the construction of the model, we found that the LASSO model possesses the most prominent stability, even though it is a non-optimal model in terms of numerical evaluation.\u003c/p\u003e \u003cp\u003e(2) \u003cb\u003eModeling based on simple Bayesian screening variables\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe Naive Bayes model assumed that the features were independent, which simplifies the plain Bayes algorithm but can sometimes sacrifice a certain level of classification accuracy. The simple Bayesian model performed poorly in our selection, as evidenced by unstable AUC values in the training set and validation set, and relatively low sensitivity, specificity, and accuracy.\u003c/p\u003e \u003cp\u003e(3) \u003cb\u003eModeling based on screening variables using SVM\u003c/b\u003e\u003c/p\u003e \u003cp\u003eBased on the six feature variables in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, four kernel functions were used in the SVM model to build the model. The polynomial, linear, radial, and sigmoid AUCs in the validation set were 0.94, 0.92, 0.97 and 0.81, respectively. The polynomial performed the best with a specificity of 0.94 and 0.91 between the two groups, but the AUC performance is unstable at 0.94 and 0.87. On the other hand, the radial function has stable and the highest AUC performance between the two groups, and we choose it as the best function in SVM to build the model.\u003c/p\u003e \u003cp\u003e(4) \u003cb\u003eModeling based on KNN screening variables\u003c/b\u003e\u003c/p\u003e \u003cp\u003eK Nearest Neighbors (KNN) is a classification algorithm in supervised learning, which is well-suited for the binary nature of our data. In our study, the optimal K value was selected based on the data characteristics and then cross-validation methods were used to build the KNN model. The KNN generally performs well, yielding better prediction results, but shows significant variation between the two groups.\u003c/p\u003e \u003cp\u003e(5) \u003cb\u003eModeling based on LightGBM variable screening\u003c/b\u003e\u003c/p\u003e \u003cp\u003eLight Gradient Boosting Machine (LightGBM) is a method for implementing Gradient Boosting Decision Trees. As the model we finally selected after thorough consideration, this model has the best values among the evaluation indicators. In addition, even though we only used the training set data for modeling, we achieved more stable validation results in the validation set with higher AUC, accuracy, and sensitivity.\u003c/p\u003e \u003cp\u003e(6) \u003cb\u003eModeling based on logistic variable screening\u003c/b\u003e\u003c/p\u003e \u003cp\u003eIn order to further evaluate the prediction performance of machine learning models, we built a traditional logistic regression statistical model with six selected variables. Most of the machine learning models have higher prediction efficacy than traditional statistical models, which may be related to the characteristics of machine learning itself. On the other hand, the AUC of the traditional regression model is close to 0.9 in both the training set and the test set, which to some extent also indicates the reliability of these metrics.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eModel Evaluation\u003c/h2\u003e \u003cp\u003eThe predictive performance of the five models was evaluated using AUC, accuracy, precision, sensitivity specificity F1-Score and Brier Score. A 10-fold cross-validation (CV) was applied to avoid overfitting. The AUC of the Radial of SVM and KNN models was 0.956 and 0.972 in the training set, but in the test set, their AUC values declined significantly to 0.948 and 0.881, respectively. While the predictive performance of LightGBM based on AUC was stable, with AUC values of 0.960 and 0.969, respectively(Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Furthermore, LightGBM demonstrates higher accuracy, precision, sensitivity, specificity and F1-Score of 0.935, 0.830, 0.929, 0.938 and 0.876, respectively, compared to the stability of SVM, LASSO, Bayesian, and KNN in the test set. In addition to the sigmoid of SVM, all five machine learning models exhibit better prediction performance than the traditional logistic regression model (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).(Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEvaluation results of the model in the training set\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eF1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eBrier Score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePolynomial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9332\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9269\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8614\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9104\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7982\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.8286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0963\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLinear\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9165\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8538\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8614\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8557\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6641\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0871\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRadial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9560\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9169\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8713\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9055\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7788\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.8224\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0631\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSigmoid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.7742\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8372\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.6931\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.5882\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.6364\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1442\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLasso\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9172\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8870\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8614\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8806\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7190\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7838\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0998\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBayes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8339\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8614\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8408\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6350\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7311\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1198\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9722\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8571\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8930\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.8245\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0569\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLightGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9596\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8937\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8955\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7398\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.8125\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0715\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLogistic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9190\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8970\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8317\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8806\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.8806\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7778\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0878\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEvaluation results of the model in the testing set\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eF1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eBrier Score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePolynomial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.8644\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8438\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8333\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8412\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6364\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7216\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1280\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLinear\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9040\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8359\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8571\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8412\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6316\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7273\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0987\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRadial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9481\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8359\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8588\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7647\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0858\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSigmoid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.8443\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8281\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8235\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6071\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.6939\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1261\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLasso\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8516\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8333\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8471\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6481\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7292\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1167\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBayes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.8797\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8438\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8353\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6296\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7083\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1603\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKnn\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.8810\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8516\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8333\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8471\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.6481\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7292\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0922\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLightGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9688\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9375\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9353\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.8298\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.8764\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0586\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLogistic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9046\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8906\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7857\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.8647\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.7021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.7416\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0998\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eBest machine model for visualization\u003c/h2\u003e \u003cp\u003eBased on the model evaluation metrics, we ultimately selected LightGBM as the top-performing machine learning model. Using SHAP for visualization, the result showed that both the day age and gestational age play a significant role in the model, with the highest contribution value. When the gestational age is smaller, the contribution value to the prediction of NEC is larger. Additionally, when the day age is 0, the contribution value is negative, indicating that NEC is less likely to occur at this time. In addition, HB levels are significantly negatively correlated with NEC, with lower HB levels being more inclined towards NEC and EOS having the other hand, parameters associated with inflammation, such as HSCRP and PDW are more inclined towards NEC with higher levels.(Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eValidation analysis included Long-Term Time Series Forecasting and External validation\u003c/h2\u003e \u003cp\u003eClinical information was collected from patients diagnosed and treated as NEC stage IA, IB, IIA, and IIB by doctors of the same seniority, according to the modified BELL staging system in our hospital from October 2022 to January 2024 and Zhongshan Boai hospital. The established model was applied to the clinical features and laboratory examination information we collected. The AUC, accuracy, sensitivity and specificity of Long-Term Time Series Forecasting were 0.926, 0.933, 0.808 and 0.974, respectively. In external validation data, the AUC, accuracy, sensitivity and specificity were 0.901, 0.872, 0.776 and 0.903. (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e)\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eNEC is a serious gastrointestinal disease in neonates, particularly in preterm infants, classified into three stages according to Bell's Modified Staging Criteria. The symptoms of stage I and II include unstable body temperature, apnea, decreased heart rate, abdominal symptoms such as increased gastric residue, mild abdominal distension, fecal occult blood, and normal or mild intestinal obstruction visible on X-ray. In addition, there may be a loss of bowel sounds, abdominal tenderness, intestinal obstruction, intestinal wall pneumatosis in stage II. Timely intervention can effectively prevent disease progression. Once the disease reaches stage III, the mortality rate was obvious increased[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Therefore, it is crucial to identify NEC as early as possible. In our study, we combined clinical characteristics with the most commonly used clinical indicators for establishing a model for predicting early NEC. Interestingly, compared to the NEC group, one-way logistic regression analysis revealed that the control group exhibited higher leukocyte, neutrophil, and lymphocyte counts. The plausible explanations for this are as follows: firstly, control subjects comprised high-risk infants with a baseline level of inflammation, whereas the NEC group's blood samples were collected during the early, less severe stages of inflammation, preceding the diagnosis. Secondly, the marginal increase in leukocytes could be attributed to the entry of Escherichia coli, endotoxin, and inflammatory factors into the bloodstream, prompting a state of acute stress and leading to a significant adhesion of activated leukocytes, primarily neutrophils, to the vascular walls, thereby resulting in a seemingly minor elevation in circulating leukocyte counts[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOur study was found that six indicators, including day age, gestational age, HB, PDW, HSCRP, and EOS were important in both traditional and machine learning models. LightGBM model indicates that the day age is the most important factor, often working in conjunction with gestational age, in the occurrence of NEC. It is well established that preterm infants have an underdeveloped gut, lower microbial abundance in the gut compared to term infants, and deficient immune function[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. This leads to an increased risk of infections, with an incidence of NEC ranging from 2%-13% in preterm infants and a mortality rate as high as 20%-30%[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. In our study, both gestational age and day age can be used as important variables in assessing the risk of NEC, consistent with previous studies[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eBlood counts and CRP are extremely common in clinical tests. Elevated levels of HSCRP can indicate early-stage inflammation, while persistently high CRP levels can reliably indicate the development of intestinal infections. Michelle Keane et al.[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] found that elevating CRP was not only correlated with the severity of NEC but also predicted whether surgery was necessary. The same observation was made in an international survey on the management of NEC[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAlterations of PLT in neonate were associated with disease severity and prognosis[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. However, in our study, PLT did not exhibit a significant difference in the initial stages of NEC. Conversely, PDW, an indicator of platelet activation, demonstrated a notable difference and served as a high-risk predictor in the model. It is probably because that in early NEC, PLT is not significantly depleted but responds to the inflammatory process by activating platelets[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Eosinophils, as a type of immune cell, can exacerbate intestinal inflammation by releasing specific proteins through degranulation[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Eosinophils was considered by Christensen RD as a white blood cell (WBC)-related biomarker for the onset of NEC[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Roberts et al.[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] found that persistent blood eosinophils is a predictor of complications during the course of NEC. In the best-constructed model, lower HB levels tend to correspond to more severe staging of NEC. Anemia is a prevalent comorbidity in preterm infants[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], and NEC, as an intestinal ischemic disease[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e], is closely associated with anemia throughout the disease's progression. The presence of intestinal necrosis, bloody stools, and bloody peritoneal fluid leads to decrease hemoglobin level. Furthermore, the induced inflammatory response affects the production of hemoglobin and the patient's survival[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. In a case-control study[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e], the researchers compared splanchnic tissue oxygen saturation, extraction, and variability of splanchnic oxygen saturation in NEC patients with different hemoglobin levels, demonstrated that low hemoglobin levels were associated with intestinal injury. This may accelerate the development of NEC. In the predictive model we selected as the best, low levels of hemoglobin were a significant point of difference between early NEC and normal controls.\u003c/p\u003e \u003cp\u003eIn terms of model constructed, logistic regression models were also effective in in predicting disease turnover in high-risk patients. In fact, there are some models that focus on distinguishing whether NEC occurs. Julia M. Pantalone et al.[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e] analyzed complete blood counts (CBCs) around illness onset in NEC cases compared with controls, and developed LDA, SVM, RF, and Logistic regression models. RF had the best performance with an AUC of 0.877, sensitivity 0.8, and specificity 0.793 for the comparison of S-NEC vs. Controls. Although RF was the best performer among all models in the comparison of S-NEC vs. M-NEC, its AUC value was still below 0.8 and had low sensitivity among all models. For intestinal perforation due to severe NEC, Irles et al.[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e] developed a back-propagation artificial neural network (ANN) based on critical variables such as neonatal platelet and neutrophil levels, birthweight, sex, gestational age and maternal age. Back-propagation ANN developed two estimation models for predicting intestinal perforation caused by NEC. The application of machine learning models for early prediction of NEC has not been widely adopted, and there are currently no machine learning models capable of predicting early NEC risk based on basic data.\u003c/p\u003e \u003cp\u003eIn this study, the most frequently utilized test metrics, along with general clinical characteristics, were employed to develop prevalent machine learning models for predicting the beginning of NEC. When comparing machine learning algorithms, LightGBM and the radial basis function kernel of SVM both demonstrated excellent predictive performance for NEC in the training set and test set. By comparison, the LightGBM model proved to be the most effective in integrating each assessment metric. The model identified several risk factors for the development of NEC in neonates, including day age, gestational age, EOS, HB, PDW and HSCRP, with day age being the most critical predictor of NEC. Accurate risk factor assessment facilitates the identification of high-risk patients. In our study, we found that gestational age combined with day age at onset was highly predictive of NEC, similar to previous research[\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. The smaller the gestational age, the greater the SHAP contribution value. This indicates that preterm infants are highly susceptible to NEC around seven days after birth. Therefore, close attention should be given at this time.\u003c/p\u003e \u003cp\u003eOur study used ML algorithms to predict NEC in early neonatal life. However, there are some limitations. It was a two-center study and the amount of sample was limited, validation of large databases is needed. In future, we will continue to refine and validate the existing machine learning model using large-scale, multicenter population data.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe LightGBM model, which considers day age, gestational age, EOS, HB, PDW, and HSCRP, demonstrates good performance in predicting NEC in the early stage. Clinicians can use this model to identify highrisk newborns requiring focused attention, enhance observation of their clinical manifestations, clarify diagnoses through imaging examinations .\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eBASO\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ebasophil\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eDCA\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDecision curve analysis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eEOS\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eEosinophil\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eHB\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHemoglobin\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eHSCRP\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHigh-sensitivity C-reactive protein\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eLASSO\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLeast Absolute Shrinkage and Selection Operator\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eLightGBM\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLight Gradient Boosting Machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eLYM\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003elymphocyte\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eML\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emachine learning\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eMONO\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emonocyte\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eMPV\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003emean platelet volume\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eNEC\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003enecrotizing enterocolitis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eNEU\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eneutrophil\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003ePDW\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eplatelet distribution width\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003ePLT\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eplatelet\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003ePCT\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eplateletcrit\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eRBC\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ered blood cell\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eRF\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRandom Forest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eROC\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ereceiver operating characteristic curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eSHAP\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSHapely Additive exPlanation\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eSVM\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSupport Vector Machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eWBC\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ewhite blood cell\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e \u003cp\u003e Research have been performed in accordance with the Declaration of Helsinki. The participants involved in this study were reviewed and approved by Ethics Committee of Zhujiang Hospital of Southern Medical University (2024-KY-123-01), ensuring that all ethical standards and requirements have been met. Informed consent was obtained from all subjects and/or their legal guardians for this study.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent for publication\u003c/strong\u003e \u003cp\u003e Written informed consent was obtained from the parents.\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe 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.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis study was supported by the Guangdong Basic and Applied Basic Research Foundation of China (2019A1515011086) and Clinical research Fund of Guangdong Medical Association(A202302026).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAll authors contributed to the study conception and design. Yanling Mou prepared the manuscrip and collected data. Jinhao Li summarized and analyzed data. Jianjun Wang, Daiyue Yu and Huirong Yang conceived and designed the experiments. Xi Zhang, Rongying Tan and Djibril Adam Mahama made contributions to acquisition of data and interpretation of data. Liucheng Yang and Kai Wu guided the project and revised the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eNot applicable\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eDue to\u0026nbsp;strict ethical and legal guidelines that govern the handling of patient data, the datasets generated and analysed during the current study are not publicly available but are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eCao Y, Jiang S, Sun J, Hei M, Wang L, Zhang H, Ma X, Wu H, Li X, Sun H, Zhou W, Shi Y, Wang Y, Gu X, Yang T, Lu Y, Du L, Chen C, Lee SK, Zhou W, et al. Assessment of Neonatal Intensive Care Unit Practices, Morbidity, and Mortality Among Very Preterm Infants in China. JAMA Netw open. 2021;4(8):e2118904. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1001/jamanetworkopen.2021.18904\u003c/span\u003e\u003cspan address=\"10.1001/jamanetworkopen.2021.18904\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThompson AM, Bizzarro MJ. Necrotizing enterocolitis in newborns: pathogenesis, prevention and management. Drugs. 2008;68(9):1227\u0026ndash;38. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2165/00003495-200868090-00004\u003c/span\u003e\u003cspan address=\"10.2165/00003495-200868090-00004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYee WH, Soraisham AS, Shah VS, Aziz K, Yoon W, Lee SK, Canadian Neonatal Network. Incidence and timing of presentation of necrotizing enterocolitis in preterm infants. Pediatrics. 2012;129(2):e298\u0026ndash;304. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1542/peds.2011-2022\u003c/span\u003e\u003cspan address=\"10.1542/peds.2011-2022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePicaud JC, De Magistris A, Mussap M, Corbu S, Dess\u0026igrave; A, Noto A, Fanos V, Cesare Marincola F. Urine NMR Metabolomics Profile of Preterm Infants With Necrotizing Enterocolitis Over the First Two Months of Life: A Pilot Longitudinal Case-Control Study. Front Mol Biosci. 2021;8:680159. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fmolb.2021.680159\u003c/span\u003e\u003cspan address=\"10.3389/fmolb.2021.680159\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVitali R, Terrin G, Palone F, Laudadio I, Cucchiara S, Boscarino G, Di Chiara M, Stronati L. Fecal High-Mobility Group Box 1 as a Marker of Early Stage of Necrotizing Enterocolitis in Preterm Neonates. Front Pead. 2021;9:672131. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fped.2021.672131\u003c/span\u003e\u003cspan address=\"10.3389/fped.2021.672131\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu XC, Li LQ, Ling KR, Guo L, Hu XY, Li C. Fecal HBD-2 and Claudin-3 may be potential biomarkers to predict the deterioration of necrotizing enterocolitis: A prospective study. Front Pead. 2022;10:1062798. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fped.2022.1062798\u003c/span\u003e\u003cspan address=\"10.3389/fped.2022.1062798\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen G, Lv X, Tang W. Fecal calprotectin as a non-invasive marker for the prediction of post-necrotizing enterocolitis stricture. Pediatr Surg Int. 2023;39(1):250. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00383-023-05534-7\u003c/span\u003e\u003cspan address=\"10.1007/s00383-023-05534-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoschino L, Verlato G, Duci M, Cavicchiolo ME, Guiducci S, Stocchero M, Giordano G, Fascetti Leon F, Baraldi E. The Metabolome and the Gut Microbiota for the Prediction of Necrotizing Enterocolitis and Spontaneous Intestinal Perforation. Syst Rev Nutrients. 2022;14(18):3859. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/nu14183859\u003c/span\u003e\u003cspan address=\"10.3390/nu14183859\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBell MJ, Ternberg JL, Feigin RD, Keating JP, Marshall R, Barton L, Brotherton T. Neonatal necrotizing enterocolitis. Therapeutic decisions based upon clinical staging. Ann Surg. 1978;187(1):1\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/00000658-197801000-00001\u003c/span\u003e\u003cspan address=\"10.1097/00000658-197801000-00001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHerzlich J, Mandel D, Marom R, Mendelsohn R, Eshel Fuhrer A, Mangel L. Blood Glucose, Lactate and Platelet Count in Infants with Spontaneous Intestinal Perforation versus Necrotizing Enterocolitis-A Pilot Study. Child (Basel Switzerland). 2023;10(6):1028. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/children10061028\u003c/span\u003e\u003cspan address=\"10.3390/children10061028\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng J, Wang H. Association Between Platelet-to-Lymphocyte Ratio and Preterm Necrotizing Enterocolitis. Front Pead. 2021;9:686880. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fped.2021.686880\u003c/span\u003e\u003cspan address=\"10.3389/fped.2021.686880\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng W, Hou J, Die X, Sun J, Guo Z, Liu W, Wang Y. Application of coagulation parameters at the time of necrotizing enterocolitis diagnosis in surgical intervention and prognosis. BMC Pediatr. 2022;22(1):259. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12887-022-03333-y\u003c/span\u003e\u003cspan address=\"10.1186/s12887-022-03333-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrishnan R, Rajpurkar P, Topol EJ. Self-supervised learning in medicine and healthcare. Nat biomedical Eng. 2022;6(12):1346\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41551-022-00914-1\u003c/span\u003e\u003cspan address=\"10.1038/s41551-022-00914-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSubspecialty Group of Neonatology, the Society of Pediatrics, Chinese Medical Association, \u0026amp;, Editorial, Board, Chinese Journal of Pediatrics. Zhonghua er ke za zhi\u0026thinsp;=\u0026thinsp;Chinese. J Pediatr. 2023;61(10):869\u0026ndash;73. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3760/cma.j.cn112140-20230630-00434\u003c/span\u003e\u003cspan address=\"10.3760/cma.j.cn112140-20230630-00434\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKordasz M, Racine M, Szavay P, Lehner M, Krebs T, Luckert C, Hau EM, Berger S, Kessler U. Risk factors for mortality in preterm infants with necrotizing enterocolitis: a retrospective multicenter analysis. Eur J Pediatrics. 2022;181(3):933\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00431-021-04266-x\u003c/span\u003e\u003cspan address=\"10.1007/s00431-021-04266-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeister AL, Doheny KK, Travagli RA, Maywood NJ.), 245(2), 85\u0026ndash;95. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1535370219891971\u003c/span\u003e\u003cspan address=\"10.1177/1535370219891971\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHumberg A, Fortmann I, Siller B, Kopp MV, Herting E, G\u0026ouml;pel W, H\u0026auml;rtel C, German Neonatal Network, German Center for Lung Research and Priming Immunity at the beginning of life (PRIMAL) Consortium. Preterm birth and sustained inflammation: consequences for the neonate. Semin Immunopathol. 2020;42(4):451\u0026ndash;68. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00281-020-00803-2\u003c/span\u003e\u003cspan address=\"10.1007/s00281-020-00803-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErnst E, Hammerschmidt DE, Bagge U, Matrai A, Dormandy JA. Leukocytes and the risk of ischemic diseases. JAMA. 1987;257(17):2318\u0026ndash;24.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampbell EL, Kao DJ, Colgan SP. Neutrophils and the inflammatory tissue microenvironment in the mucosa. Immunol Rev. 2016;273(1):112\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/imr.12456\u003c/span\u003e\u003cspan address=\"10.1111/imr.12456\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHealy DB, Ryan CA, Ross RP, Stanton C, Dempsey EM. Clinical implications of preterm infant gut microbiome development. Nat Microbiol. 2022;7(1):22\u0026ndash;33. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41564-021-01025-4\u003c/span\u003e\u003cspan address=\"10.1038/s41564-021-01025-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaplina A, Kononova S, Zaikova E, Pervunina T, Petrova N, Sitkin S. Necrotizing Enterocolitis: The Role of Hypoxia, Gut Microbiome, and Microbial Metabolites. Int J Mol Sci. 2023;24(3):2471. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijms24032471\u003c/span\u003e\u003cspan address=\"10.3390/ijms24032471\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKorpela K, Blakstad EW, Moltu SJ, Str\u0026oslash;mmen K, Nakstad B, R\u0026oslash;nnestad AE, Br\u0026aelig;kke K, Iversen PO, Drevon CA, de Vos W. Intestinal microbiota development and gestational age in preterm neonates. Sci Rep. 2018;8(1):2453. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-018-20827-x\u003c/span\u003e\u003cspan address=\"10.1038/s41598-018-20827-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJones IH, Hall NJ. Contemporary Outcomes for Infants with Necrotizing Enterocolitis-A Systematic Review. J Pediatr. 2020;220:86\u0026ndash;e923. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jpeds.2019.11.011\u003c/span\u003e\u003cspan address=\"10.1016/j.jpeds.2019.11.011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSamuels N, van de Graaf RA, de Jonge RCJ, Reiss IKM, Vermeulen MJ. Risk factors for necrotizing enterocolitis in neonates: a systematic review of prognostic studies. BMC Pediatr. 2017;17(1):105. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12887-017-0847-3\u003c/span\u003e\u003cspan address=\"10.1186/s12887-017-0847-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeane M, Fallon R, Riordan A, Shaw B. (2015) Markedly raised levels of C-reactive protein are associated with culture-proven sepsis or necrotising enterocolitis in extremely preterm neonates. Acta paediatrica (Oslo, Norway: 1992), 104(7), e289\u0026ndash;e293. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/apa.12978\u003c/span\u003e\u003cspan address=\"10.1111/apa.12978\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZani A, Eaton S, Puri P, Rintala R, Lukac M, Bagolan P, Kuebler JF, Hoellwarth ME,Wijnen R, Tovar J, Pierro A, \u0026amp; EUPSA Network (2015) International survey on the management of necrotizing enterocolitis. European journal of pediatric surgery: official journal of Austrian Association of Pediatric Surgery \u0026hellip; et al]\u0026thinsp;=\u0026thinsp;Zeitschrift fur Kinderchirurgie,25(1), 27\u0026ndash;33. https://doi.org/10.1055/s-0034-1387942.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGo H, Ohto H, Nollet KE, Takano S, Kashiwabara N, Chishiki M, Maeda H, Imamura T, Kawasaki Y, Momoi N, Hosoya M. Using Platelet Parameters to Anticipate Morbidity and Mortality Among Preterm Neonates: A Retrospective Study. Front Pead. 2020;8:90. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fped.2020.00090\u003c/span\u003e\u003cspan address=\"10.3389/fped.2020.00090\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoupenova M, Clancy L, Corkrey HA, Freedman JE. Circulating Platelets as Mediators of Immunity, Inflammation, and Thrombosis. Circul Res. 2018;122(2):337\u0026ndash;51. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1161/CIRCRESAHA.117.310795\u003c/span\u003e\u003cspan address=\"10.1161/CIRCRESAHA.117.310795\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan der Meijden PEJ, Heemskerk JWM. Platelet biology and functions: new concepts and clinical perspectives. Nat Rev Cardiol. 2019;16(3):166\u0026ndash;79. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41569-018-0110-0\u003c/span\u003e\u003cspan address=\"10.1038/s41569-018-0110-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGurtner A, Borrelli C, Gonzalez-Perez I, Bach K, Acar IE, N\u0026uacute;\u0026ntilde;ez NG, Crepaz D, Handler K, Vu VP, Lafzi A, Stirm K, Raju D, Gschwend J, Basler K, Schneider C, Slack E, Valenta T, Becher B, Krebs P, Moor AE, \u0026hellip; Arnold I C (2023) Active eosinophils regulate host defence and immune responses in colitis. Nature, 615(7950), 151\u0026ndash;157. https://doi.org/10.1038/s41586-022-05628-7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChristensen RD, Lambert DK, Gordon PV, Baer VL, Gerday E, Henry E. Neonates presenting with bloody stools and eosinophilia can progress to two different types of necrotizing enterocolitis. J perinatology: official J Calif Perinat Association. 2012;32(11):874\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/jp.2011.163\u003c/span\u003e\u003cspan address=\"10.1038/jp.2011.163\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWahidi LS, Sherman J, Miller MM, Zaghouani H, Sherman MP. Early Persistent Blood Eosinophilia in Necrotizing Enterocolitis Is a Predictor of Late Complications. Neonatology. 2015;108(2):137\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1159/000431305\u003c/span\u003e\u003cspan address=\"10.1159/000431305\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZdravic D, Yougbare I, Vadasz B, Li C, Marshall AH, Chen P, Kjeldsen-Kragh J, Ni H. Fetal and neonatal alloimmune thrombocytopenia. Semin Fetal Neonatal Med. 2016;21(1):19\u0026ndash;27. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.siny.2015.12.004\u003c/span\u003e\u003cspan address=\"10.1016/j.siny.2015.12.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePatel RM, Knezevic A, Shenvi N, Hinkes M, Keene S, Roback JD, Easley KA, Josephson CD. Association of Red Blood Cell Transfusion, Anemia, and Necrotizing Enterocolitis in Very Low-Birth-Weight Infants. JAMA. 2016;315(9):889\u0026ndash;97. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1001/jama.2016.1204\u003c/span\u003e\u003cspan address=\"10.1001/jama.2016.1204\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoberts AG, Younge N, Greenberg RG. Neonatal Necrotizing Enterocolitis: An Update on Pathophysiology, Treatment, and Prevention. Paediatr Drugs. 2024;26(3):259\u0026ndash;75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40272-024-00626-w\u003c/span\u003e\u003cspan address=\"10.1007/s40272-024-00626-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGanz T. Anemia of Inflammation. N Engl J Med. 2019;381(12):1148\u0026ndash;57. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1056/NEJMra1804281\u003c/span\u003e\u003cspan address=\"10.1056/NEJMra1804281\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalteren WS, Bos AF, van Oeveren W, Hulscher JBF, Kooi EMW. Neonatal anemia relates to intestinal injury in preterm infants. Pediatr Res. 2022;91(6):1452\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41390-021-01903-x\u003c/span\u003e\u003cspan address=\"10.1038/s41390-021-01903-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePantalone JM, Liu S, Olaloye OO, Prochaska EC, Yanowitz T, Riley MM, Buland JR, Brozanski BS, Good M, Konnikova L. Gestational Age-Specific Complete Blood Count Signatures in Necrotizing Enterocolitis. Front Pead. 2021;9:604899. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fped.2021.604899\u003c/span\u003e\u003cspan address=\"10.3389/fped.2021.604899\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIrles C, Gonz\u0026aacute;lez-P\u0026eacute;rez G, Carrera Mui\u0026ntilde;os S, Michel Macias C, S\u0026aacute;nchez G\u0026oacute;mez C, Mart\u0026iacute;nez-Zepeda A, Cordero Gonz\u0026aacute;lez G, Laresgoiti Servitje E. Estimation of Neonatal Intestinal Perforation Associated with Necrotizing Enterocolitis by Machine Learning Reveals New Key Factors. Int J Environ Res Public Health. 2018;15(11):2509. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijerph15112509\u003c/span\u003e\u003cspan address=\"10.3390/ijerph15112509\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSharma R, Hudak ML, Tepas JJ 3rd, Wludyka PS, Marvin WJ, Bradshaw JA, Pieper P. Impact of gestational age on the clinical presentation and surgical outcome of necrotizing enterocolitis. J perinatology: official J Calif Perinat Association. 2006;26(6):342\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/sj.jp.7211510\u003c/span\u003e\u003cspan address=\"10.1038/sj.jp.7211510\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"machine learning, necrotizing enterocolitis, diagnose, prediction","lastPublishedDoi":"10.21203/rs.3.rs-4556691/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4556691/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e: To improve the prognosis of necrotizing enterocolitis (NEC) in newborns, early identification and timely preventive interventions play an essential role. Based on the current situation, establishing a novel and simple prediction model is of great clinical significance.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e: The clinical data of NEC neonates in Zhujiang Hospital of Southern Medical University from October 2010 to October 2022 were collected, and 429 non-NEC patients in the same period were selected as the control group by random sampling method. After that, all participants were randomly divided into training group (70%) and testing group (30%). Combining relevant clinical features and laboratory results, five machine learning (ML) algorithms and classical logistic regression models were established. To evaluate the performance of each model, the area under the receiver operating characteristic curve (ROC), accuracy, sensitivity, and specificity of various models were compared. 10-folds cross-validation was used to find the best hyperparameters for each model. Decision curve analysis (DCA) was further used to evaluate the performance of the established models for clinical applications, and create a column-line graph, ranking the feature importance in model by SHapely Additive exPlanation (SHAP). The column plots were calibrated using calibration curves. In addition, the established model was validated in time series analysis as well as in another medical center.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: Six important features were finally included for modeling, including the Day (OR=1.15; 95% CI: 1.07-1.23; \u003cem\u003eP\u003c/em\u003e=0.001), Gestational age (OR=0.77; 95% CI: 0.62-0.95; \u003cem\u003eP\u003c/em\u003e=0.016), Eosinophil (EOS) (OR=3.76; 95% CI: 1.76-8.02; \u003cem\u003eP\u003c/em\u003e=0.001), Hemoglobin (HB) (OR=0.98; 95% CI: 0.97-1.00; \u003cem\u003eP\u003c/em\u003e=0.011), Platelet distribution width (PDW) (OR=1.21; 95% CI: 1.08-1.35; \u003cem\u003eP\u003c/em\u003e=0.001) and High-sensitivity C-reactive protein (HSCRP) (OR=1.03; 95% CI: 1.01-1.06; \u003cem\u003eP\u003c/em\u003e=0.007). While the logistic regression model achieved an AUC of 0.919, accuracy of 0.897, sensitivity of 0.832, F1-score of 0.778, and a Brier score of 0.0878 in the training group, the AUCs for the five machine learning models ranged from 0.774 to 0.972. Among these models, the LightGBM model performed the best, with an AUC of 0.960, accuracy of 0.894, sensitivity of 0.901, F1-score of 0.813, and a Brier score of 0.072.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e: The LightGBM machine learning model can effectively identify neonatal patients at higher risk of NEC based on Day age, Gestational age, EOS, HB, PDW, and HSCRP levels. This model is useful for assisting in clinical decision-making.\u003c/p\u003e","manuscriptTitle":"An Early Prediction of Neonatal Necrotizing Enterocolitis in High-Risk Newborns- Based on Two Medical Center Clinical Databases","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-03 16:16:20","doi":"10.21203/rs.3.rs-4556691/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ca879bbe-2912-4710-a666-3836d6c824ab","owner":[],"postedDate":"July 3rd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-12-31T08:23:23+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-03 16:16:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4556691","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4556691","identity":"rs-4556691","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}