Newborn early-onset septicemia risk prediction method and system based on maternal factors

Abstract

The invention discloses a neonatal early-onset septicemia risk prediction method and system based on maternal factors, and the method comprises the steps: firstly collecting maternal data, including maternal clinical and laboratory data; carrying out variable processing and screening on the maternal data to obtain significant variables related to the risk of neonatal early-onset septicemia; establishing a multivariable statistical prediction model based on the significant variables; and outputting a risk score of the neonatal early-onset septicemia by using the prediction model, wherein the risk score is used for individualized risk assessment. The method has the advantages that multivariate analysis can be integrated to assist clinical decision making, and the EOS risk can be predicted individually.

Description

Technical Field

[0001] This invention relates to the field of medical and health technology, specifically to a method and system for predicting the risk of early-onset sepsis (EOS) in newborns based on maternal clinical and laboratory indicators. Background Technology

[0002] Neonatal sepsis is one of the major threats to neonatal survival. Based on the time of onset, neonatal sepsis is generally divided into early-onset sepsis (EOS, occurring within 72 hours of birth) and late-onset sepsis (LOS). EOS is mainly caused by vertical transmission from mother to newborn during the perinatal period. Common pathogens are bacteria, especially Gram-negative bacteria such as Group B Streptococcus and Escherichia coli. EOS usually has a rapid onset, often accompanied by pneumonia and multi-organ involvement, presenting with nonspecific symptoms and insidious clinical signs, leading to delayed diagnosis and thus increasing the risk of adverse outcomes.

[0003] Due to the high mortality rate of neonatal sepsis (EOS), early diagnosis and timely treatment are crucial for improving neonatal survival. Given the significant physiological and clinical differences between newborns and older children, diagnostic criteria and predictive indicators for neonatal sepsis are constantly evolving. Currently, clinical assessment of neonatal inflammatory status mainly relies on white blood cell count and neutrophil percentage. However, these indicators primarily reflect existing inflammation and disease progression, have limited ability to predict the occurrence of sepsis, and involve invasive procedures, hindering early non-invasive identification.

[0004] In clinical practice, the prediction of neonatal early-onset sepsis (EOS) requires a comprehensive assessment of multiple factors, including clinical manifestations, laboratory tests, and maternal and neonatal related variables, in order to achieve early intervention. Therefore, there is an urgent need for a method and system for predicting the risk of neonatal early-onset sepsis that can reduce invasive procedures on newborns and achieve early identification of EOS risk. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is: how to provide a method and system for predicting the risk of early-onset neonatal sepsis that can integrate multivariate analysis to assist clinical decision-making and improve the survival rate of newborns.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A method for predicting the risk of early-onset neonatal sepsis based on maternal factors includes the following steps: S1. Collect maternal data, including maternal clinical and laboratory data; S2. Perform variable processing and screening on the maternal data to obtain significant variables related to the risk of early-onset neonatal sepsis. S3. Establish a multivariate statistical prediction model based on the significant variables; S4. Use the prediction model to output a risk score for early-onset neonatal sepsis for individualized risk assessment.

[0007] Furthermore, the maternal data indicators include at least one of the following: history of pregnancy-related diseases, mode of delivery, multiple pregnancy, preterm birth, intrauterine fetal distress, placental abnormalities, intrahepatic cholestasis of pregnancy, chorioamnionitis, puerperal infection, cervical insufficiency, white blood cell count, monocyte percentage, platelet-to-white blood cell ratio, hemoglobin glycated index, and globulin.

[0008] Furthermore, in step S2, for categorical variables, univariate analysis is used to identify maternal risk factors associated with early-onset neonatal sepsis as significant categorical variables.

[0009] Furthermore, for continuous variables, the optimal cutoff value was determined by ROC curve analysis and then divided, followed by a chi-square test to determine the ROC continuous variable. At the same time, RCS analysis was used to assess the linear or nonlinear relationship between the continuous variable and the risk of early-onset neonatal sepsis to determine the RCS continuous variable. Venn diagrams were used to visualize the ROC and RCS continuous variables, and the continuous variables in the overlapping part were identified as significant continuous variables.

[0010] Preferably, in step S3, variables that differ between significant categorical variables and significant continuous variables are merged, and the variance inflation factor and tolerance are used to assess collinearity between variables, and variables with a variance inflation factor greater than 10 or a tolerance less than 0.1 are removed.

[0011] Furthermore, in step S3, a multivariate Poisson regression model is used to establish a prediction model, and confounding factors are adjusted to calculate the event occurrence ratio and its confidence interval. The confounding factors include at least the maternal age and the year of neonatal admission.

[0012] Furthermore, in step S3, the model performance is evaluated by the area under the receiver operating characteristic curve, and the model stability is verified by Bootstrap resampling.

[0013] Furthermore, in step S3, the independent predictors of the prediction model include at least one of the following: chorioamnionitis, puerperal infection, preterm birth, intrauterine fetal distress, multiple pregnancy, elevated monocyte percentage, elevated platelet-to-white blood cell ratio, elevated hemoglobin glycated index, decreased globulin, cesarean section as the mode of delivery, cervical insufficiency, placental abnormalities, and intrahepatic cholestasis of pregnancy.

[0014] A risk prediction system for early-onset neonatal sepsis based on maternal factors, comprising: The data acquisition module is used to collect clinical and laboratory data from the mother. The data processing and filtering module is used to process and filter the maternal data to obtain significant variables related to the risk of neonatal early-onset sepsis. The modeling module is used to build a multivariate statistical prediction model based on the significant variables; The risk scoring module is used to output a risk score for early-onset neonatal sepsis using the prediction model and to conduct individualized risk assessment.

[0015] Furthermore, the data acquisition module is used to collect at least one maternal data indicator from the following: pregnancy history, delivery method, multiple pregnancy, premature birth, intrauterine fetal distress, placental abnormalities, intrahepatic cholestasis of pregnancy, chorioamnionitis, puerperal infection, cervical insufficiency, white blood cell count, monocyte percentage, platelet-to-white blood cell ratio, hemoglobin glycated index, and globulin. The modeling module includes a multivariate Poisson regression modeling unit, which is used to build a predictive model after removing highly collinear variables, adjust for maternal age and neonatal admission year, calculate the event incidence ratio and its confidence interval, evaluate model performance by the area under the receiver operating characteristic curve, and verify model stability by Bootstrap resampling. The risk scoring module is used to construct a nodal chart based on the final model and generate a risk score of 0–100.

[0016] Compared with the prior art, the present invention has the following advantages: 1. Non-invasive: The model is based entirely on maternal indicators, requiring no invasive procedures on newborns, and is easily accepted by patients.

[0017] 2. Prenatal prediction: Indicators can be obtained before delivery, supporting early risk assessment and intervention.

[0018] 3. Innovative Indicators: The HGI and PWR indicators are introduced for the first time to reveal their correlation with EOS risk and enhance predictive value. Attached Figure Description

[0019] Figure 1 A schematic diagram illustrating the process of incorporating maternal and infant data.

[0020] Figure 2 A graph showing the relationship between the parent categorical variable and EOS risk.

[0021] Figures 3-8 Restricted cubic spline plot for continuous variables.

[0022] Figure 9 and Figure 10To investigate the relationship between maternal factors and the risk of early-onset neonatal sepsis (EOS).

[0023] Figure 11 and Figure 12 The images show the ROC curve and noctilinear plot of a multivariate Poisson regression model.

[0024] Figure 13 For model performance evaluation. Detailed Implementation

[0025] The present invention will be further described in detail below with reference to the embodiments.

[0026] A method for predicting the risk of early-onset neonatal sepsis based on maternal factors is proposed. Based on the applicant's multicenter retrospective cohort study and risk prediction model project, this multicenter retrospective cohort study collected clinical data of neonates and their mothers hospitalized in the Neonatal Department of Chongqing Maternal and Child Health Center and Yunyang Maternal and Child Health Hospital from January 2017 to December 2023.

[0027] The data inclusion criteria are as follows: 1. Newborns ≤72 hours old must be admitted to the hospital; 2. The mother's pregnancy and delivery records are complete; 3. Newborns diagnosed with EOS, or as a control group (no infection symptoms during hospitalization and not diagnosed with infection).

[0028] The diagnosis of EOS is based on the "Expert Consensus on the Diagnosis and Treatment of Neonatal Sepsis".

[0029] The data exclusion criteria are as follows: 1. Congenital malformations or chromosomal abnormalities in newborns; 2. Key clinical or laboratory data are missing; 3. Infant infection is clearly not of maternal origin, such as hospital-acquired infection; 4. Artificial termination of pregnancy or non-live birth of newborns.

[0030] Data collection involved extracting clinical and laboratory data from the hospital's electronic medical record (HIS) and laboratory information system (LIS). Neonatal data included sex, multiple pregnancy, preterm birth, macrosomia, and signs of distress. Maternal data included pregnancy history (gestational diabetes, gestational hypertension, intrahepatic cholestasis of pregnancy, etc.), delivery method, and prenatal infection status. Maternal laboratory parameters were obtained from the initial tests upon admission, including complete blood count (white blood cells, hemoglobin, platelets, etc.), liver function tests (globulins, total bilirubin, etc.), and kidney function tests (creatinine, urea, etc.). All data were independently extracted and verified by two trained researchers to ensure completeness and accuracy.

[0031] Calculation of derived indicators: Commonly used blood and biochemical derived indicators are calculated to assess the relationship between maternal inflammation and metabolic status and neonatal eosinophilic disease (EOS) risk. Derived indicators include: Platelet-to-white blood cell ratio (PWR) = PLT / WBC Neutrophil-to-lymphocyte ratio (NLR) = N# / L# Monocyte to lymphocyte ratio (MLR) = M# / L# Neutrophil-to-monocyte ratio to lymphocyte ratio (NMLR) = (M# + N#) / L# Systemic Inflammatory Response Index (SIRI) = (N# × M#) / L# Systemic Immune Inflammatory Index (SII) = (PLT × N#) / L# Derived neutrophil to lymphocyte ratio (DNLR) = N# / (WBC − L#) The aspartate aminotransferase to alanine aminotransferase ratio (SLR) = AST / ALT Blood urea nitrogen to creatinine ratio (UCR) = UREA / CR Glycated hemoglobin index (HGI) = HbA1c − Pre_HbA1c Prognostic Nutritional Index (PNI) = 10 × ALB + 0.005 × L# Statistical Analysis: Statistical analysis was performed using R software (v4.5.0). Normality of continuous variables was assessed using the Shapiro-Wilk test. Normally distributed data were expressed as mean ± standard deviation (Mean ± SD) and analyzed using the independent samples t-test; non-normally distributed data were expressed as median (IQR) and analyzed using the Mann–Whitney U test. Categorical variables were expressed as frequency and percentage, and comparisons between groups were performed using the chi-square test or Fisher's exact test.

[0032] Univariate analysis was used to screen for maternal risk factors associated with EOS. For categorical variables, the chi-square test was used to assess EOS risk; for continuous variables, two methods were employed for variable selection: 1. Determining the optimal cutoff value based on the ROC curve, and then performing a chi-square test after dichotomizing the continuous variable; 2. Using RCS analysis to assess the linear or non-linear relationship between the continuous variable and EOS. Significant variables were visualized using Venn diagrams, and intersection variables were included in the multivariate model to improve stability and interpretability.

[0033] Before multivariate regression, variance inflation factor (VIF) and tolerance were used to assess collinearity. A VIF > 10 or a tolerance < 0.1 indicated severe collinearity, requiring variable exclusion or consolidation based on clinical and statistical principles. Univariate significant variables were included in the multivariate Poisson regression model to calculate the event incidence ratio (IRR) and 95% CI. Model performance was assessed using the area under the ROC curve (AUC).

[0034] An initial model (Model 1) and a final model (Model 2) were constructed. To assess the stability and overfitting risk of the final model, logistic regression was used to verify the consistency of the Poisson regression results. Model 2 underwent 1,000 bootstrap resampling iterations, with each iteration reconstructing the model and calculating AUC, Brier score, calibration intercept, and slope to estimate the model's optimism bias. A nomogram was constructed based on the final Poisson regression model to achieve individualized risk prediction. Data visualization was performed using R 4.5.0 and Origin 2024 software. All tests were two-tailed, with P < 0.05 considered statistically significant.

[0035] From 2017 to 2024, this study included 111,993 pregnant women, of whom 79,570 were successfully matched with corresponding newborn records. Figure 1 As shown in Table 1, the overall incidence of EOS was 0.56% (449 / 79,570). There were no statistically significant differences between the EOS-positive and EOS-negative groups in maternal age, pre-pregnancy weight, height, pre-pregnancy body mass index (BMI), premature rupture of membranes (PROM), threatened preterm labor (TPL), fetal growth restriction (FGR), abnormal obstetric history (AOH), smoking history, and alcohol consumption history (ACH). Compared with the control group, the maternal levels of ALB, PNI, prealbumin (PA), MCV, A / G ratio, and SLR were decreased in neonates with EOS, while MLR, SII, GGT, NMLR, and NLR were increased, all with statistically significant differences (all P < 0.05). To further explore the relationship between maternal age and the risk of eosinophilia (EOS), maternal age was divided into five groups: 25, 25–29, 30–34, 35–39, and ≥40 years old. The 25–29 age group (0.60%) and the 30–34 age group (0.61%) had the highest incidence of EOS, and the differences were statistically significant (P<0.001). Figure 9 The incidence of EOS is shown for maternal age groups. The bar chart illustrates the proportion of EOS cases in different maternal age categories (40 years). The highest risk of EOS was observed in newborns born to mothers aged 25-34 years.

[0036] Table 1 Basic parameters of continuous variables variable Control group (mean ± standard deviation) Control group, N EOS (mean ± standard deviation) EOS, N P ALB 34.77 ± 2.78 79141 34.1 ± 3.2 429 <0.001 UCR 0.08 ± 0.02 79141 0.08 ± 0.02 429 <0.001 PNI 347.67 ± 27.84 79141 341.06 ± 32.02 429 <0.001 PA 210.79 ± 34.62 79141 201.79 ± 40.21 429 <0.001 MCV 91.77 ± 6.1 79141 90.65 ± 6.18 429 <0.001 RBC 3.89 ± 0.45 79141 3.96 ± 0.47 429 <0.001 A / G 1.18 ± 0.38 79141 1.16 ± 0.3 429 0.002 L# 1.55 ± 0.48 79141 1.47 ± 0.46 429 0.002 MLR 0.38 ± 0.18 79141 0.43 ± 0.28 429 0.002 UREA 3.72 ± 1.96 79141 3.65 ± 2.7 429 0.002 SII 1041.86 ± 696.97 79141 1213.21 ± 974.6 429 0.003 MCH 31.04 ± 2.58 79141 30.73 ± 2.6 429 0.008 Pre-pregnancy weight 54.26 ± 8.79 71268 55.06 ± 8.13 376 0.011 Newborn weight 3231.08 ± 491.16 67677 3024.21 ± 794.78 385 0.013 GGT 16.47 ± 16.5 79141 18.01 ± 16.36 429 0.023 L% 16.05 ± 5.56 79141 15.32 ± 6.21 429 0.024 NMLR 6.21 ± 3.77 79141 7.29 ± 5.68 429 0.026 SLR 1.5 ± 0.66 79141 1.46 ± 0.65 429 0.033 NLR 5.83 ± 3.63 79141 6.86 ± 5.44 429 0.033 TP 64.85 ± 4.86 79141 64.27 ± 5.81 429 0.033 IBIL 7.38 ± 2.6 79141 7.22 ± 2.99 429 0.033 HGB 120.09 ± 13.55 79141 121.03 ± 14.45 429 0.038 ALP 159.53 ± 64.15 79141 155.47 ± 62.04 429 0.045 SIRI 3.36 ± 3.11 79141 4.47 ± 6 429 0.051 M# 0.55 ± 0.19 79141 0.57 ± 0.22 429 0.054 RDWSD 46.13 ± 4.02 79141 45.76 ± 3.93 429 0.070 LDH 177.2 ± 39.15 79141 182.45 ± 48.61 429 0.071 M% 5.42 ± 1.36 79141 5.5 ± 1.49 429 0.076 HCT 35.51 ± 3.63 79141 35.71 ± 3.9 429 0.080 Preconception BMI 21.44 ± 3.26 64093 21.63 ± 2.89 324 0.091 N% 77.68 ± 6.4 79141 78.33 ± 7.23 429 0.095 TBIL 9.22 ± 3.32 79141 9.14 ± 3.76 429 0.115 CYSC 1.27 ± 0.3 79141 1.29 ± 0.32 429 0.145 RDW 13.91 ± 1.26 79141 14.02 ± 1.36 429 0.147 ALT 18.23 ± 29.39 79141 24.14 ± 51.86 429 0.161 MCHC 337.93 ± 11.26 79141 338.67 ± 11.69 429 0.203 Age 30.02 ± 4.02 79141 29.76 ± 3.59 429 0.261 CR 46.95 ± 14.27 79141 47.59 ± 16.77 429 0.284 Height 1.59 ± 0.05 64456 1.59 ± 0.05 329 0.291 TBA 4.53 ± 5.62 79141 5.26 ± 8.27 429 0.409 PLT 180.3 ± 51.57 79141 183.11 ± 54.68 429 0.433 E# 0.06 ± 0.06 79141 0.06 ± 0.08 429 0.452 B# 0.23 ± 0.13 79141 0.23 ± 0.13 429 0.482 DNLR 0.92 ± 0.02 79141 0.92 ± 0.03 429 0.487 E% 0.62 ± 0.67 79141 0.61 ± 0.72 429 0.508 DBIL 1.86 ± 1.34 79141 1.94 ± 1.33 429 0.519 N# 8.15 ± 3.12 79141 8.61 ± 4.05 429 0.608 AST 21.31 ± 19.02 79141 24.26 ± 27.93 429 0.610 WBC 10.33 ± 3.31 79141 10.74 ± 4.14 429 0.661 UA 331.56 ± 82.05 79141 334.16 ± 89.4 429 0.817 PWR 18.97 ± 7.42 79141 19.22 ± 8.5 429 0.896 GLB 30.07 ± 3.61 79141 30.16 ± 4.06 429 0.908 B# 0.02 ± 0.01 79141 0.02 ± 0.01 429 0.933 In the table, EOS represents early-onset neonatal sepsis; SD is the standard deviation; N is the number of patients; WBC is the white blood cell count; PLT is the platelet count; RBC is the red blood cell count; HGB is the hemoglobin; HCT is the hematocrit; MCV is the mean corpuscular volume; MCH is the mean corpuscular hemoglobin level; MCHC is the mean corpuscular hemoglobin concentration; RDW is the red blood cell distribution width; RDWSD is the red blood cell distribution width (standard deviation); N% is the percentage of neutrophils; M% is the percentage of monocytes; N# is the absolute neutrophil count; M# is the absolute monocyte count; L% is the percentage of lymphocytes; E% is the percentage of eosinophils; L# is the absolute lymphocyte count; E# is the absolute eosinophil count; B% is the percentage of basophils; B# is the absolute basophil count; PA is prealbumin; TP is total protein; ALB is albumin; GLB is globulin; A / G is white blood cell count. Protein / globulin ratio; TBIL (total bilirubin); DBIL (direct bilirubin); IBIL (indirect bilirubin); TBA (total bile acids); ALT (alanine aminotransferase); AST (aspartate aminotransferase); ALP (alkaline phosphatase); GGT (gamma-glutamyl transferase); LDH (lactate dehydrogenase); CR (creatinine); UA (uric acid); CYSC (cystatin C); PWR (platelet-to-leukocyte ratio); NLR (neutrophil-to-lymphocyte ratio); MLR (monocyte-to-lymphocyte ratio); NMLR (neutrophil + monocyte) to lymphocyte ratio; SIRI (systemic inflammatory response index); SII (systemic immune inflammatory index); DNLR (derived neutrophil-to-lymphocyte ratio); SLR (aspartate aminotransferase to alanine aminotransferase ratio); UCR (urea to creatinine ratio); PNI (prognostic nutritional index); HGI (hemoglobin glycated index).

[0037] Maternal categorical variables associated with EOS risk: Univariate analysis showed that the following maternal factors significantly increased the risk of EOS: Preterm rupture of membranes (PPROM) RR=3.53, 95%CI: 2.57–4.84; Preterm birth RR=4.28, 95%CI: 3.44–5.33; Intrauterine fetal distress (IFD) RR=2.68, 95%CI: 2.15–3.34; Placental abnormalities (PI) RR=1.85, 95%CI: 1.37–2.50; Intrahepatic cholestasis of pregnancy (ICP) RR=2.16, 95%CI: 1.52–3.08; Hypertension RR=1.73, 95%CI: 1.27–2.37; Eclampsia Early pregnancy risk ratio (RR) = 1.76, 95% CI: 1.17–2.65; Cervical insufficiency risk ratio (RR) = 3.57, 95% CI: 2.02–6.31; Chorioamnionitis risk ratio (RR) = 15.68, 95% CI: 12.47–19.73; Puerperal infection risk ratio (RR) = 4.79, 95% CI: 1.81–12.66; Multiple pregnancy risk ratio (RR) = 2.88, 95% CI: 2.16–3.86; Mode of delivery risk ratio (RR) = 1.22, 95% CI: 1.01–1.47; High-risk pregnancy risk ratio (RR) = 1.41, 95% CI: 1.08–1.82; In vitro fertilization (IVF) risk ratio (RR) = 1.72, 95% CI: 1.31–2.26. A previous pregnancy history (RR=0.72, 95% CI: 0.60–0.87) and a previous delivery history (RR=0.53, 95% CI: 0.41–0.68) significantly reduced the risk of EOS. Umbilical cord entanglement and hepatitis B infection were not significantly associated with EOS (P>0.05). Figure 2 As shown.

[0038] Figure 2 This univariate analysis included maternal factors associated with the risk of early-onset neonatal sepsis (EOS). In the figure, PPROM represents premature rupture of membranes; IFD represents intrauterine fetal distress; PI represents placental abnormalities; ICP represents intrahepatic cholestasis of pregnancy; IVF represents in vitro fertilization; WBC represents white blood cell count; RBC represents red blood cell count; HGB represents hemoglobin; HCT represents hematocrit; MCHC represents mean corpuscular hemoglobin concentration; GLB represents globulins; TBIL represents total bilirubin; TBA represents total bile acids; ALT represents alanine aminotransferase; AST represents aspartate aminotransferase; ALP represents alkaline phosphatase; GGT represents gamma-glutamyl transferase; LDH represents lactate dehydrogenase; CYSC represents cystatin C; N% represents the percentage of neutrophils; M% represents the percentage of monocytes; N# represents the absolute neutrophil count; and M# represents the absolute monocyte count.

[0039] Maternal continuous variables related to EOS risk: To assess the impact of maternal continuous variables on EOS, this embodiment first used ROC curves to determine the optimal cutoff values ​​for all continuous variables, and then performed a dichotomous post-analysis based on these values. The results showed that an increase in maternal WBC levels (RR = 2.59, 95% CI 1.87–3.57) was observed. P <0.001), AST (RR = 1.78, 95%CI 1.32–2.41, P <0.001), alkaline phosphatase (ALP) (RR = 1.60, 95% CI 1.11–2.32, P = 0.018), and higher levels of pre-pregnancy weight, height, pre-pregnancy BMI, neutrophil percentage (N%), monocyte percentage (M%), N#, M#, total bilirubin (TBIL), total bile acids (TBA), ALT, gamma-glutamyl transferase (GGT), lactate dehydrogenase (LDH), and cystatin C (CYSC) significantly increased the risk of EOS. Conversely, higher levels of globulin (GLB) (RR = 0.56, 95% CI 0.38–0.82, ) significantly increased the risk of EOS. P =0.006), hemoglobin (HGB) (RR = 0.77, 95% CI 0.63–0.93, P = 0.009), and elevated levels of red blood cell count (RBC), hematocrit (HCT), and mean corpuscular hemoglobin concentration (MCHC) significantly reduced the risk of EOS.

[0040] RCS analysis was further employed to assess potential nonlinear associations and enhance robustness. Significant nonlinear relationships were found between EOS risk and the following parent variable: WBC ( P <0.001), HGB ( P = 0.029), mean corpuscular volume (MCV) ( P = 0.019), mean corpuscular hemoglobin (MCH) ( P = 0.039), M% ( P = 0.005), N# ( P = 0.001), Total Protein (TP) ( P = 0.003), TBIL ( P = 0.001), UREA ( P = 0.034), PWR ( P = 0.009), SIRI ( P = 0.045) and HGI ( P= 0.002). Specifically, the risk of EOS increases significantly when the population value exceeds or falls below the following threshold: WBC > 18.66 × 10 9 / L, HGB<86.35 g / L, MCV<92.50 fL, MCH<31.50 pg, M%<3.72% or>5.30%, N#>16.28 × 10 9 / L, UREA > 11.89 mmol / L, PWR > 17.93, and SIRI > 14.16. Furthermore, TP and TBIL showed a U-shaped association with EOS risk, with the lowest predicted risk at TP = 66.47 g / L and TBIL = 8.70 µmol / L. Figures 3-8 Restricted cubic spline (RCS) analysis was used to analyze maternal continuous variables associated with the risk of early-onset neonatal sepsis (EOS). In the figure, WBC represents white blood cell count; HGB represents hemoglobin; MCV represents mean corpuscular volume; MCH represents mean corpuscular hemoglobin; M% represents monocyte percentage; N# represents absolute neutrophil count; TP represents total protein; TBIL represents total bilirubin; PWR represents the platelet-to-white blood cell ratio; SIRI represents the systemic inflammatory response index; and HGI represents the glycated hemoglobin index.

[0041] To identify robust predictors of EOS, this embodiment uses Venn diagrams to visualize the overlap between variables identified by two independent methods: univariate analysis based on bivariate continuous variables using optimal cutoff values, and RCS modeling. A total of 12 overlapping variables were identified, including WBC, RBC, HGB, HCT, M%, N#, GLB, TBIL, SIRI, DNLR, HGI, and PWR. Figure 10 Venn diagrams were used to visualize the intersection of parent continuous variables significantly associated with EOS risk identified by two methods: univariate analysis based on cutoff values ​​and restricted cubic spline (RCS) modeling. Variables jointly identified by both methods were considered robust predictors and selected for subsequent multivariate modeling. These variables showed consistent associations with EOS in both methods and were therefore selected for further modeling. Additionally, parent categorical variables showing significant associations in previous univariate analyses were also included as candidates for model development.

[0042] Given the potential for multicollinearity among the selected variables, variance inflation factor (VIF) and tolerance statistics were used for collinearity diagnosis. Variables with high collinearity were excluded to improve model stability and interpretability. The remaining predictors were incorporated into a multivariate Poisson regression model (Model 1) and adjusted for maternal age and neonatal admission year (Table 1). 2From Model 1, the maternal factors significantly associated with EOS risk were retained in the final predictive model (Model 2). Among them, chorioamnionitis (IRR = 16.16, 95% CI 10.20–16.98) was also included. P <0.001) and puerperal infection (IRR = 3.91, 95% CI 1.56–9.83, P = 0.004) was associated with the highest risk. The risk of EOS in newborns from twin pregnancies was significantly higher than that from singleton pregnancies (IRR = 1.65, 95% CI 1.15–2.35). P = 0.006). Delivery method was also associated with EOS risk: Cesarean section significantly reduced the risk of EOS (IRR = 0.67, 95% CI 0.54–0.82). P = 0.001). Furthermore, the parental GLB level was negatively correlated with EOS risk (IRR = 0.58, 95% CI 0.40–0.85). P = 0.005), while PWR (IRR = 1.33, 95% CI 1.05–1.68, P = 0.018), HGI (IRR = 1.72, 95% CI 1.42–2.08, P <0.001) and M% (IRR = 1.43, 95% CI 1.17–1.75, P An increase in <0.001) was associated with increased risk. Other maternal conditions, such as preterm birth, IFD, PI, ICP, and cervical insufficiency, were also found to be independently associated with increased risk of EOS, as shown in Table 3.

[0043] Table 2. The association between parent factors and EOS risk was assessed using a multivariate Poisson regression model (Model 1). variable IRR z 95% CI ll 95% CI ul P Maternal age 0.99 -0.70 0.97 1.02 0.487 Year of admission 0.92 -2.91 0.87 0.97 0.004 PPROM 0.99 -0.07 0.67 1.46 0.944 Premature birth 3.40 7.72 2.49 4.65 0.000 IFD 2.18 6.12 1.70 2.80 0.000 PI 1.47 2.43 1.08 1.99 0.015 ICP 1.67 2.77 1.16 2.40 0.006 hypertension 1.33 1.20 0.83 2.11 0.232 Preeclampsia 0.89 -0.39 0.49 1.62 0.699 Cervical insufficiency 1.91 2.12 1.05 3.48 0.034 Chorioamnionitis 11.38 17.77 8.70 14.88 0.000 Puerperal infection 3.56 2.72 1.43 8.89 0.006 Multiple pregnancies 1.60 2.19 1.05 2.44 0.029 delivery method 0.66 -4.05 0.53 0.80 0.000 WBC 1.25 1.05 0.82 1.90 0.296 HGB 0.95 -0.44 0.75 1.20 0.660 M% 1.52 3.92 1.23 1.87 0.000 GLB 0.59 -2.73 0.40 0.86 0.006 SIRI 1.33 1.58 0.93 1.89 0.114 DNLR 1.22 1.30 0.90 1.64 0.195 HGI 1.65 4.26 1.31 2.08 0.000 PWR 1.43 2.88 1.12 1.82 0.004 TBIL 1.31 1.90 0.99 1.73 0.057 High-risk pregnancy 1.20 1.33 0.92 1.57 0.185 IVF 0.91 -0.49 0.63 1.32 0.624 pregnancy 0.88 -1.14 0.70 1.10 0.256 childbirth 0.78 -1.62 0.57 1.05 0.105

[0044] In the table, IRR is the event rate ratio; 95% CI is the 95% confidence interval; 11 is the lower limit; 11 is the upper limit; PPROM is premature rupture of membranes in preterm birth; IFD is intrauterine fetal distress; PI is placental abnormality; ICP is intrahepatic cholestasis of pregnancy; IVF is in vitro fertilization; WBC is white blood cell count; HGB is hemoglobin; M% is monocyte percentage; GLB is globulin; TBIL is total bilirubin; N% is neutrophil percentage; N# is absolute neutrophil count; M# is absolute monocyte count; PWR is platelet-to-white blood cell ratio; SIRI is the systemic inflammatory response index; DNLR is the derived neutrophil-to-lymphocyte ratio; HGI is the hemoglobin glycated index.

[0045] Table 3 shows the results of adjusting key maternal covariates using multivariate Poisson regression (Model 2) and logistic regression (sensitivity analysis) to examine their association with the incidence of EOS.

[0046] In the table, IRR is the event rate ratio; 95% CI is the 95% confidence interval; 11 is the lower limit; 11 is the upper limit; IFD is intrauterine fetal distress; PI is placental abnormality; ICP is intrahepatic cholestasis of pregnancy; M% is the percentage of monocytes; GLB is globulin; PWR is the platelet-to-white blood cell ratio; and HGI is the glycated hemoglobin index.

[0047] To evaluate the predictive performance of Model 2, receiver operating characteristic (ROC) curves were plotted and the AUC was calculated. The model demonstrated good discriminative ability in predicting EOS, with an AUC of 0.762 (…). P <0.001), indicating good discriminative ability for EOS, such as Figure 11 As shown, a nomogram was constructed based on a multivariate Poisson regression model to facilitate individualized risk prediction and improve clinical applicability. The nomogram contains 13 maternal predictors. Each variable is assigned a weighted score proportional to its contribution to the risk of neonatal EOS. By summing the individual scores, clinicians can use the scale at the bottom of the nomogram to estimate the predicted probability of neonatal EOS. This tool provides a practical and intuitive method for early risk stratification and decision-making in neonatal care. Figure 12 As shown.

[0048] Sensitivity analysis: To ensure the robustness of the results, this embodiment used the variables included in Model 2 for multivariate logistic regression analysis. The results were consistent with those of Poisson regression (all P values ​​< 0.05), as shown in Table 3. In the bootstrap results, the apparent performance indicators showed that the model had good discriminative ability for EOS (AUC = 0.762), a Brier score of 0.0053, a calibration intercept close to zero (-0.021), and a calibration slope close to 1 (0.995). After bootstrap correction, the model's performance remained stable (AUC = 0.757, Brier score = 0.0053, calibration intercept = -0.102, calibration slope = 0.978), indicating consistent performance on the resampled dataset, low risk of overfitting, and robust predictive ability. Furthermore, the bootstrap calibration plot showed a high degree of consistency between predicted probabilities and observed risks, providing additional support for the clinical applicability of this model in personalized EOS risk assessment. Figure 13 As shown.

[0049] This embodiment aims to explore the potential association between maternal factors and the risk of neonatal hemorrhagic angina (EOS), and to develop a predictive model for assessing the likelihood of EOS occurring prenatally or shortly postnatally. This model is expected to support earlier identification and intervention, thereby contributing to improved neonatal outcomes. This embodiment included 79,570 successfully matched mother-infant pairs, and a comprehensive analysis of the association between maternal factors and neonatal EOS risk was conducted. A predictive model was built using multivariate Poisson regression, which demonstrated good discriminative performance for EOS risk (AUC = 0.762). The model was further visualized as a nomogram containing 13 maternally relevant predictors, enabling individualized risk assessment of EOS.

[0050] Studies have found that newborns born to mothers aged 25 to 34 years have the highest risk of eosinophilic ovarian disease (EOS), with an incidence rate of 0.60% to 0.61%. This increased risk may be attributed to higher birth density and a higher incidence of pregnancy complications (such as chorioamnionitis) in this age group. Several perinatal maternal factors are significantly associated with the occurrence of EOS. Chorioamnionitis (IRR = 16.16, 95% CI 10.20 to 16.98) and puerperal infection (IRR = 3.91, 95% CI 1.56 to 9.83) were identified as the strongest predictors, likely due to their direct impact on intrauterine infection. Mechanistically, maternal reproductive tract pathogens (commonly group B streptococci, etc.) are a significant risk factor. E. coli Ureaplasma and Mycoplasma can ascend from the cervix to the chorion-amnion, causing chorioamnionitis and contaminated amniotic fluid. 。The fetus may inhale or swallow contaminated amniotic fluid, or acquire pathogens through skin and mucous membrane contact, which can colonize or invade the fetus before or immediately after birth. Inflammation can also increase placental barrier permeability, potentially leading to hematogenous transplacental transmission to the fetus in the presence of maternal bacteremia. At the maternal-fetal interface, chorioamnionitis activates innate immune pathways in the placenta and fetal membranes via TLR2 / TLR4 recognition of Gram-positive or Gram-negative pathogen-associated molecular patterns (PAMPs), triggering the MyD88–NF-κB signaling cascade. This results in the release of pro-inflammatory cytokines such as IL-1β, IL-6, and TNF-α, neutrophil infiltration, and complement activation. These processes may contribute to fetal inflammatory response syndrome and increase neonatal susceptibility to eosinophilic lesions (EOS). Furthermore, cervical insufficiency, intrauterine fetal distress, preterm birth, premature rupture of membranes, and intrahepatic cholestasis of pregnancy are all significantly associated with an increased risk of EOS, highlighting the crucial role of adverse intrauterine environment and delivery complications. Twin pregnancies are also associated with a higher risk of EOS, possibly due to a higher preterm birth rate and more frequent clinical interventions. Furthermore, vaginal delivery in cases of maternal infection increases the chance of exposure to pathogens during delivery, while elective cesarean section may reduce the risk of EOS (IRR = 0.67, 95% CI 0.54–0.82).

[0051] Regarding maternal laboratory parameters, the study identified several hematological and biochemical biomarkers significantly associated with EOS risk. Elevated levels of HGI, M%, PWR, and SIRI, as well as decreased levels of GLB and HGB, were associated with increased EOS risk. These findings suggest that maternal inflammatory status, immune alterations, and liver function may play a crucial role in determining fetal susceptibility to infection. Further analysis using restricted cubic spline functions revealed nonlinear associations between several maternal biomarkers (e.g., WBC, HGB, GLB, TP, and TBIL) and EOS risk, highlighting a potential threshold effect and providing quantitative insights for clinical surveillance. Specifically, a maternal WBC count >18.66 × 10⁻⁶ was associated with EOS risk. 9 / L and / or PWR values ​​>17.93 were significantly associated with an increased risk of EOS. The white blood cell threshold determined in this study exceeded the upper limit of the reference range for pregnant women in most guidelines (15 × 10⁻⁶). 9 Leukocytosis ( / L) suggests that more pronounced leukocytosis may be particularly associated with the risk of eosinophilia (EOS). Leukocytosis during pregnancy may indicate intrauterine infection, such as chorioamnionitis, or persistent maternal inflammation. This is influenced by inflammatory mediators and proteolytic enzymes. ofThese conditions can impair the integrity of the fetal membranes and epithelium, thereby weakening the physical and chemical barrier function of the fetal or neonatal mucosa and increasing the risk of bacterial translocation into the bloodstream. Similarly, the PWR threshold (17.93) is not explicitly incorporated into current obstetric or neonatal guidelines, highlighting the need for further validation before widespread adoption. Nevertheless, PWR is an inflammation-sensitive ratio that may reflect an imbalance in the immune coagulation axis, potentially caused by relative leukopenia or thrombocytosis, both of which can disrupt maternal-fetal immune balance and increase the risk of EOS. These thresholds can serve as early warning indicators of high-risk pregnancies, prompting increased perinatal monitoring. Maternal TP and TBIL levels also exhibit a U-shaped relationship with EOS risk. Low TP or TBIL levels may indicate maternal malnutrition, impaired liver function, or immunoglobulin deficiency, leading to inadequate intrauterine immune protection. Conversely, elevated TP or TBIL levels may be associated with cholestasis, hepatocellular damage, or systemic inflammation, which can impair placental function or alter intrauterine microbiota colonization, thereby promoting the development of EOS.

[0052] This embodiment explores the potential association between maternal factors and neonatal eosinophilic obstructive uterine disease (EOS) risk. Maternal factors may provide a valuable method for assessing EOS risk because they are often available before or during delivery, enabling early identification of newborns who can benefit from close monitoring or timely intervention. Compared to neonatal clinical signs and laboratory indicators, maternal factors are generally more objective, readily available, non-invasive, and pose no risk to the newborn, suggesting their potential applicability in various clinical settings. Furthermore, maternal indicators may provide indirect information about the intrauterine environment, which helps elucidate the mechanisms leading to EOS.

[0053] To ensure the robustness of the research results, this embodiment conducted a sensitivity analysis of multivariate logistic regression on the variables in Model 2. The results were consistent with those of the Poisson regression model, further supporting the reliability of the conclusions. Integrating multiple statistical methods, including cutoff value-based grouping, restricted cubic spline models, variable selection based on Venn diagrams, and multivariate regression analysis, enhanced the scientific rigor of variable selection and improved the stability of the model.

[0054] Furthermore, in this embodiment, a user-friendly web-based EOS prediction tool was developed based on this model. Clinicians only need to input 13 corresponding indicators for the pregnant woman to automatically generate a risk score for neonatal EOS, ranging from 0 to 100 points, with higher scores indicating a greater risk of EOS. This tool not only provides quantitative risk assessment but also helps clinicians quickly identify high-risk pregnant women and newborns, enabling targeted early intervention and close monitoring. It is worth emphasizing that this model can complete risk prediction before the newborn's birth, providing a scientific basis for the development of empirical anti-infection strategies and multidisciplinary collaboration between obstetrics and neonatology, thereby significantly improving the foresight and individualization of perinatal management. Through this innovative platform, clinicians can more efficiently conduct risk stratification management, implement timely preventive measures for high-risk pregnant women and newborns, potentially reducing the incidence of EOS, optimizing medical resource allocation, and improving overall perinatal safety.

[0055] The maternal-factor-based risk prediction model for early neonatal sepsis (EOS) constructed in this embodiment has several advantages. First, previous research on EOS risk prediction is limited, and existing models are mainly based on neonatal clinical or laboratory indicators. This not only involves invasive procedures on newborns but also has a certain time lag, delaying early intervention and treatment for high-risk infants. In contrast, the model in this embodiment is entirely based on maternal indicators, causing no harm to the newborn and is more easily accepted and promoted by patients and their families. Second, all indicators are derived from maternal clinical information, including gestational age, pregnancy complications, time of membrane rupture, infection history, and related laboratory tests. These can be obtained prenatally or at delivery, without relying on invasive sampling or laboratory testing after birth. Data acquisition is convenient and highly complete. This model can achieve risk assessment before birth, providing a scientific basis for early monitoring of high-risk newborns, development of empirical anti-infection strategies, and multidisciplinary collaboration between obstetrics and neonatology, thereby enabling prospective and individualized intervention management. Furthermore, since maternal factors are generally available at different levels of medical institutions, this model has good potential for widespread application and can assist primary hospitals in optimizing perinatal management. Finally, this model innovatively introduces two indicators—HGI and PWR. Existing studies have shown that HPI is associated with the risk of various cardiometabolic diseases, while PWR has reference value in the prognosis or risk assessment of diseases such as acute myeloid leukemia and postoperative acute kidney injury, but these have not been used in EOS studies before. This study found that these two indicators are significantly associated with the risk of EOS, further enhancing the model's innovativeness and predictive value.

[0056] In summary, this embodiment establishes a neonatal EOS prediction model based on maternal factors, demonstrating good discriminative performance and potential clinical applicability. This nomogram, as a visual risk assessment tool, can individually predict EOS risk, providing a new method for the early identification and precise management of high-risk newborns during the perinatal period.

[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the risk of early-onset neonatal sepsis based on maternal factors, characterized in that, Includes the following steps: S1. Collect maternal data, including maternal clinical and laboratory data; S2. Perform variable processing and screening on the maternal data to obtain significant variables related to the risk of early-onset neonatal sepsis. S3. Establish a multivariate statistical prediction model based on the significant variables; S4. Use the prediction model to output a risk score for early-onset neonatal sepsis for individualized risk assessment.

2. The method for predicting the risk of early-onset neonatal sepsis based on maternal factors as described in claim 1, characterized in that, The maternal data indicators include at least one of the following: history of pregnancy-related diseases, mode of delivery, multiple pregnancy, preterm birth, intrauterine fetal distress, placental abnormalities, intrahepatic cholestasis of pregnancy, chorioamnionitis, puerperal infection, cervical insufficiency, white blood cell count, monocyte percentage, platelet-to-white blood cell ratio, hemoglobin glycated index, and globulin.

3. The method for predicting the risk of early-onset neonatal sepsis based on maternal factors as described in claim 1 or 2, characterized in that, In step S2, for categorical variables, univariate analysis is used to identify maternal risk factors associated with early-onset neonatal sepsis as significant categorical variables.

4. The method for predicting the risk of early-onset neonatal sepsis based on maternal factors as described in claim 3, characterized in that, For continuous variables, the optimal cutoff value was determined by ROC curve analysis and then divided, followed by a chi-square test to determine the ROC continuous variable. At the same time, RCS analysis was used to assess the linear or nonlinear relationship between the continuous variable and the risk of early-onset neonatal sepsis to determine the RCS continuous variable. Venn diagrams were used to visualize the ROC and RCS continuous variables, and the continuous variables in the overlapping part were identified as significant continuous variables.

5. The method for predicting the risk of early-onset neonatal sepsis based on maternal factors as described in claim 4, characterized in that, In step S3, variables that differ between significant categorical and significant continuous variables are merged, and the variance inflation factor and tolerance are used to assess collinearity between variables, eliminating variables with a variance inflation factor greater than 10 or a tolerance less than 0.

1.

6. The method for predicting the risk of early-onset neonatal sepsis based on maternal factors as described in claim 4, characterized in that, In step S3, a multivariate Poisson regression model is used to establish a prediction model, and confounding factors are adjusted to calculate the event occurrence ratio and its confidence interval. The confounding factors include at least the maternal age and the year of neonatal admission.

7. The method for predicting the risk of early-onset neonatal sepsis based on maternal factors as described in claim 4, characterized in that, In step S3, the model performance is evaluated by the area under the receiver operating characteristic curve, and the model stability is verified by Bootstrap resampling.

8. The method for predicting the risk of early-onset neonatal sepsis based on maternal factors as described in claim 1, characterized in that, In step S3, the independent predictors of the prediction model include at least one of the following: chorioamnionitis, puerperal infection, preterm birth, intrauterine fetal distress, multiple pregnancy, elevated monocyte percentage, elevated platelet-to-leukocyte ratio, elevated hemoglobin glycated index, decreased globulin, cesarean section as the mode of delivery, cervical insufficiency, placental abnormalities, and intrahepatic cholestasis of pregnancy.

9. A risk prediction system for early-onset neonatal sepsis based on maternal factors, characterized in that, include: The data acquisition module is used to collect clinical and laboratory data from the mother. The data processing and filtering module is used to process and filter the maternal data to obtain significant variables related to the risk of neonatal early-onset sepsis. The modeling module is used to build a multivariate statistical prediction model based on the significant variables; The risk scoring module is used to output a risk score for early-onset neonatal sepsis using the prediction model and to conduct individualized risk assessment.

10. The neonatal early-onset sepsis risk prediction system based on maternal factors as described in claim 9, characterized in that, The data acquisition module is used to collect at least one maternal data indicator from the following: pregnancy history, delivery method, multiple pregnancy, premature birth, intrauterine fetal distress, placental abnormalities, intrahepatic cholestasis of pregnancy, chorioamnionitis, puerperal infection, cervical insufficiency, white blood cell count, monocyte percentage, platelet-to-white blood cell ratio, hemoglobin glycated index, and globulin. The modeling module includes a multivariate Poisson regression modeling unit, which is used to build a predictive model after removing highly collinear variables, adjust for maternal age and neonatal admission year, calculate the event incidence ratio and its confidence interval, evaluate model performance by the area under the receiver operating characteristic curve, and verify model stability by Bootstrap resampling. The risk scoring module is used to construct a nodal chart based on the final model and generate a risk score of 0–100.

Images