Premature infant bronchopulmonary dysplasia prediction model construction method and prediction device

Abstract

The invention discloses a premature bronchial lung dysplasia prediction model construction method and a prediction device. The method comprises the following steps: S1, obtaining feature data of a premature with the gestational age less than 32 weeks; s2, carrying out regression analysis on the obtained feature data by utilizing single-factor logistic regression, screening out single-factor predictive variables related to the BPD of the premature infant, and converting Plt; a single-factor predictive variable of 0.05 is incorporated into Lasso regression analysis, and an independent predictive factor related to BPD is screened out; s3, introducing the screened independent prediction factors into multivariable logistic logistic regression, and constructing a multivariable logistic regression model through forward stepwise regression, backward stepwise regression and bidirectional stepwise regression methods; selecting a model with the minimum AIC value based on an Akaike information criterion, and constructing a premature infant BPD prediction model by using variables in the model with the minimum AIC value; only perinatal period factors are needed during prediction of the device, and the influence of clinical management factors is avoided; and after the prediction model is converted into the column diagram, the method is more suitable for clinical rapid prediction application.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer models for disease prediction, and in particular relates to a method for constructing a prediction model for bronchopulmonary dysplasia in premature infants and a prediction device. Background Art

[0002] Bronchopulmonary dysplasia (BPD) is a common chronic lung disease in premature infants and one of the main causes of death and disability in premature infants. With the improvement of medical treatment, the survival rate of extremely premature infants has increased significantly, and the incidence of BPD has also increased accordingly. In 2019, the National Institute of Child Health and Human Development (NICHD) database showed that the incidence of BPD in premature infants with a gestational age of <32 weeks was as high as 71.1%. A multicenter study in my country in 2020 reported that the incidence of BPD in extremely premature infants with a gestational age of 22-28 weeks was as high as 72.2%. BPD was first reported by Northway et al. in 1967. The most widely used diagnostic criteria in clinical practice are the versions proposed by NICHD in 2001, which are diagnosed at 28 days after birth and graded at 36 weeks of corrected gestational age (PMA), 56 days after birth, or at discharge. However, with the advancement and diversification of respiratory support technology, this standard can no longer fully meet clinical needs. In 2018, Higgins et al. proposed a new diagnostic standard for BPD, which unified the evaluation time to 36 weeks of PMA and subdivided the condition based on ventilation mode and inspired oxygen concentration. This standard has been gradually accepted and widely used in clinical diagnosis and epidemiological research.

[0003] It is currently believed that BPD is caused by multiple factors before, during, and after birth, based on immature lung development. Although there is no effective treatment, prevention and early comprehensive management have become the main strategies, with an emphasis on early prediction and intervention. Currently, a variety of BPD prediction models have been reported at home and abroad. However, most of these models focus on risk factors within 14 days after birth, and mainly use the 2001 NICHD diagnostic criteria, and have not been validated by multicenter large-scale data. In addition, due to differences in race and medical conditions, these models may not be fully applicable to the Chinese population. With the introduction of new diagnostic criteria, the assessment time of BPD has been postponed to 36 weeks of PMA. In addition to prenatal and early postnatal factors, diseases within 28 days after birth also play an important role in the occurrence and development of BPD.

[0004] A nomogram, also known as a Nomogram chart, is based on multiple-factor regression analysis. It integrates multiple predictive indicators and then uses calibrated line segments to be plotted on the same plane according to a certain ratio to express the mutual relationship between various variables in the prediction model. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems in the related art to some extent. For this reason, the object of the present invention is to provide a method for constructing a prediction model for bronchopulmonary dysplasia in premature infants and a prediction device.

[0006] To achieve the above object, according to the first aspect of the present invention, the method for constructing a prediction model for bronchopulmonary dysplasia in premature infants provided by the present invention includes the steps of:

[0007] S1: Obtain the characteristic data of premature infants with a gestational age less than 32 weeks, where the characteristic data includes perinatal maternal information data, perinatal information data, and neonatal information data;

[0008] S2: Use univariate logistic regression to perform regression analysis on the obtained characteristic data, preliminarily screen out the univariate predictive variables related to BPD in premature infants, and incorporate the univariate predictive variables with P < 0.05 into the Lasso regression analysis to screen out the independent predictive factors related to BPD;

[0009] S3: Incorporate the screened independent predictive factors into multivariate logistic regression, and construct a multivariate logistic regression model through forward stepwise regression, backward stepwise regression, and bidirectional stepwise regression methods; and select the model with the minimum AIC value based on the Akaike information criterion, and construct a prediction model for BPD in premature infants using multiple variables in the model with the lowest AIC value;

[0010] According to an embodiment of the present invention, the perinatal maternal information data includes any one or more of the following: age, race, pregnancy weight gain, gestational hypertension, gestational diabetes, intrahepatic cholestasis of pregnancy, abnormal thyroid function during pregnancy, chorioamnionitis, autoimmune disease, duration of premature rupture of membranes, Streptococcus agalactiae infection or Ureaplasma urealyticum infection;

[0011] The perinatal information data includes any one or more of the following: prenatal steroid use, prenatal magnesium sulfate use, antibiotic use within 24 hours before delivery, whether cervical cerclage was performed before delivery, prenatal fetal medical intervention, conception method, multiple pregnancy situation, delivery method, 1-minute Apgar score, 5-minute Apgar score, or neonatal asphyxia situation;

[0012] The neonatal information data includes any one or more of the following: gender, gestational age, birth weight, whether small for gestational age (SGA), ratio of birth weight to gestational age (RBG), neonatal respiratory distress syndrome (NRDS), use of pulmonary surfactant (PS), persistent pulmonary hypertension of the newborn (PPHN), whether complicated with pneumothorax, whether complicated with pulmonary hemorrhage, the first respiratory support mode and duration, whether invasive respiratory support is used and the time, early-onset sepsis, and hemodynamically significant patent ductus arteriosus, antibiotic use time, caffeine use, or the occurrence of bronchopulmonary dysplasia (BPD).

[0013] According to an embodiment of the present invention, the variables for constructing the preterm BPD prediction model include: gestational age, birth weight, whether cervical cerclage was performed antenatally, whether complicated with pneumothorax, whether complicated with pulmonary hemorrhage, initial ventilation duration, whether invasive ventilation was performed and the invasive ventilation time.

[0014] According to an embodiment of the present invention, the prediction formula of the preterm BPD prediction model is:

[0015] P = 1 / (1 + exp(-(3.935 - 0.176 * ga - 0.001 * bw + 0.717 * cerclageofcervix + 0.715 * pntx + 0.624 * pnmh + 0.813 * imv + 0.071 * imvtime + 0.050 * irstime)));

[0016] Wherein, P represents the predicted probability value, ga represents the gestational age at birth, bw represents the birth weight, cerclageofcervix represents antenatal cervical cerclage, pntx represents pneumothorax, pnmh represents pulmonary hemorrhage, imv represents whether invasive ventilation was performed, imvtime represents the invasive ventilation time, and irstime represents the initial ventilation duration.

[0017] According to an embodiment of the present invention, after step S3, the following step is further included:

[0018] S4: Drawing a nomogram prediction model according to the preterm BPD prediction model.

[0019] According to an embodiment of the present invention, the prediction formula of the nomogram prediction model is:

[0020] Total score = score corresponding to invasive ventilation time + score corresponding to initial ventilation time + score corresponding to invasive ventilation + score corresponding to pulmonary hemorrhage + score corresponding to pneumothorax + score corresponding to cervical cerclage + score corresponding to birth weight + score corresponding to gestational age at birth; wherein, the probability corresponding to the total score is the risk probability of preterm infants developing BPD.

[0021] According to an embodiment of the present invention, the variables constitute eight scales; wherein,

[0022] The first scale is the scale for the duration of invasive respiratory support, with a value range of 0 - 70.17 days, corresponding to scores of 0 - 8.8 points;

[0023] The second scale is the scale for the initial respiratory support time, with a value range of 0 - 59 days, corresponding to scores of 0 - 5.3 points;

[0024] The third scale is the scale for whether there is invasive ventilation, with a value range of 0 - 1, corresponding to scores of 0 - 1.4 points;

[0025] The fourth scale is the scale for whether there is combined pulmonary hemorrhage, with a value range of 0 - 1, corresponding to scores of 0 - 1.2 points;

[0026] The fifth scale is the scale for whether there is combined pneumothorax, with a value range of 0 - 1, corresponding to scores of 0 - 1.3 points;

[0027] The sixth scale is the scale for whether there is cervical cerclage before delivery, with a value range of 0 - 1, corresponding to scores of 0 - 1.3 points;

[0028] The seventh scale is the scale for birth weight, with a value range of 570 - 2690 g, corresponding to scores of 0 - 3.4 points;

[0029] The eighth scale is the scale for gestational age at birth, with a value range of 23.29 - 31.86, corresponding to scores of 0 - 2.7 points.

[0030] According to an embodiment of the present invention, after step S5, the method further includes the steps:

[0031] S5: Obtain one or more sets of validation data sets;

[0032] S6: According to the validation data set, use any one or more of the ROC curve, calibration curve, or DCA curve to evaluate the constructed nomogram prediction model for bronchopulmonary dysplasia in premature infants.

[0033] According to still another aspect of the present invention, an embodiment of the present invention provides a nomogram prediction device for bronchopulmonary dysplasia in premature infants. The nomogram prediction device for bronchopulmonary dysplasia in premature infants includes a computer device, and the computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the above - constructed nomogram prediction model for bronchopulmonary dysplasia in premature infants.

[0034] Through the above technical solution, the method and prediction device for constructing a nomogram prediction model for bronchopulmonary dysplasia in premature infants of the present invention. The method includes step S1: Obtain the characteristic data of premature infants with a gestational age less than 32 weeks, and the characteristic data includes perinatal maternal information data, perinatal information data, and neonatal information data;

[0035] S2: Use univariate logistic regression to perform regression analysis on the obtained feature data, preliminarily screen out univariate predictive variables related to BPD in premature infants, and incorporate univariate predictive variables with P < 0.05 into Lasso regression analysis to screen out independent predictive factors related to BPD; S3: Incorporate the screened independent predictive factors into multivariate logistic regression, and construct a multivariate logistic regression model through forward stepwise regression, backward stepwise regression, and bidirectional stepwise regression methods; and select the model with the smallest AIC value based on the Akaike information criterion, and construct a prediction model for BPD in premature infants using multiple variables in the model with the lowest AIC value; S4: Draw a nomogram prediction model according to the prediction model for BPD in premature infants. The device containing this BPD prediction model only requires perinatal factors and is not affected by clinical management factors. It has good prediction efficacy in predicting BPD in premature infants under the new diagnostic criteria, and has high sensitivity and specificity. After internal and external data verification, it is helpful for clinical diagnosis. After converting the prediction model into a nomogram, it is more simple and intuitive for clinical application.

[0036] Other features and advantages of the present invention will be described in detail in the subsequent specific implementation section. Description of the Drawings

[0037] Figure 1 is a flowchart of a method for constructing a prediction model for bronchopulmonary dysplasia in premature infants provided by an embodiment of the present invention;

[0038] Figure 2 is another flowchart of a method for constructing a prediction model for bronchopulmonary dysplasia in premature infants provided by an embodiment of the present invention;

[0039] Figure 3 is a schematic structural diagram of a prediction device for bronchopulmonary dysplasia in premature infants provided by an embodiment of the present invention;

[0040] Figure 4A is a curve graph showing the change characteristics of variable coefficients for variable screening based on Lasso regression provided by an embodiment of the present invention;

[0041] Figure 4B is a result graph of Lasso regression cross-validation for variable screening based on Lasso regression provided by an embodiment of the present invention;

[0042] Figure 5 is a nomogram model graph for predicting the occurrence of BPD in premature infants provided by an embodiment of the present invention;

[0043] Figure 6A is a curve graph for validating the prediction model using the training set of the receiver operating characteristic (ROC) curve provided by an embodiment of the present invention;

[0044] Figure 6B It is a graph for validating a prediction model using an internal validation set of a Receiver Operating Characteristic (ROC) curve provided by an embodiment of the present invention;

[0045] Figure 6C It is a graph for validating a prediction model using an external validation set of a Receiver Operating Characteristic (ROC) curve provided by an embodiment of the present invention;

[0046] Figure 7A It is a calibration scatter plot evaluation graph of a prediction model using a training set provided by an embodiment of the present invention;

[0047] Figure 7B It is a calibration scatter plot evaluation graph of a prediction model using an internal validation set provided by an embodiment of the present invention;

[0048] Figure 7C It is a calibration scatter plot evaluation graph of a prediction model using an external validation set provided by an embodiment of the present invention;

[0049] Figure 8A It is a decision curve analysis graph of a prediction model using a training set provided by an embodiment of the present invention;

[0050] Figure 8B It is a decision curve analysis graph of a prediction model using an internal validation set provided by an embodiment of the present invention;

[0051] Figure 8C It is a decision curve analysis graph of a prediction model using an external validation set provided by an embodiment of the present invention. Detailed implementation manners

[0052] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below. Preferred embodiments of the present invention are given below. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0054] On the one hand, referring to Figure 1 , an embodiment of the present invention provides a method for constructing a prediction model for bronchopulmonary dysplasia in premature infants, including the steps:

[0055] S1: Obtain the characteristic data of preterm infants with a gestational age less than 32 weeks, where the characteristic data includes perinatal maternal information data, perinatal information data, and neonatal information data;

[0056] S2: Use univariate logistic regression to perform regression analysis on the obtained characteristic data, preliminarily screen out the univariate predictive variables related to BPD in preterm infants, incorporate the univariate predictive variables with P < 0.05 into the Lasso regression analysis, and screen out the independent predictive factors related to BPD;

[0057] S3: Incorporate the screened independent predictive factors into multivariate logistic regression, and construct a multivariate logistic regression model through forward stepwise regression, backward stepwise regression, and bidirectional stepwise regression methods; and select the model with the smallest AIC value based on the Akaike information criterion, and construct a BPD prediction model for preterm infants using multiple variables in the model with the lowest AIC value;

[0058] S4: Draw a nomogram prediction model according to the BPD prediction model for preterm infants.

[0059] Specifically, in step S1, obtain the characteristic data of preterm infants with a gestational age less than 32 weeks. Among them, the data in the embodiments of the present invention come from the Shenzhen Data Collaboration Network, and preterm infants with a gestational age less than 32 weeks treated in 28 tertiary neonatal intensive care units (NICUs) in Shenzhen are included, and the time range is from January 2022 to December 2023. All participating hospitals received standardized training before the project started. Inclusion criteria: The inclusion criteria are as follows:

[0060] (1) Preterm infants with a gestational age less than 32 weeks;

[0061] (2) Length of hospital stay ≥ 7 days. Exclusion criteria are as follows: 1. Infants who died within 7 days after birth;

[0062] Those with chromosomal or genetic abnormalities, combined with multiple malformations, such as congenital lung diseases (congenital pulmonary hypoplasia, pulmonary sequestration, congenital bronchopulmonary cysts, pulmonary aplasia, and congenital pulmonary arteriovenous fistula, etc.), congenital heart diseases (such as atrial septal defect, ventricular septal defect, pulmonary valve stenosis, aortic valve stenosis, tetralogy of Fallot, etc.) or other systemic malformations.

[0063] Among them, the dataset of Peking University Shenzhen Hospital is selected as the external validation set, and the remaining datasets are randomly divided into a training set and an internal validation set by a computer method in a ratio of 7:3. Whether BPD is combined during hospitalization is used as the main outcome indicator.

[0064] The characteristic data includes perinatal maternal information data, perinatal information data, and neonatal information data;

[0065] Among them, the perinatal maternal information data include any one or more of the following: age, race, pregnancy weight gain, gestational hypertension, gestational diabetes, intrahepatic cholestasis of pregnancy, abnormal thyroid function during pregnancy, chorioamnionitis, autoimmune diseases, time of premature rupture of membranes, group B streptococcus infection or ureaplasma urealyticum infection;

[0066] Among them, the perinatal information data include any one or more of the following: prenatal steroid use, prenatal magnesium sulfate use, antibiotic use within 24 hours before delivery, whether cervical cerclage was performed prenatally, prenatal fetal medical intervention, conception method, multiple pregnancy situation, delivery method, 1-minute Apgar score, 5-minute Apgar score or neonatal asphyxia situation;

[0067] Among them, the neonatal information data include any one or more of the following: gender, gestational age, birth weight, whether small for gestational age (SGA), ratio of birth weight to gestational age (RBG), neonatal respiratory distress syndrome (NRDS), use of pulmonary surfactant (PS), persistent pulmonary hypertension of the newborn (PPHN), whether complicated with pneumothorax, whether complicated with pulmonary hemorrhage, first respiratory support mode and duration, whether invasive respiratory support was used and the time, early-onset sepsis and hemodynamically significant patent ductus arteriosus, antibiotic use time, caffeine use or bronchopulmonary dysplasia (BPD) occurrence;

[0068] In the embodiment of the present invention, according to whether BPD is combined at 36 weeks of postmenstrual age (PMA), the enrolled study subjects are divided into a BPD group and a non-BPD group. The BPD diagnostic criteria are based on the revised criteria in 2018: premature infants with a gestational age ≤ 32 weeks, accompanied by imaging-confirmed persistent parenchymal lung lesions, and requiring a certain level of respiratory support or inspired oxygen concentration for at least 3 consecutive days at 36 weeks of PMA to maintain arterial oxygen saturation at 90% - 95%.

[0069] Finally, a training set composed of 863 individuals and an internal validation set composed of 370 individuals are included in the training set. In the training set, there are 204 cases of BPD (23.64%), while in the internal validation set, there are 104 cases of BPD (28.11%). In addition, the external validation set includes 103 individuals, and 25 cases of BPD (24.27%) are recorded.

[0070] In step S2, univariate logistic regression is used to perform regression analysis on the obtained feature data. For example, SPSS 25.0 software can be used to perform univariate logistic regression analysis on 45 clinical variables in the obtained feature data. The results show that 18 indicators, including gestational age at birth, birth weight, maternal weight gain during pregnancy, antenatal cervical cerclage, 1-minute Apgar score, 5-minute Apgar score, asphyxia, administration of pulmonary surfactant after birth, neonatal respiratory distress syndrome, pneumothorax, pulmonary hemorrhage, persistent pulmonary hypertension, hemodynamically significant patent ductus arteriosus, first respiratory support mode and duration, whether invasive respiratory support is used and the time, and caffeine use time, have statistical differences (P<0.05). (Table 1) shows the univariate predictive variables related to BPD.

[0071] Table 1 Results of univariate logistic regression

[0072]

[0073]

[0074]

[0075] To further screen and reduce the dimension, the univariate predictive variables with P<0.05 in the univariate logistic regression analysis are incorporated into the LASSO regression analysis, as Figure 4A and Figure 4B shown. The Lasso regression analyzes 18 all univariate predictive variables with P<0.05, Figure 4A showing the change trajectory of the regression coefficient value of each predictive variable being compressed to 0 as the regularization strength λ gradually increases. Figure 4B shows the trend of the binomial deviance (used to measure the goodness of fit of the model) changing with the regularization parameter λ. The lowest point represents the just-fitted state of the model, and the λ at this time is the minimum λ. The minimum λ value is marked by a vertical line, and 11 independent predictive variables with non-zero coefficients are determined. Finally, 11 independent predictive factors related to BPD are obtained, including gestational age at birth, birth weight, antenatal cervical cerclage, 1-minute Apgar score, pneumothorax, pulmonary hemorrhage, persistent pulmonary hypertension, first respiratory support mode and duration, whether invasive respiratory support is used and the time (Table 2).

[0076] Table 2 Results of Lasso regression

[0077] Variables Lasso regression coefficient Gestational age at birth (GA) -0.1807702234 Birth weight (BW) -0.0006400681 Prenatal cervical cerclage 0.5292237707 1-minute Apgar score (AP1) -0.0026390875 Pneumothorax (PNTX) 0.4081112533 Pulmonary hemorrhage (PNMH) 0.2983355374 Persistent pulmonary hypertension of the newborn (PPHN) 0.0323433129 Invasive mechanical ventilation (IMV) 0.2036667680 Initial respiratory support (IRS) 0.5586316896 Duration of initial respiratory support (IRStime) 0.0401867020 Duration of invasive mechanical ventilation (IMVtime) 0.0668792565

[0078] In step S3, the 11 indicators (gestational age at birth, birth weight, antenatal cervical cerclage, 1-minute Apgar score, pneumothorax, pulmonary hemorrhage, persistent pulmonary hypertension, initial respiratory support mode and duration, whether invasive respiratory support is used and the time) screened by the above Lasso regression were included in the multivariate logistic regression. A multivariate logistic regression model was constructed by forward stepwise regression, backward stepwise regression and stepwise regression methods. The model was selected based on the Akaike information criterion (AIC), and the model with the smallest AIC value was selected as the optimal model. A probability prediction model with better performance was selected to construct a prediction model for preterm BPD.

[0079] The final results showed that the model composed of 8 key indicators (variables), namely gestational age at birth, birth weight, whether there was cervical cerclage before delivery, whether pneumothorax was combined, whether pulmonary hemorrhage was combined, initial ventilation duration, whether invasive ventilation was performed and the time of invasive ventilation, was the optimal model (Table 3). A prediction model for preterm BPD was constructed using these 8 variables.

[0080] Table 3 Results of multivariate logistic regression analysis

[0081]

[0082] The logistic regression equation of the prediction model for preterm BPD finally obtained was:

[0083] f(x) = 3.935 - 0.176(ga) - 0.001(bw) + 0.717(cerclage of cervix) + 0.715(pntx) + 0.624(pnmh) + 0.813(imv) + 0.071(imvtime) + 0.050(irstime).

[0084] The formula for calculating the predicted probability value of the prediction model for preterm BPD according to this equation was:

[0085] P = 1 / (1 + exp(-(3.935 - 0.176*ga - 0.001*bw + 0.717*cerclage of cervix + 0.715*pntx + 0.624*pnmh + 0.813*imv + 0.071*imvtime + 0.050*irstime)));

[0086] Among them, P represents the predicted probability value, ga represents the gestational age at birth, bw represents the birth weight, cerclageofcervix represents antenatal cervical cerclage, pntx represents pneumothorax, pnmh represents pulmonary hemorrhage, imv represents whether there is invasive ventilation, imvtime represents the invasive ventilation time, and irstime represents the initial ventilation duration. The device containing this BPD prediction model only requires perinatal factors and is not affected by clinical management factors. It has good predictive efficacy in predicting BPD in preterm infants under the new diagnostic criteria, with high sensitivity and specificity, and has been verified by internal and external data, which is helpful for clinical diagnosis.

[0087] Refer to Figure 2 and Figure 5 , according to an embodiment of the present invention, after step S3, step S4 is further included. In step S4, a nomogram prediction model is drawn according to the preterm infant BPD prediction model. Since applying the predicted probability value equation obtained in step S4 clinically requires a large number of complex operations, therefore, in this step S4, the stata nomolog command can be used to convert the equation into a nomogram (as shown in Figure 5 ). There are corresponding scale lines for all 8 variables in the nomogram. When applying, find the corresponding points of the child's indicators on the number axis, and at the same time record the scores projected onto the lower scale lines. Add the scores of the eight variables to get the total score. Finally, find the predicted probability value corresponding to the position of the total score, which is the possibility that the child has BPD obtained through this nomogram. The prediction formula of the nomogram prediction model is:

[0088] Total score value = score corresponding to invasive ventilation time + score corresponding to initial ventilation time + score corresponding to invasive ventilation + score corresponding to pulmonary hemorrhage + score corresponding to pneumothorax + score corresponding to cervical cerclage + score corresponding to birth weight + score corresponding to gestational age at birth; among them, the probability corresponding to the total score value is the risk probability of preterm infants developing BPD.

[0089] Refer to Figure 5 , in the nomogram, the 8 variables form eight scales; among them,

[0090] The first row is the scale of invasive respiratory support time, with a value range of 0 - 70.17 days and corresponding scores of 0 - 8.8 points;

[0091] The second row is the scale of initial respiratory support time, with a value range of 0 - 59 days and corresponding scores of 0 - 5.3 points;

[0092] The third row is the scale of whether there is invasive ventilation, with a value range of 0 - 1 and corresponding scores of 0 - 1.4 points;

[0093] The fourth row is the scale of whether there is combined pulmonary hemorrhage, with a value range of 0 - 1 and corresponding scores of 0 - 1.2 points;

[0094] The fifth row is whether to combine the pneumothorax scale, with a value range of 0 - 1 and corresponding scores of 0 - 1.3 points;

[0095] The sixth row is whether to perform cervical cerclage before delivery scale, with a value range of 0 - 1 and corresponding scores of 0 - 1.3 points;

[0096] The seventh row is the birth weight scale, with a value range of 570 - 2690 g and corresponding scores of 0 - 3.4 points;

[0097] The eighth row is the gestational age at birth scale, with a value range of 23.29 - 31.86 and corresponding scores of 0 - 2.7 points.

[0098] Obtain the single - item scores above each index, and add them up to get the total score. Then, the predicted probability value corresponding to the total score below is the probability that the model predicts the patient has BPD. For example, a single - fetus premature infant with a gestational age at birth of 30 weeks and a birth weight of 1100 g, the corresponding scores (the score corresponding to gestational age is 0.7 points, and the score corresponding to weight is 2.6 points), without cervical cerclage before delivery (corresponding score is 0 points), given CPAP assisted ventilation for 11.8 days (corresponding score is 1.1 points) and then changed to high - flow assisted ventilation and successfully weaned off oxygen after 10 days, without complications of pneumothorax (corresponding score is 0 points) and pulmonary hemorrhage (corresponding score is 0 points) during the course, and without using invasive ventilation (corresponding score is 0 points), then the total score of this patient added up in the model is 4.4, and the corresponding predicted probability value is 0.15. Another example, a single - fetus premature infant with a gestational age at birth of 27 + 4 weeks and a birth weight of 800 g, the corresponding scores (the score corresponding to gestational age is 1.3 points, and the score corresponding to weight is 3 points), without cervical cerclage before delivery (corresponding score is 0 points), given invasive assisted ventilation for 35.4 days (the corresponding score for invasive is 1.4 points, the corresponding score for the initial ventilation time is 3.2 points, and the corresponding score for invasive ventilation is 4.5 points) and then changed to nCPAP assisted ventilation, with pneumothorax occurring on the 5th day after birth (corresponding score is 1 point), without complications of pulmonary hemorrhage (corresponding score is 0 points), then the total score of this patient added up in the model is 14.4, and the corresponding predicted probability value is 0.97. After converting the prediction model into a nomogram, it is more simple, intuitive, and convenient for clinical application.

[0099] Refer to Figure 2 , according to an embodiment of the present invention, after constructing the nomogram prediction of bronchopulmonary dysplasia in premature infants through the above steps, the constructed model can also be verified. After step S5, the following steps are further included:

[0100] S6: Obtain one or more groups of validation data sets;

[0101] S7: According to the validation dataset, evaluate the constructed nomogram prediction model for preterm infants with BPD using any one or more of the ROC curve, calibration curve, or DCA curve evaluation curves.

[0102] For example, the constructed nomogram prediction model for preterm infants with BPD can be evaluated by the ROC curve and the area under the ROC curve (AUC) can be calculated to evaluate the ability of the model to distinguish BPD from the control group (as Figure 6A shown), and the result is that the AUC is 0.83 (95% CI 0.80–0.86). It is verified using the internal validation set (as Figure 6B shown), and it can be seen that it is 0.87 (95% CI 0.83–0.91). At the same time, it is verified using the external validation set (as Figure 6C shown), and it can be seen that it is 0.96 (95% CI 0.93–0.99). All AUC values exceed 0.8, indicating that the model has good discrimination ability. The positive predictive values of the training set, internal validation set, and external validation set are 0.713, 0.727, and 0.996 respectively, while the negative predictive values are 0.836, 0.836, and 0.929 respectively. The cut-off value of the ROC curve calculated according to the optimal Youden index is 0.215. The above results show that the constructed nomogram prediction model for preterm infants with BPD in the present invention has good predictive efficacy.

[0103] For example, the deviation between the risk prediction value and the actual observed value of the constructed nomogram for preterm infants with BPD can be evaluated by plotting a calibration curve. In terms of model calibration, the predicted values of the training set, internal validation set, and external validation set are all close to the diagonal line. The p-values obtained from the Hosmer-Lemeshow goodness-of-fit test are 0.579, 0.426, and 0.901 respectively, all exceeding 0.05. Similarly, the Brier scores of these sets are 0.128, 0.129, and 0.06 respectively, all lower than 0.25. These results together indicate that the model shows good consistency between the actual and predicted incidences. It can be seen that there is good consistency between the two (as Figure 7A , Figure 7B , Figure 7C shown).

[0104] For example, the clinical utility of the constructed nomogram prediction model for preterm infants with BPD can be evaluated by plotting DCA. In terms of clinical relevance, the DCA decision curve analysis shows that there are significant net benefits in the training set, internal validation set, and external validation set. It can be seen that the model can achieve better net benefits (as Figure 8A , Figure 8B , Figure 8C shown).

[0105] Through the above verification, the following conclusion can be drawn: The preterm BPD prediction model constructed by the present invention based on 8 perinatal indicators including gestational age at birth, birth weight, whether cervical cerclage was performed prenatally, whether pneumothorax was complicated, whether pulmonary hemorrhage was complicated, initial ventilation duration, whether invasive ventilation was performed and the duration of invasive ventilation has high application value for predicting the risk of BPD in preterm infants with a gestational age less than 32 weeks. Moreover, after being visually converted into a nomogram, it has greater clinical practicability.

[0106] In a second aspect, an embodiment of the present invention further provides a device for constructing a preterm bronchopulmonary dysplasia prediction model, including a computer device. The computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above method for constructing a preterm bronchopulmonary dysplasia prediction model is implemented.

[0107] The device for constructing a preterm bronchopulmonary dysplasia prediction model provided by the embodiment of the present invention constructs a preterm BPD prediction model based on 8 perinatal indicators including gestational age at birth, birth weight, whether cervical cerclage was performed prenatally, whether pneumothorax was complicated, whether pulmonary hemorrhage was complicated, initial ventilation duration, whether invasive ventilation was performed and the duration of invasive ventilation. The constructed model has high application value for predicting the risk of BPD in preterm infants with a gestational age less than 32 weeks.

[0108] Refer to Figure 3 In a third aspect, an embodiment of the present invention further provides a preterm bronchopulmonary dysplasia prediction device. The preterm bronchopulmonary dysplasia prediction device includes a computer device. The computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-constructed preterm bronchopulmonary dysplasia prediction model is implemented.

[0109] This BPD prediction model device only requires perinatal factors and is not affected by clinical management factors. It has good prediction efficacy in predicting BPD in preterm infants under the new diagnostic criteria, and has high sensitivity and specificity. After being verified by internal and external data, it is helpful for clinical diagnosis. Moreover, after the prediction model is converted into a nomogram, it is more simple and intuitive to use, facilitating clinical application.

[0110] The described computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0111] It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice within the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0112] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For the parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0113] The steps in the method embodiments of the present invention can be adjusted, combined, and deleted according to actual needs.

[0114] The modules or units in the system embodiments of the present invention can be combined, divided, and deleted according to actual needs.

[0115] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic preset hardware, or a combination of computer software and electronic preset hardware. Whether these functions are executed in the form of preset hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0116] In the embodiments provided by the present invention, it should be understood that the disclosed apparatus / computer device and method can be implemented in other ways. For example, the apparatus / computer device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the apparatus or unit can be in electrical, mechanical or other forms.

[0117] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0118] In addition, it should be noted that, in the above specific embodiments, the various specific technical features described can be combined in any suitable way without conflict. To avoid unnecessary repetition, the present invention will not separately describe various possible combination methods.

[0119] Furthermore, any combination can be made between different embodiments of the present invention as long as it does not violate the idea of the present invention, and it should also be regarded as the content disclosed by the present invention.

Claims

1. A method for constructing a prediction model for bronchopulmonary dysplasia in premature infants, characterized in that: Includes steps: S1: Acquire characteristic data of premature infants with a gestational age of less than 32 weeks, wherein the characteristic data includes perinatal maternal information data, perinatal information data, and neonatal information data; S2: Univariate logistic regression was used to perform regression analysis on the acquired characteristic data, and univariate predictive variables associated with bronchopulmonary dysplasia in premature infants were preliminarily screened out. Univariate predictive variables with P < 0.05 were included in the Lasso regression analysis to screen out independent predictive factors associated with bronchopulmonary dysplasia. S3: The screened independent predictors were included in the multivariate logistic regression, and a multivariate logistic regression model was constructed by forward stepwise regression, backward stepwise regression, and bidirectional stepwise regression methods. The model with the smallest AIC value was selected based on the Akaike information criterion, and multiple variables in the model with the lowest AIC value were used to construct a prediction model for bronchopulmonary dysplasia in premature infants.

2. The method for constructing a prediction model for bronchopulmonary dysplasia in premature infants according to claim 1, characterized in that: The perinatal maternal information data include: any one or more of age, race, weight gain during pregnancy, gestational hypertension, gestational diabetes, intrahepatic cholestasis of pregnancy, thyroid dysfunction during pregnancy, chorioamnionitis, autoimmune disease, premature rupture of membranes, Streptococcus agalactiae infection or Ureaplasma urealyticum infection; The perinatal information data include: any one or more of the following: antenatal steroid use, antenatal magnesium sulfate use, antibiotic use within 24 hours before delivery, whether cervical cerclage was performed before delivery, antenatal fetal medical intervention, mode of conception, multiple births, mode of delivery, 1-minute Apgar score, 5-minute Apgar score, or neonatal asphyxia; The neonatal information data include: gender, gestational age, birth weight, whether small for gestational age, ratio of birth weight to gestational age, neonatal respiratory distress syndrome, use of pulmonary surfactant, persistent pulmonary hypertension of the newborn, whether pneumothorax is present, whether pulmonary hemorrhage is present, first respiratory support mode and duration, whether invasive respiratory support is used and the time, any one or more of early-onset sepsis and hemodynamically significant patent ductus arteriosus, duration of antibiotic use, caffeine use or occurrence of bronchopulmonary dysplasia.

3. The method for constructing a prediction model for bronchopulmonary dysplasia in premature infants according to claim 2, characterized in that: The variables used to construct the prediction model for bronchopulmonary dysplasia in premature infants included gestational age, birth weight, whether cervical cerclage was performed before delivery, whether pneumothorax was present, whether pulmonary hemorrhage was present, initial ventilation duration, whether invasive ventilation was used, and the duration of invasive ventilation.

4. The method for constructing a prediction model for bronchopulmonary dysplasia in premature infants according to claim 3, characterized in that: The prediction formula of the bronchopulmonary dysplasia prediction model for premature infants is: P=1 / (1+exp(-(3.935-0.176*ga-0.001*bw+0.717*cerclageofcervix+0.715*pntx+0.624*pnmh+0.813*imv+0.071*imvtime+0.050*irstime))); Among them, P represents the predicted probability value, ga represents the gestational age at birth, and bw represents the birth weight. cerclageofcervix indicates prenatal cervical cerclage, pntx indicates pneumothorax, pnmh indicates pulmonary hemorrhage, imv indicates whether invasive ventilation is required, imvtime indicates the time of invasive ventilation, and irstime indicates the duration of initial ventilation.

5. The method for constructing a prediction model for bronchopulmonary dysplasia in premature infants according to claim 4, characterized in that: After step S3, the method further includes the following steps: S4: A nomogram prediction model is obtained based on the bronchopulmonary dysplasia prediction model for premature infants.

6. The method for constructing a prediction model for bronchopulmonary dysplasia in premature infants according to claim 5, characterized in that: The prediction formula of the nomogram prediction model is: Total score = score corresponding to invasive ventilation time + score corresponding to initial ventilation time + score corresponding to invasive ventilation + score corresponding to pulmonary hemorrhage + score corresponding to pneumothorax + score corresponding to cervical cerclage + score corresponding to birth weight + score corresponding to gestational age at birth; Among them, the probability corresponding to the total score is the risk probability of bronchopulmonary dysplasia in premature infants.

7. The method for constructing a prediction model for bronchopulmonary dysplasia in premature infants according to claim 6, characterized in that: The variables constitute eight scales; among them, The first scale is the scale of invasive respiratory support time, with a range of 0-70.17 days and a corresponding score of 0-8.8 points; The second scale is the initial respiratory support time scale, with a range of 0-59 days and a corresponding score of 0-5.3 points; The third scale is whether invasive ventilation is used, with a value range of 0-1, corresponding to a score of 0-1.4 points; The fourth scale is whether there is pulmonary hemorrhage, with a value range of 0-1, corresponding to a score of 0-1.2 points; The fifth scale is whether there is pneumothorax, with a value range of 0-1, corresponding to a score of 0-1.3 points; The sixth scale is whether cervical cerclage was performed before delivery, with a value range of 0-1, corresponding to a score of 0-1.3 points; The seventh scale is the birth weight scale, which ranges from 570 to 2690 g and corresponds to a score of 0 to 3.4 points; The eighth scale is the gestational age at birth scale, with a value range of 23.29-31.86 and a corresponding score of 0-2.7 points.

8. The method for constructing a prediction model for bronchopulmonary dysplasia in premature infants according to any one of claims 5 to 7, characterized in that: Step S5 is followed by the following steps: S5: Obtain one or more verification data sets; S6: Based on the validation data set, use any one or more evaluation curves including ROC curve, calibration curve or DCA curve to evaluate the constructed bronchopulmonary dysplasia prediction model for premature infants.

9. A device for predicting bronchopulmonary dysplasia in premature infants, comprising a computer device, wherein the computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the prediction model for bronchopulmonary dysplasia in premature infants constructed according to any one of claims 1 to 8 is implemented.

Images