| BackgroundPlacental implantation is a globally recognized obstetric complication..Accurate assessment,early diagnosis,and multidisciplinary collaboration are key to reducing adverse outcomes.There is a lack of relevant prognostic models for systematically evaluating perinatal outcomes of placental implantation,and they fail to suggest the need for preventive treatment based on the actual severityObjectiveBased on data from 1003 cases of patients with placenta implatation,analysis of the clinical features and clinical outcomes,clear influence the outcomes in patients with placental implantation perioperative related factors,explore the multidisciplinary cooperation security system for improving the perinatal outcomes of patients with placental implantation application value,and further build the prognosis of perinatal outcome in patients with placenta mplantation prediction model and optimize it..Materials and methodsCollected between January 2009 and December 2018,the guangzhou severe maternal treatment center hospital cesarean delivery and discharge diagnosis for patients with more than 28 weeks of single fetal placental implantation data of 1003 cases were retrospectively analyzed,depending on the type of implanted into disaster,based on the safety of the multidisciplinary cooperation system is divided into 2009-2014 and 2015-2018,to explore the safety of the types of placental implantation and multidisciplinary cooperation system’s impact on perinatal outcome To analyze the correlation between the general situation of pregnant women.multidisciplinary teams,surgical methods and hysterectomy,to explore the application value of posterior hysterectomy in placenta previa with implantation,and to build a prognosis prediction model for perinatal outcomes of placenta previa patients based on traditional model and machine learning model.If the measurement data were normally distributed,they were expressed in the form of mean±standard deviation.T-test was used for the difference comparison between the two groups,and analysis of variance was used for the difference comparison between the three groups.If normal distribution is not followed,median and quartile were used.Mann-whitney U test was used for the difference comparison between the two groups,and Kruskal-Wallis H test was used for the difference comparison between the three groups.Enumeration data were expressed by case number and composition ratio,and comparison of composition ratio between two or more groups was performed by 2 test.Univariate and multivariate logistic regression models were used for correlation factor analysis.Multiple interpolation method was used to generate data for missing values for analysis.75%of the population was randomly selected for modeling and 25%of the population for verification.A prognostic model for perinatal outcomes of placenta accreta patients was established using four traditional methods and machine learning.The predictive visual nomogram of the prognostic risk of placenta accreta was generated based on traditional methods.C statistic was used to evaluate the distinguishing ability of prediction model.The Calibration curve is adopted to demonstrate the Calibration ability of the model.R software and SPSS software were used for data processing and analysis.All statistical tests were bilateral,and P<0.05 was considered statistically significantResults1.Analysis of clinical features and outcomes based on 1003 cases of placental implantation:10-year results from a single centerThe incidence of placental implantation was 1.6%over 10 years.It rose from 0.78%in 2009-2014 to 2.35%in 2015-2018.78.66%of the patients induced placenta previa,and 62.01%had pernicious placenta previa with implantation,accounting for 51.20%of the patients with central placenta previa.Age,advanced age,parity and cesarean section increased gradually(P<0.05).The hospital stay,average postpartum blood loss,ICU admission rate and urinary system injury,NICU admission rate and neonatal mortality rate were decreased year by year,while the average gestational age and birth weight increased year by year.The risk factors of pernicious placenta previa with implantation were more significant(P<0.05),and the adverse perinatal outcomes were more serious(P<0.05)The application of the clinical treatment technology had a trends of "Λ" type in 2014,and the posterior uterine repair relative to hysterectomy was rising year by year.2.Analysis of related factors affecting perinatal outcomes of placentalimplantation patientsUnivariate analysis showed that:gestational week≥32 weeks after admission,mode of conception:IVF-ET,non-preplacenta implantation,non-pernicious placenta previa with implantation significantly reduced postpartum blood loss,gravidity ≥3 times,parity≥1 time,history of induced abortion,history of cesarean section significantly increased postpartum blood loss;Multidisciplinary team management,IVF-ET conception,hospitalized gestational weeks≥32 weeks,non-placenta previa implantation,non-perniciou placenta previa with implantation significantly reduced the risk of hysterectomy,gravidity≥3 times,parity≥1 time,history of induced abortion,and history of cesarean section significantly increased the risk of hysterectomy.Multidisciplinary team management,unnatural pregnancy,induced abortion≥3 times,gestational age≥32 weeks in hospital,gestational age≥33 weeks in labor,non-placenta previa implantation,non-dangerous placenta previa with implantation significantly reduced the length of hospital stay,parity≥2 times,and cesarean section history significantly increased the length of hospital stayMultivariate analysis showed that multidisciplinary team management and non-invasive placenta previa with implantation were independent protective factors for hysterectomy,and age≥35 years was independent risk factor for hysterectomy.Multidisciplinary team management,unnatural conception,and gestational week≥32 weeks were independent factors for reducing the length of hospital stay of pregnant women,and gestational week≥33 weeks were independent factors for increasing the length of hospital stay of pregnant women3.Multidisciplinary cooperative safety management system can improve perinatal outcomes of placental implantion patientsUnivariate and multivariate analyses showed that the risk of hysterectomy,bladder/ureter injury,length of hospital stay,and ICU admission were significantly lower in the multidisciplinary safety management system than in the control group(P<0.05).Further subgroup analysis shows that the results are robust and reliable.4 The value of posterior hysterectomy in placenta previa with implantationGravity,parity,abortions and cesarean sections in the posterior hysterectomy group was lower than that in the hysterectomy group,and the average gestational age of delivery was higher than that in the hysterectomy group The perinatal outcomes of the patients with placenta previa with implantation were better than those of the hysterectomy group(P<0.05)5 Based on the traditional model and machine learning model,the prediction model of perinatal outcomes of placental implantation patients was constructedDescription of population characteristics of modeling group and verification group;A total of 750 patients with placental implantation were included in the modeling cohort,among whom 37.20%were advanced pregnant women,51.87%had a cesarean section,80.53%had placenta previa,62.93%had pernicious placenta previa,and the incidence of postpartum hemorrhage was 54.40%.A total of 253 subjects who met the inclusion criteria were included in the verification cohort Among them,38.34%were elderly pregnant women,56.13%had a cesarean section,73.12%had placenta previa,59.29%had pernicious placenta previa,and 50.20%had postpartum hemorrhage.There was no difference in the distribution of clinical features between the two groupsSingle factor analysis results of modeling group and validation group;Hospital in the model group:year,gravidity,parity,induced abortion(1,2),cesarean sectiontimes,hospital gestational age(>39 weeks),prenatal diagnosis,placenta previa(central),pernicious placenta previa,cystoscope+tube,uterus repair,postpartum hemorrhage,uterine artery ligation are associated with significant risk for hysterectomy.2015-2018 versus 2009-2014,》39 weeks versus<32 weeks,prenatal diagnosis,uterine artery ligation,and uterine repair significantly reduced the risk of hysterectomy(P<0.05).The risk of hysterectomy was significantly increased with the increase of gravidity,parity,abortion,cesarean section,placenta previa and pernicious placenta previa(P<0.05).In the verification group,gravidity,parity,cesarean section,placenta previa(central type),pernicious placenta previa,cystoscope+catheterization,uterine repair,and postpartum hemorrhage were all significantly associated with the risk of hysterectomy.The risk of hysterectomy was significantly increased with increased number of gravidity,parity,and cesarean sections,hypertensive disorders during pregnancy,and postpartum hemorrhage(P<0.05)Four models were constructed by traditional methods,namely(1)complete model,(2)stepwise regression model,(3)Bootstrap complete model and(4)bootstrap stepwise regression model.Based on the above prognostic factors,the predictive nomogram of the prognostic risk of hysterectomy for placental implanted patients was successfully constructed.The internal verification results showed that the C statistics of the prediction differentiation ability of the four model prediction models for hysterectomy were 0.893,0.888,0.893.0.888 respectively in the modeling queue and the verification queue:0.838,0.833,0.837,0.834 respectively,showing that the model had good differentiation ability.The calibration curve results show that the prediction model has high accuracy.The prognostic prediction model was successfully constructed by machine learning,and the prediction differentiation ability C statistic was 0.9133 and 0.8363 respectively in the modeling group and the verification group.Compared with traditional models,machine learning models are more discriminative,more concise and more conducive to clinical application.Conclusion1.The incidence of placenta accreta is increasing year by year,and the main type is the dangerous placenta previa with implantation.Its high risk factor also rises,and perinatal adverse outcome decreases year by year.2.The main risk factors of perinatal outcome were cesarean section and central placenta previa.3.A multidisciplinary cooperation safety system can effectively reduce adverse perinatal outcomes,and posterior hysterectomy can effectively reduce hysterectomy during placenta previa with implantation.4.Accurate and reliable prognostic prediction models for perinatal outcomes of placenta accretion patients were successfully constructed through traditional methods and machine learning.The model constructed by machine learning was more concise,and prospective studies were needed to further verify the clinical practicality and continuously improve the model. |
PDF Full Text Request