ABSTRACT
Objective:To develop and validate a predictive model for adverse outcomes in women with hypertensive disorders of pregnancy (HDP).Methods:We retrospectively analyzed the clinical data of patients diagnosed with HDP and delivered at the First Affiliated Hospital of Soochow University or Sichuan Provincial Maternity and Child Health Care Hospital between May 1, 2011, and April 30, 2019. These patients were categorized as the adverse outcome group or the control group with adverse outcomes within 48 h after admission. Univariate analysis, least absolute shrinkage, selection operator (LASSO), and multivariable logistic regression were employed to analyze factors influencing the adverse outcomes and develop a predictive model. The receiver operating characteristic (ROC) curve and calibration plot was used to assess the predictive performance. Bootstrapping was used for the internal validation and the retrospective dataset of patients with HDP from the First Affiliated Hospital of Soochow University from May 1, 2019, to April 30, 2020, for the external validation. A graphic nomogram was created through R software based on the model.Results:(1) Of the 2 978 HDP patients who were included in the development set, 356 were in the adverse outcome group, accounting for 12.0%; of the 233 patients who were included in the external validation set, 40 presented with adverse outcomes within 48 h after admission, accounting for 17.2%. (2) Nine optimal predictors were identified based on the LASSO regression analysis and multivariable logistic regression, consisting of gestational age on admission, routine prenatal care, number of symptoms, mean arterial pressure, platelet count, fibrinogen, albumin, serum urea, and serum creatinine, based on which the logistic predictive model was established. (3) The ROC curve for this predictive model achieved an area under the curve (AUC) of 0.878 (95% CI: 0.858-0.897), and the ideal cut-off value for predicted probability was 0.136, with a sensitivity of 0.778 (95% CI: 0.731-0.820) and specificity of 0.848(95% CI: 0.834-0.862). The model was well-calibrated as the Hosmer-Lemeshow test showed that P>0.05. The calibration plot of the model had a slope of 1 and an intercept of 0. (4) The model showed good consistency in the internal validation and had an AUC of 0.872 (95% CI: 0.807-0.937) in the external validation. The Hosmer-Lemeshow test showed that the P value was >0.05, and the calibration slope was 1.001. (5) A nomogram was constructed for convenient clinical use. Conclusion:A relatively accurate prediction model for adverse outcomes in HDP patients was established, which could be used as a valuable quantitative tool for assessing HDP-related complications.
ABSTRACT
Objective:To establish a model for predicting cesarean delivery after failure of trial of labor among low-risk term primipara.Methods:This study retrospectively analyzed the clinical data of low-risk primiparas, with singleton cephalic full-term fetus, who delivered in the Department of Obstetrics and Gynecology of the First Affiliated Hospital of Soochow University from January 1, 2011 to August 31, 2017. Women experienced cesarean delivery(CS) following failed trial of labor were grouped as CS group, while those successfully delivered normally as vaginal delivery group(VD group). Chi-square test, t-test and multivariate logistic regression analysis were used for statistical analysis. Influencing factors of CS after a failed trial of labor were screened to establish the prediction model. The area under the receiver operating characteristic curve (AUC) and Hosmer-Lemeshow goodness-of-fit test were used to assess the performance of the model. A nomogram was established using R programming language based on the predictive model. Results:(1) This study recruited 6 551 subjects and among them, 576 (8.8%) women experienced CS after a failed trial of labor and the rest 5 975(91.2%) delivered vaginally. (2) The women in CS group were older [(27.5±3.1) vs (26.8±3.0) years, t=-4.963, P<0.01] and shorter in height [(159.5±4.2) vs (161.7±4.6) cm, t=11.548, P<0.01] , had higher pre-pregnancy body mass index (BMI) [(21.5±2.6) vs (20.8±2.5) kg/m 2, t=-6.743, P<0.01] and higher weight gain during pregnancy [(14.8±4.2) vs (14.1±4.2) kg, t=-3.446, P<0.01] and delivered later [(282±7) vs (278±7) d, t=-10.499, P<0.01] compared with those in VD group. The incidence of premature rupture of membranes (PROM) [26.4% (152/576) vs 20.7% (1 238/5 975) , χ2=10.101, P<0.01], labor induction [oxytocin: 26.4% (152/576) vs 16.3% (976/5 975), artificial rupture of membranes: 46.5% (268/576) vs 36.6% (2 189/5 975), application of cervical dilator balloon: 2.6% (15/576) vs 1.1% (65/5 975) and Propess: 4.7% (27/576) vs 2.5% (149/5 975), χ2=134.918, P<0.01], and the proportion of cases with meconium-stained amniotic fluid [ Ⅰ: 5.2% (30/576) vs 3.5% (209/5 975), Ⅱ: 5.7% (33/576) vs 2.5% (150/5 975), Ⅲ/bloody: 13.7% (79/576) vs 1.8% (105/5 975), χ2=307.664, P<0.01] were all higher in CS group than in VD group. There were more male infants [58.0% (334/576) vs 49.1% (2 934/5 975), χ2=16.576, P<0.01] and higher neonatal birth weight [(3 528±389) vs (3 344±368) g, t=-11.431, P<0.01] in the CS group as well. (3) Multivariate logistic regression analysis showed that maternal age and height, pre-pregnancy BMI, weight gain during pregnancy, gestational age at delivery, PROM, labor induction with oxytocin, artificial rupture of membrane, application of cervical dilator balloon and Propess, meconium-stained amniotic fluid, and fetal gender were all independent factors for CS. Two prediction models and nomograms were established according to fetal gender was involved or not. (4) The AUC of the prediction model not involving fetal gender was 0.774 (95% CI: 0.763-0.784) and the cut-off value was >8.7% with the sensitivity and specificity of 0.707 and 0.706, while that involving fetal gender was 0.782 (95% CI: 0.771-0.791) with the sensitivity and specificity of 0.785 and 0.645, respectively, when the cut-off value was >7.4%. The Hosmer-Lemeshow goodness-of-fit test showed that the two models fitted well (both P>0.05). Results of the internal validation using Bootstrap method indicated that the CS rates predicted by both models were consistent with the real data. Conclusions:The established models could effectively and accurately predict CS in term, singleton, cephalic, and low-risk primipara after failure of trial of labor, which might be a tool for clinicians to inform pregnant women to choose an appropriate delivery mode, thus improving maternal and infant outcomes.