Determinants of labor epidural analgesia uptake and associations with maternal-neonatal outcomes: a stratified cohort study with risk prediction modeling

Abstract

Background

Labor pain is a major physiological and psychological stressor for women during childbirth. Epidural analgesia is widely recognized as the most effective form of pain relief in labor, yet its utilization varies greatly, and uncertainty remains regarding its determinants and potential impact on maternal and neonatal outcomes. This study aimed to identify factors influencing the use of epidural analgesia and to evaluate its associations with labor progression and postpartum hemorrhage.

Methods

This retrospective cohort included 1035 women who delivered at our hospital between 2020 and 2023. Univariable and multivariable logistic regression analyses were performed to identify independent determinants of epidural analgesia use. A clinical prediction model was developed and compared with eleven machine learning classifiers using receiver operating characteristic analysis. Associations between epidural analgesia and maternal-neonatal outcomes, particularly postpartum hemorrhage, were evaluated.

Results

Younger maternal age, primiparity, and lower pre-delivery fibrinogen levels were independently associated with the use of epidural analgesia. The logistic regression model demonstrated good discrimination (AUC = 0.70, 95% CI: 0.66–0.75), comparable to that of the machine learning models. Among vaginal deliveries, epidural analgesia was associated with longer total, first-stage, and second-stage labor durations (P < 0.001) but did not increase the incidence of postpartum hemorrhage or adverse maternal-neonatal outcomes. Across the entire cohort, higher pre-delivery D-dimer (OR = 1.54, 95% CI: 1.08–2.20, P = 0.017) and fibrinogen (OR = 1.34, 95% CI: 1.07–1.67, P = 0.010) levels were independent risk factors for postpartum hemorrhage, whereas epidural analgesia itself was not.

Conclusions

Maternal age, parity, and coagulation status are key factors influencing the choice of epidural analgesia. Although epidural analgesia prolongs labor, it does not increase postpartum hemorrhage or other adverse outcomes. These findings support the safety of epidural analgesia and highlight the importance of assessing coagulation function to improve individualized labor management and prevent postpartum complications.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-026-08883-y.

Background

Labor pain represents one of the most intense physiological and psychological experiences, with profound implications for both maternal well-being and perinatal outcomes. Effective pain management during labor not only improves maternal comfort and satisfaction but also reduces stress-related neuroendocrine responses that may influence uterine contractility and fetal oxygenation [1, 2]. Emerging evidence suggests that managing pain through a combination of epidural analgesia, movement, water immersion, and relaxation techniques can significantly reduce tension and anxiety, thereby facilitating more efficient labor progression [3, 4]. Among available options, epidural analgesia (EA) is recognized as one of the most effective techniques, consistently demonstrated to markedly relieve pain and enhance maternal satisfaction, in combination with multimodal strategies such as systemic opioids, relaxation training, and manual interventions [5].

Nevertheless, global utilization of EA remains heterogeneous, varying widely by geographic region, institutional resources, and sociocultural norms [6, 7]. Despite guideline support for broader use when indicated, in many regions the rate of EA use remains below 30%, partly due to limited anesthesiology capacity and misconceptions about its risks [8]. Even in high-resource settings, maternal decision-making is influenced by factors such as age, parity, prior delivery experience, and educational background [9]. Understanding these determinants is essential for optimizing individualized counseling and promoting equitable access to obstetric anesthesia services.

Beyond pain control, the broader obstetric and neonatal safety profile of EA has been the subject of ongoing debate. Earlier studies associated EA with prolongation of the first and second stages of labor and with higher rates of operative vaginal delivery, as well as an increased risk of maternal intrapartum fever and lower neonatal Apgar scores [10–12]. In contrast, more recent large-scale studies indicate that EA does not significantly increase cesarean delivery rates and is not detrimental to short-term neonatal outcomes [13, 14]. This evolution in evidence reflects technical advances in EA, lower concentrations of local anesthetics, and improved labor management practices.

Postpartum hemorrhage (PPH), the leading direct cause of maternal mortality worldwide, remains a major public health challenge despite advances in obstetric care [15]. Given its complex and multifactorial etiology involving uterine atony, trauma, and coagulopathy, potential iatrogenic contributors including labor analgesia have drawn increasing scrutiny [16]. Although evidence from large population-based studies indicates that EA does not increase PPH risk among women with uncomplicated pregnancies, recent data suggest that it may be associated with higher blood loss or PPH incidence in specific high-risk groups, particularly those with gestational hypertension or pre-eclampsia [17]. Conversely, some studies have reported potential hemodynamic and stress-modulating benefits of EA that could help reduce bleeding in selected populations [18]. Clarifying the causal relationship between EA and PPH thus represents a current research priority in obstetric anesthesia.

In the present study, using a large clinical cohort, our primary objective was to systematically identify the independent determinants of EA uptake and to develop a robust clinical prediction model for EA selection. As a secondary but essential objective, we evaluated the associations of EA with labor progression and maternal-neonatal outcomes, with a specific focus on PPH, to provide the necessary safety evidence to support the implementation of our predictive findings. By centering our investigation on these determinants while validating their safety outcomes, these findings aim to provide actionable evidence for rational and individualized use of EA, thereby informing clinical decision-making and improving the safety and experience of delivery.

Methods

Design and participants

This retrospective cohort included 1035 consecutive women who delivered at our hospital between January 1, 2022 and December 31, 2023 with complete clinical data. Participants were categorized by the use of EA into EA and non-EA groups, and further stratified by final mode of delivery into vaginal delivery and intrapartum transfer to cesarean section (no elective cesarean deliveries were included). Usual intrapartum care pathways and analgesia protocols were maintained throughout and were not altered for the study. The study adhered to the Declaration of Helsinki and was approved by the Institutional Review Board of Jinan Maternal and Child Health Hospital (approval No. 2023–1–038). Written informed consent was obtained from all participants.

Inclusion and exclusion criteria

Inclusion criteria were: (1) singleton pregnancy; (2) gestational age ≥ 28 weeks; and (3) complete clinical records. Exclusion criteria were: (1) elective cesarean delivery; (2) multiple gestation; (3) major fetal congenital anomalies; and (4) contraindications to EA such as severe cardiac, hepatic, or renal dysfunction, or coagulopathy. In total, 1035 women met the criteria and were included in the final analysis.

Epidural analgesia and delivery management

Labor analgesia was jointly managed by the departments of anesthesiology and obstetrics. Following admission, all parturients received standardized counseling regarding analgesia options, benefits, and risks. EA was performed for eligible and consenting women when cervical dilation reached ≥ 3 cm during the active phase of labor or at the obstetrician’s discretion. The epidural catheter was inserted at the L2-L3 or L3-L4 interspace by an experienced anesthesiologist under aseptic conditions. The initial loading dose consisted of 10 mL of 0.1% ropivacaine combined with sufentanil 0.3 μg/mL. Analgesia was subsequently maintained using a patient-controlled epidural analgesia (PCEA) regimen according to the hospital’s standardized protocol.

Labor progression was managed in strict accordance with institutional obstetric guidelines, and detailed records were kept for the duration of each labor stage and the mode of membrane rupture. For women who required intrapartum conversion to cesarean delivery because of labor arrest, fetal distress, malpresentation, intra-amniotic infection, or maternal request, surgical decisions were made by senior obstetricians after comprehensive clinical evaluation and in line with current recommendations. Operative details, including the indication for surgery, operative time, and estimated blood loss, were recorded systematically. Peripartum laboratory monitoring followed standard protocols and included measurements before and within 24 h after delivery of complete blood count, coagulation and fibrinolytic parameters, and liver function indices.

Data collection

Data were extracted from the electronic medical record, anesthesia information, and laboratory information systems, with two investigators independently abstracting and cross-checking all entries for consistency. Demographic and obstetric history variables included maternal age, parity, pre-pregnancy body mass index (BMI), use of assisted reproductive technology (ART), gestational age at delivery, and gestational weight gain. Pregnancy complications comprised hypertensive disorders of pregnancy, gestational diabetes, hypothyroidism, premature rupture of membranes, cord around the neck, and preterm birth. Pre-delivery laboratory indices included hemoglobin (HGB), white blood cell count (WBC), lymphocyte count (LYM), neutrophil count (NEU), platelet count (PLT), albumin (ALB), alanine aminotransferase (ALT), aspartate aminotransferase (AST), D-dimer, fibrinogen, and prothrombin time (PT). Fetal ultrasound parameters comprised biparietal diameter, head circumference, abdominal circumference, and femur length. Intrapartum process data included mode of membrane rupture, durations of each labor stage, reasons for intrapartum conversion to cesarean delivery, and intraoperative measures. Maternal-neonatal outcomes included postpartum blood loss, 1- and 5-min Apgar scores, neonatal intensive care unit (NICU) admission, neonatal sex, birth weight and length of hospital stay, as well as post-delivery laboratory reassessments. Logical inconsistencies and outliers were adjudicated by chart review before constructing the final analytic dataset.

Study outcomes

The primary objective of this study was to identify the independent determinants of EA uptake and to develop a clinical prediction model for EA selection. The variables considered for the model included maternal demographic and obstetric characteristics (age, parity, pre-pregnancy BMI, use of assisted reproduction, weight gain during pregnancy, and gestational age), as well as key obstetric comorbidities, such as hypertensive disorders of pregnancy (including gestational hypertension and preeclampsia), gestational diabetes mellitus, hypothyroidism, premature rupture of membranes, cord around the neck, and preterm birth, were treated as essential covariates in all predictive and outcome models. Additionally, pre-delivery laboratory parameters (HGB, WBC, LYM, NEU, PLT, ALB, ALT, AST, D-dimer, fibrinogen, and PT) and fetal ultrasound measurements (biparietal diameter, head circumference, abdominal circumference, and femur length) were included to capture the complete physiological and fetal status prior to delivery.

The secondary objective was to evaluate the safety profile of EA by assessing its impact on labor course and maternal-neonatal outcomes. These outcomes included: (1) labor duration (total, first stage, second stage, and third stage); (2) incidence of PPH, defined as blood loss ≥ 500 mL within 24 h after vaginal delivery or ≥ 1000 mL after cesarean delivery [19]; and (3) other maternal-neonatal endpoints, including neonatal Apgar scores at 1 and 5 min, NICU admission, length of hospital stay, and changes in laboratory parameters.

Statistical analysis

All analyses were performed using SPSS version 26.0, GraphPad Prism version 8.0, and R version 4.2.2. Normality of continuous variables was assessed with the Shapiro–Wilk test and visual inspection of histograms. Normally distributed data are presented as mean ± standard deviation and compared with the independent-samples t test; non-normally distributed data are presented as median (interquartile range) and compared with the Mann–Whitney U test. Categorical variables are summarized as counts (percentages) and compared using the χ² test or Fisher’s exact test, as appropriate. To identify independent determinants of EA use, variables significant in univariable analyses together with clinically relevant covariates were entered into a multivariable binary logistic regression model, with results reported as odds ratios (ORs) and 95% confidence intervals (CIs); effects were visualized with forest plots. The clinical prediction model was constructed based on the regression coefficients of the identified independent predictors, and discriminative performance was evaluated by the receiver operating characteristic (ROC) curve. For model development, the dataset was randomly divided into training, test, and validation subsets in a 7: 2: 1 ratio. The machine-learning classifiers were implemented using the caret and mlr3 packages in R. The conventional multivariable logistic model served as the baseline comparator against 11 machine-learning algorithms, including logistic regression (LR), random forest (RF), gradient boosting (GB), eXtreme Gradient Boosting (XGBoost), support vector machine (SVM), decision tree (DT), ExtraTrees, AdaBoost, Naïve Bayes, linear discriminant analysis (LDA), and quadratic discriminant analysis (QDA). Model hyperparameters were tuned using grid search and fivefold cross-validation within the training set to optimize performance and control for overfitting. Model discriminative performance was assessed by the area under the receiver operating characteristic curve (AUC), and visual comparison via bar plots across datasets. Furthermore, to provide a comprehensive evaluation of the clinical model’s classification ability, sensitivity and specificity were calculated, with the optimal cutoff value determined by the maximum Youden index. All tests were two-sided, and a P value < 0.05 was considered statistically significant.

Results

Baseline characteristics

A total of 1035 participants were included, of whom 752 (72.7%) received EA and 283 (27.3%) did not. Baseline characteristics are summarized in Table 1. The proportion of women aged ≥ 35 years was significantly lower in the EA group than in the non-EA group (8.78% vs. 20.14%, P < 0.001), whereas primiparity was more common in the EA group (69.81% vs. 44.88%, P < 0.001). There were no significant differences in pre-pregnancy BMI, use of ART, gestational weight gain, gestational age at delivery, hypertensive disorders of pregnancy, gestational diabetes, hypothyroidism, premature rupture of membranes, or cord around the neck (P > 0.05). Preterm birth was less frequent in the EA group (3.72% vs. 8.83%, P = 0.001). Pre-delivery laboratory indices and fetal ultrasound parameters were generally balanced between groups (P > 0.05).

Table 1.

Baseline characteristics of the participants

Characteristic	Total (n = 1035)	Epidural anesthesia (n = 752)	No epidural anesthesia (n = 283)	P-value
Age (year)				< 0.001*
< 35	912 (88.12)	686 (91.22)	226 (79.86)
≥ 35	123 (11.88)	66 (8.78)	57 (20.14)
Parity				< 0.001*
0	652 (63.00)	525 (69.81)	127 (44.88)
≥ 1	383 (37.00)	227 (30.19)	156 (55.12)
BMI (kg/m2)				0.193
< 30	784 (75.75)	578 (76.86)	206 (72.79)
≥ 30	251 (24.25)	174 (23.14)	77 (27.21)
Assisted reproduction				0.493
Yes	24 (2.32)	16 (2.13)	8 (2.83)
No	1011 (97.68)	736 (97.87)	275 (97.17)
Weight gain during pregnancy (kg)				0.106
< 15	447 (43.19)	313 (41.62)	134 (47.35)
≥ 15	588 (56.81)	439 (58.38)	149 (52.65)
Gestational age (day)	279 (273–283)	280 (274–283)	276 (269–281)	0.299
Obstetric complications
Hypertensive disorders of pregnancy	52 (5.02)	34 (4.52)	18 (6.36)	0.263
Gestational diabetes mellitus	201 (19.42)	143 (19.02)	58 (20.49)	0.598
Hypothyroidism	139 (13.43)	109 (14.49)	30 (10.60)	0.125
Premature rupture of membranes	324 (31.30)	247 (32.85)	77 (27.21)	0.084
Cord around the neck	359 (34.69)	270 (35.90)	89 (31.45)	0.188
Preterm birth	53 (5.12)	28 (3.72)	25 (8.83)	0.001*
Laboratory examination before delivery
HGB (g/L)	125 (118–132)	125 (118–132)	124 (118–133)	0.632
WBC (109/L)	8.98 (7.67–10.33)	8.93 (7.50–10.28)	9.17 (7.88–10.62)	0.332
LYM (109/L)	1.70 (1.42–2.05)	1.70 (1.42–2.05)	1.70 (1.42–2.05)	0.297
NEU (109/L)	6.60 (5.39–7.95)	6.58 (5.35–7.84)	6.70 (5.68–8.28)	0.537
PLT (109/L)	203 (173–239)	203 (174–239)	203 (173–239)	0.231
ALB (g/L)	36.70 (35.20–38.00)	36.70 (35.30–38.08)	36.60 (35.20–37.90)	0.877
ALT (IU/L)	11 (8–15)	11 (8–15)	10 (8–14)	0.715
AST (IU/L)	16 (14–18)	16 (14–19)	15 (13–18)	0.901
D-dimer (mg/L)	1.06 (0.74–1.54)	1.07 (0.75–1.53)	0.99 (0.73–1.57)	0.123
FIB (g/L)	4.31 (3.87–4.87)	4.31 (3.87–4.85)	4.38 (3.87–5.00)	0.909
PT (s)	11.10 (10.80–11.40)	11.10 (10.80–11.40)	11.10 (10.70–11.50)	0.649
Fetal obstetric ultrasound indicators
Biparietal diameter (mm)	93 (90–95)	93 (90–95)	92 (89–95)	0.249
Head circumference (mm)	330 (321–338)	330 (322–338)	328 (319.25–337.75)	0.174
Abdomen circumference (mm)	337 (323–350)	338 (324–351)	336 (321–349.75)	0.111
Femur length (mm)	71 (69–73)	71 (69–73)	71 (68–73)	0.911

Data are presented as median (interquartile range) or n (%)

P < 0.05 was considered statistically significant and indicated by an asterisk

Abbreviations: BMI Body mass index, HGB Hemoglobin, WBC White blood cell count, LYM Lymphocyte count, NEU Neutrophil count, PLT Platelet count, ALB Albumin, ALT Alanine aminotransferase, AST Aspartate aminotransferase, FIB Fibrinogen, PT Prothrombin time

Independent determinants of EA use

Univariable and multivariable logistic regression results are shown in Table 2. Univariable analysis showed that age, parity, gestational age at delivery, preterm birth, biparietal diameter, and head circumference were associated with EA use (P < 0.05; Fig. 1A). In the multivariable model, age (OR = 2.24, 95% CI: 1.36–3.70, P = 0.002), parity (OR = 2.54, 95% CI: 1.77–3.67, P < 0.001), and pre-delivery fibrinogen (OR = 0.84, 95% CI: 0.73–0.98, P = 0.023) remained independently associated with EA use (Fig. 1B).

Table 2.

Univariable and multivariate logistic regression analysis for epidural anesthesia

Characteristics	Univariable		Multivariate
	OR (95%CI)	P-value	OR (95%CI)	P-value
Age ≥ 35 vs. < 35 years	2.62 (1.78–3.85)	< 0.001*	2.24 (1.36–3.70)	0.002*
Parity ≥ 1 vs. 0	0.35 (0.27–0.47)	< 0.001*	2.54 (1.77–3.67)	< 0.001*
BMI ≥ 30 vs. < 30 kg/m2	1.24 (0.91–1.70)	0.174	1.29 (0.87–1.92)	0.209
Assisted reproduction vs. no	1.34 (0.67–3.16)	0.507	1.25 (0.37–4.18)	0.716
Weight gain during pregnancy ≥ 15 vs. < 15 kg	0.79 (0.60–1.04)	0.098	1.04 (0.73–1.49)	0.829
Gestational age (day)	1.03 (1.02–1.04)	< 0.001*	1.02 (1.00–1.05)	0.072
Hypertensive disorders of pregnancy vs. no	1.43 (0.80–2.58)	0.230	1.76 (0.85–3.64)	0.130
Gestational diabetes mellitus vs. no	1.10 (0.78–1.54)	0.592	1.04 (0.67–1.59)	0.883
Hypothyroidism vs. no	0.70 (0.46–1.08)	0.103	0.78 (0.46–1.32)	0.348
Premature rupture of membranes vs. no	0.76 (0.57–1.03)	0.082	0.88 (0.60–1.29)	0.510
Cord around the neck vs. no	0.82 (0.61–1.10)	0.180	0.95 (0.66–1.35)	0.758
Preterm birth vs. no	0.40 (0.23–0.70)	0.001*	1.00 (0.29–3.47)	0.995
HGB (g/L)	1.01 (1.00–1.02)	0.167	1.02 (1.00–1.03)	0.050
WBC (109/L)	0.95 (0.89–1.00)	0.068	0.99 (0.70–1.41)	0.956
LYM (109/L)	1.02 (0.78–1.34)	0.891	0.90 (0.53–1.51)	0.684
NEU (109/L)	0.94 (0.89–1.00)	0.061	0.96 (0.67–1.38)	0.834
PLT (109/L)	1.00 (1.00–1.00)	0.806	1.00 (1.00–1.01)	0.318
ALB (g/L)	1.04 (0.98–1.10)	0.192	0.97 (0.91–1.04)	0.364
ALT (IU/L)	1.00 (0.98–1.02)	0.903	1.02 (0.99–1.05)	0.259
AST (IU/L)	0.99 (0.98–1.01)	0.516	0.99 (0.96–1.02)	0.431
D-dimer (mg/L)	0.86 (0.74–1.01)	0.068	0.82 (0.67–1.01)	0.058
FIB (g/L)	0.89 (0.78–1.01)	0.067	0.84 (0.73–0.98)	0.023*
PT (s)	0.85 (0.66–1.09)	0.192	0.99 (0.72–1.37)	0.950
Biparietal diameter (mm)	1.05 (1.01–1.10)	0.008*	1.05 (0.96–1.14)	0.277
Head circumference (mm)	1.01 (1.00–1.03)	0.038*	0.99 (0.97–1.02)	0.634
Abdomen circumference (mm)	1.01 (1.00–1.01)	0.089	1.01 (0.99–1.02)	0.277
Femur length (mm)	1.04 (0.99–1.08)	0.121	0.95 (0.88–1.03)	0.199

P < 0.05 was considered statistically significant and indicated by an asterisk

The covariables in the multivariate regression models were maternal age, parity, BMI, assisted reproduction, weight gain during pregnancy, gestational age, hypertensive disorders of pregnancy, gestational diabetes mellitus, hypothyroidism, premature rupture of membranes, cord around the neck, preterm birth, pre-delivery HGB, WBC, LYM, NEU, PLT, ALB, ALT, AST, D-dimer, FIB, PT, biparietal diameter, head circumference, abdomen circumference and femur length

Abbreviations: OR Odds ratio, CI Confidence interval, BMI Body mass index, HGB Hemoglobin, WBC White blood cell count, LYM Lymphocyte count, NEU Neutrophil count, PLT Platelet count, ALB Albumin, ALT Alanine aminotransferase, AST Aspartate aminotransferase, FIB Fibrinogen, PT Prothrombin time

Fig. 1.

Univariable and multivariable logistic regression analyses of factors associated with epidural analgesia (EA) use. A Forest plot showing univariable logistic regression results for maternal and perinatal variables associated with EA use. B Forest plot showing multivariable logistic regression results. ORs are shown with 95% CIs. Abbreviations: OR, odds ratio; CI, confidence interval; BMI, body mass index; HGB, hemoglobin; WBC, white blood cell count; LYM, lymphocyte count; NEU, neutrophil count; PLT, platelet count; ALB, albumin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; FIB, fibrinogen; PT, prothrombin time

Clinical prediction model construction and performance

A clinical prediction model for EA selection was developed based on the identified independent determinants. The logistic regression-based model demonstrated an AUC of 0.70 (95% CI: 0.66–0.75, P < 0.001), indicating acceptable discrimination (Fig. 2). To provide a more granular assessment of performance, we calculated that the model achieved a sensitivity of 62.38% and a specificity of 71.08% at the optimal Youden index threshold.

Fig. 2.

ROC curve of the logistic regression-based clinical prediction model for epidural analgesia (EA) selection. The ROC curve of the multivariable logistic regression model shows an area under the curve (AUC) of 0.70 (95% CI: 0.66–0.75, P < 0.001), indicating acceptable discriminative ability for predicting the likelihood of EA use

To further evaluate the robustness of our clinical model, a comparative analysis was performed against 11 machine-learning classifiers using a 7:2:1 random split for training (n = 725), testing (n = 207), and validation (n = 103). While complex ensemble models, including Random Forest, XGBoost, and Decision Tree, achieved near-perfect discrimination on the training set (AUC = 1.000), their performance decreased on the independent test and validation sets, indicating a degree of overfitting (Fig. 3A-C). In comparison, the logistic regression-based clinical model demonstrated consistent stability across all datasets, with AUC values of 0.747, 0.638, and 0.617 for the training, test, and validation sets, respectively (Fig. 3D). On the final validation set, the AUC of our clinical model remained comparable to more complex algorithms, such as Random Forest (AUC = 0.703) and SVM (AUC = 0.641), while providing superior interpretability for clinical decision-making. These results suggest that the logistic regression-based clinical model serves as a balanced and reliable tool for predicting EA demand in a real-world obstetric setting.

Fig. 3.

Comparison of machine-learning model performance for predicting epidural analgesia (EA) use. A-C ROC curves of 11 supervised machine-learning classifiers across the training, test, and validation datasets, respectively. D Bar plot summarizing AUC values for all models across the three datasets

Maternal and neonatal outcomes by delivery pathway

To validate the clinical safety of the aforementioned determinants and the prediction model, we further evaluated the maternal and neonatal outcomes across different delivery pathways. Among the 889 women who delivered vaginally, the total duration of labor as well as the first and second stages were significantly longer in the EA group (n = 676) than in the non-EA group (n = 213) (P < 0.001). No significant differences were observed in the third stage or in the mode of membrane rupture (P > 0.05; Table 3). Postpartum laboratory testing showed higher WBC, NEU, D-dimer, fibrinogen, and PT levels in the EA group (P < 0.05). Regarding clinical outcomes, total hospital stays, pre-delivery stays, post-delivery stays, and neonatal birth weight were higher in the EA group (P < 0.05), whereas rates of PPH, neonatal sex, Apgar scores, placental weight, umbilical cord length, and NICU admission did not differ significantly (P > 0.05).

Table 3.

Impact of epidural analgesia on perinatal outcomes among women with vaginal delivery

Characteristic	Total (n = 889)	Epidural anesthesia (n = 676)	No epidural anesthesia (n = 213)	P-value
Total stage of labor (mins)	455.00 (318.50–610.00)	505.00 (380.00–660.00)	299.00 (222.50–405.00)	< 0.001*
First stage of labor (mins)	390.00 (270.00–540.00)	450.00 (320.00–570.00)	270.00 (190.00–360.00)	< 0.001*
Second stage of labor (mins)	45.00 (20.00–78.00)	58.00 (30.00–86.75)	18.00 (12.00–29.50)	< 0.001*
Third stage of labor (mins)	10.00 (10.00–10.00)	10.00 (10.00–10.00)	10.00 (10.00–10.00)	0.291
Mode of breaking membrane				0.129
Artificial membrane rupture	22 (2.47)	20 (2.96)	2 (0.94)
Natural rupture	867 (97.53)	656 (97.04)	211 (99.06)
Laboratory examination after delivery
HGB (g/L)	114 (106–122)	114 (106–121)	116 (105–123)	0.261
WBC (109/L)	10.49 (8.91–12.36)	10.75 (9.15–12.51)	9.53 (8.50–11.66)	< 0.001*
LYM (109/L)	1.87 (1.55–2.23)	1.89 (1.56–2.24)	1.80 (1.51–2.21)	0.323
NEU (109/L)	7.87 (6.52–9.54)	8.11 (6.70–9.67)	7.12 (6.02–8.65)	< 0.001*
PLT (109/L)	184.00 (154.00–215.00)	182.50 (153.00–214.00)	188.00 (158.50–218.50)	0.225
D-dimer (mg/L)	1.05 (0.71–1.64)	1.08 (0.75–1.67)	0.93 (0.64–1.54)	0.003*
FIB (g/L)	4.31 (3.79–4.85)	4.41 (3.87–5.00)	4.08 (3.64–4.70)	< 0.001*
PT (s)	11.30 (10.90–11.70)	11.30 (11–11.70)	11.20 (10.80–11.60)	0.034*
Postpartum hemorrhage	44 (4.95)	35 (5.18)	9 (4.23)	0.718
Total hospital stay days	4.00 (4.00–6.00)	5.00 (4.00–6.00)	4.00 (3.00–5.00)	< 0.001*
Pre-delivery hospital stay days	1.00 (0.00–2.00)	1.00 (1.00–2.00)	1.00 (0.00–1.00)	< 0.001*
Post-delivery hospital stay days	3.00 (3.00–4.00)	3.00 (3.00–4.00)	3.00 (3.00–4.00)	< 0.001*
Neonatal sex				0.813
Female	418 (47.02)	316 (46.75)	102 (47.89)
Male	471 (52.98)	360 (53.25)	111 (52.11)
Apgar score 1 min postpartum (points)				0.462
0–9	42 (4.72)	30 (4.44)	12 (5.63)
10	847 (95.28)	646 (95.56)	201 (94.37)
Apgar score 5 min postpartum (points)				0.634
0–9	6 (0.67)	4 (0.59)	2 (0.94)
10	883 (99.33)	672 (99.41)	211 (99.06)
Birth weight (g)	3300 (3100–3600)	3350 (3100–3600)	3300 (3000–3500)	0.012*
Placental weight (g)	590 (550–640)	600 (550–650)	580 (540–630)	0.109
Umbilical cord length (cm)	60 (55–60)	60 (55–60)	60 (55–60)	0.332
NICU	39 (4.39)	26 (3.85)	13 (6.10)	0.179

Data are presented as median (interquartile range) or n (%)

P < 0.05 was considered statistically significant and indicated by an asterisk

Abbreviations: HGB Hemoglobin, WBC White blood cell count, LYM Lymphocyte count, NEU Neutrophil count, PLT Platelet count, FIB Fibrinogen, PT Prothrombin time, NICU Neonatal intensive care unit

Among 146 women who converted to cesarean during labor (EA n = 76, non-EA n = 70), the distribution of indications differed between groups, while operative time was comparable. As shown in Table 4, the EA group had higher proportions of conversion for abnormal fetal position (34.21% vs. 22.86%), fetal distress (34.21% vs. 25.71%), and intra-amniotic infection (18.42% vs. 2.86%), whereas the non-EA group more often converted at maternal request (44.29% vs. 10.53%). Postoperatively, WBC, NEU, and D-dimer were higher in the EA group (P < 0.05). Total, pre-delivery, and post-delivery hospital stays, as well as neonatal birth weight, were also higher in the EA group (P < 0.05). Rates of PPH, neonatal sex, Apgar scores, and NICU admission did not differ significantly (P > 0.05).

Table 4.

Impact of epidural analgesia on perioperative outcomes among women undergoing intrapartum conversion to cesarean delivery

Characteristic	Total (n = 146)	Epidural anesthesia (n = 76)	No epidural anesthesia (n = 70)	P-value
Reasons for conversion from vaginal delivery to cesarean section				< 0.001*
Abnormal fetal position	42 (28.77)	26 (34.21)	16 (22.86)
Fetal distress	44 (30.14)	26 (34.21)	18 (25.71)
Amniotic cavity infection	16 (10.96)	14 (18.42)	2 (2.86)
Patient's personal wishes	39 (26.71)	8 (10.53)	31 (44.29)
Maternal comorbidities	5 (3.43)	2 (2.63)	3 (4.29)
Surgery time (mins)	43 (38–48.25)	43 (39–48)	43 (36–50)	0.366
Laboratory examination after delivery
HGB (g/L)	115.50 (107.75–125.00)	118.50 (106.25–125.65)	114.00 (108.00–123.25)	0.641
WBC (109/L)	13.42 (11.38–15.68)	13.68 (11.80–16.52)	12.70 (10.89–15.20)	0.039*
LYM (109/L)	1.37 (1.10–1.68)	1.35 (1.09–1.69)	1.40 (1.12–1.68)	0.472
NEU (109/L)	11.21 (9.24–13.47)	11.57 (9.68–14.02)	10.83 (8.85–12.63)	0.044*
PLT (109/L)	192.00 (169.00–226.00)	198.50 (169.00–229.00)	187.00 (170.50–220.75)	0.826
D-dimer (mg/L)	1.47 (0.97–2.04)	1.67 (1.07–2.26)	1.26 (0.91–1.81)	0.012*
FIB (g/L)	5.33 (4.70–5.81)	5.57 (4.43–6.08)	5.32 (5.00–5.65)	0.383
PT (s)	11.80 (11.38–12.30)	11.80 (11.40–12.40)	11.80 (11.30–12.23)	0.636
Postpartum hemorrhage	1 (0.68)	0 (0.00)	1 (1.43)	0.479
Total hospital stay days	6.00 (5.00–8.00)	7.00 (6.00–8.00)	5.50 (5.00–6.00)	< 0.001*
Pre-delivery hospital stay days	1.00 (0.00–3.00)	2.00 (1.00–3.00)	1.00 (0.00–2.00)	< 0.001*
Post-delivery hospital stay days	5.00 (4.00–5.00)	5.00 (4.00–5.75)	5.00 (4.00–5.00)	0.036*
Neonatal sex				0.740
Female	66 (45.21)	33 (43.42)	33 (47.14)
Male	80 (54.79)	43 (56.78)	37 (52.86)
Apgar score 1 min postpartum (points)				0.893
0–9	10 (6.85)	5 (6.58)	5 (7.14)
10	136 (93.15)	71 (93.42)	65 (92.86)
Apgar score 5 min postpartum (points)				0.607
0–9	3 (2.05)	1 (1.32)	2 (2.86)
10	143 (97.95)	75 (98.68)	68 (97.14)
Birth weight (g)	3450.00 (3200.00–3650.00)	3500.00 (3350.00–3700.00)	3350.00 (3087.50–3600.00)	0.002*
NICU	15 (10.27)	5 (6.58)	10 (14.29)	0.173

Data are presented as median (interquartile range) or n (%)

P < 0.05 was considered statistically significant and indicated by an asterisk

Abbreviations: HGB Hemoglobin, WBC White blood cell count, LYM Lymphocyte count, NEU Neutrophil count, PLT Platelet count, FIB Fibrinogen, PT Prothrombin time, NICU Neonatal intensive care unit

Risk factors for PPH

Finally, mnivariable and multivariable logistic regression were used to identify independent risk factors for PPH across the entire cohort (Table 5). Univariable analysis suggested that pre-delivery D-dimer, abdominal circumference, and neonatal birth weight were associated with PPH. In the multivariable model, higher D-dimer (OR = 1.54, 95% CI: 1.08–2.20, P = 0.017) and fibrinogen (OR = 1.34, 95% CI: 1.07–1.67, P = 0.010) were independent risk factors for PPH. Use of EA itself was not significantly associated with increased PPH risk (OR = 0.59, 95% CI: 0.22–1.58, P = 0.289).

Table 5.

Univariable and multivariate logistic regression analysis for postpartum hemorrhage

Characteristics	Univariable		Multivariate
	OR (95%CI)	P-value	OR (95%CI)	P-value
Age ≥ 35 vs. < 35 years	0.87 (0.36–2.10)	0.759	1.42 (0.38–5.30)	0.602
Parity ≥ 1 vs. 0	0.88 (0.48–1.61)	0.671	0.73 (0.32–1.66)	0.456
BMI ≥ 30 vs. < 30 kg/m2	0.88 (0.45–1.72)	0.699	1.09 (0.47–2.56)	0.840
Weight gain during pregnancy ≥ 15 vs. < 15 kg	0.79 (0.43–1.46)	0.455	0.69 (0.30–1.58)	0.384
Gestational age (day)	1.04 (1.00–1.08)	0.087	1.03 (0.97–1.10)	0.332
Epidural anesthesia vs. no	0.75 (0.37–1.54)	0.432	0.59 (0.22–1.58)	0.289
Hypertensive disorders of pregnancy vs. no	2.39 (0.32–17.69)	0.394	1.64 (0.19–14.37)	0.656
Gestational diabetes mellitus vs. no	1.32 (0.58–3.01)	0.504	1.84 (0.59–5.72)	0.290
Hypothyroidism vs. no	0.84 (0.37–1.91)	0.669	0.80 (0.28–2.25)	0.667
Premature rupture of membranes vs. no	1.27 (0.65–2.48)	0.494	1.09 (0.47–2.49)	0.847
Cord around the neck vs. no	1.48 (0.76–2.91)	0.251	1.18 (0.54–2.57)	0.685
Preterm birth vs. no	0.41 (0.06–3.03)	0.383	9.04 (0.39–207.64)	0.169
HGB (g/L)	1.00 (0.97–1.02)	0.842	1.00 (0.97–1.04)	0.974
WBC (109/L)	1.05 (0.93–1.12)	0.454	1.02 (0.45–2.30)	0.958
LYM (109/L)	0.67 (0.35–1.25)	0.208	0.94 (0.28–3.14)	0.913
NEU (109/L)	1.07 (0.95–1.22)	0.274	0.98 (0.43–2.27)	0.969
PLT (109/L)	1.00 (0.99–1.00)	0.159	1.00 (0.99–1.00)	0.237
ALB (g/L)	0.97 (0.86–1.09)	0.632	0.94 (0.80–1.10)	0.398
ALT (IU/L)	1.00 (0.97–1.04)	0.812	1.00 (0.95–1.06)	0.996
AST (IU/L)	1.00 (0.97–1.04)	0.859	1.03 (0.97–1.08)	0.402
D-dimer (mg/L)	1.45 (1.15–1.83)	0.002*	1.54 (1.08–2.20)	0.017*
FIB (g/L)	1.12 (0.91–1.39)	0.274	1.34 (1.07–1.67)	0.010*
PT (s)	1.14 (0.66–1.95)	0.636	1.60 (0.87–2.97)	0.133
Biparietal diameter (mm)	1.02 (0.94–1.11)	0.639	0.93 (0.78–1.12)	0.450
Head circumference (mm)	1.01 (0.98–1.03)	0.642	0.99 (0.93–1.05)	0.731
Abdomen circumference (mm)	1.02 (1.00–1.04)	0.037*	1.01 (0.98–1.05)	0.377
Femur length (mm)	1.07 (0.97–1.19)	0.181	1.11 (0.94–1.33)	0.227
Birth weight (g)	1.00 (1.00–1.00)	0.045*	1.00 (1.00–1.00)	0.313

P < 0.05 was considered statistically significant and indicated by an asterisk

The covariables in the multivariate regression models were maternal age, parity, BMI, weight gain during pregnancy, gestational age, epidural anesthesia, hypertensive disorders of pregnancy, gestational diabetes mellitus, hypothyroidism, premature rupture of membranes, cord around the neck, preterm birth, pre-delivery HGB, WBC, LYM, NEU, PLT, ALB, ALT, AST, D-dimer, FIB, PT, biparietal diameter, head circumference, abdomen circumference, femur length and birth weight

Abbreviations: OR Odds ratio, CI Confidence interval, BMI Body mass index, HGB Hemoglobin, WBC White blood cell count, LYM Lymphocyte count, NEU Neutrophil count, PLT Platelet count, ALB Albumin, ALT Alanine aminotransferase, AST Aspartate aminotransferase, FIB Fibrinogen, PT Prothrombin time

Discussion

The clinical value of labor analgesia extends beyond pain relief and improved birth experience to encompass peripartum safety and resource allocation. However, the determinants guiding clinical use of EA and its potential association with severe complications such as PPH remain central questions in obstetric anesthesia that warrant further investigation. Using a large clinical cohort, we systematically examined key factors associated with EA selection and focused on the effects of EA on labor course and maternal-neonatal outcomes. We found that maternal age, parity, and pre-delivery coagulation indices jointly influenced EA uptake, and a prediction model based on clinical variables showed good discriminative performance. Further analyses indicated that EA prolonged the duration of labor among women delivering vaginally but did not increase the risk of PPH or other adverse maternal-neonatal outcomes. Pre-delivery coagulation status, rather than EA itself, was identified as an independent risk factor for PPH. These findings provide evidence to elucidate decision-making around EA, to objectively evaluate its safety profile, and to refine strategies for PPH risk management.

Regarding determinants of EA use, younger maternal age and primiparity emerged as significant correlates in our cohort, suggesting that both demographic characteristics and obstetric history influence analgesia decisions. While previous studies have reported greater EA uptake among older or nulliparous women [6, 20], our findings may reflect a generational shift, with younger women showing stronger awareness and acceptance of labor analgesia. This trend may be attributed to higher health literacy and a stronger proactive demand for a painless childbirth experience among the younger generation. Furthermore, nulliparous women frequently experience higher levels of birth-related anxiety or tokophobia due to the unpredictability of their first labor, which significantly increases their inclination toward EA compared to multiparous women who may have established different coping mechanisms from previous deliveries.

We also observed statistical associations between EA use and pre-delivery coagulation status, specifically fibrinogen levels. Notably, although the reported fibrinogen values remained within the conventional physiological range for late pregnancy, the observed negative association with EA uptake suggests a subtle biobehavioral link. Fibrinogen is not only a coagulation substrate but also an acute-phase reactant [21, 22]; thus, its baseline levels may serve as a biological surrogate for a parturient’s systemic stress or inflammatory profile. Parturients with lower-normal fibrinogen levels may possess different neuroendocrine stress responses or systemic pain thresholds, which could indirectly influence their psychological threshold for seeking analgesia; it may also reflect shared decision-making by clinicians and patients after considering bleeding risk and anticipated intrapartum interventions [19, 23]. This observation provides a novel perspective that a woman’s decision to opt for EA is likely a multifactorial process influenced by an interplay of psychological readiness and baseline physiological status.

From a clinical perspective, these observations offer a new biobehavioral framework for understanding EA selection, where baseline physiological markers may serve as surrogates for an individual’s stress response or pain threshold. Additionally, the prediction model constructed from routinely available admission variables demonstrated good discrimination. The practical utility of this model extends to precision obstetric care, as the early identification of women with a high propensity for EA facilitates more efficient allocation of anesthesia resources and personnel. Furthermore, recognizing these independent determinants allows for targeted antenatal counseling, ensuring that individualized pain management strategies are aligned with the unique physiological and obstetric profile of each patient, thereby enhancing the overall delivery experience and safety.

While the primary utility of our findings lies in predicting EA demand, confirming the safety of these clinical choices is paramount for bedside application. Therefore, we further examined the impact of EA on obstetric outcomes to ensure that the identified high-uptake groups are not at increased risk for complications. Our results clearly show that EA is associated with longer total, first-stage, and second-stage labor among vaginal deliveries. This accords with the prevailing evidence in the field [24–26]. Potential mechanisms include sympathetic blockade with changes in uteroplacental hemodynamics and pelvic floor relaxation, which may reduce the presenting part’s pressure on the cervix and slow cervical dilation. Notably, pain relief and labor progression are closely linked; effective pain management through EA and complementary relaxation techniques can significantly reduce maternal anxiety, potentially encouraging more efficient labor. However, we did not evaluate the specific degree of pain relief in this cohort, which represents a limitation in our analysis of labor-related outcomes. Since total labor duration exceeding 12 h is a recognized risk factor for severe PPH [27, 28], the stability of the association between EA and PPH risk must be interpreted within the context of labor efficiency and the success of analgesia. Importantly, although labor was prolonged in our cohort, it did not translate into worse maternal or neonatal outcomes. These findings support the relative safety of EA and alleviate concerns that EA might increase serious adverse outcomes despite its effect on labor duration.

In the intrapartum conversion to cesarean subgroup, indications for conversion differed between the EA and non-EA groups. Women with EA were more likely to convert for clear obstetric indications (fetal distress, malpresentation, infection), possibly because adequate analgesia facilitates clinical monitoring and timely intervention, whereas conversion at maternal request was more frequent without EA. This highlights the role of EA in maintaining compliance during labor [29, 30]. Higher postoperative WBC, NEU counts, and D-dimer in the EA group may relate to longer labor or underlying infection risk, or reflect greater intensity of perioperative management rather than a direct effect of EA [5, 31]. These laboratory differences should be interpreted cautiously in light of the absence of differences in major outcomes.

Another key finding is that EA itself was not an independent risk factor for PPH, consistent with several recent high-quality observational studies [13, 32]. By contrast, higher pre-delivery D-dimer and fibrinogen levels independently predicted PPH. Elevated D-dimer indicates activation of fibrinolysis and potential intravascular coagulation [33]. Notably, the positive association between fibrinogen and PPH in our data is not entirely consistent with the conventional view that low fibrinogen predicts severe PPH [34, 35]. Possible explanations include measurement timing in late pregnancy rather than during early bleeding, influence of pregnancy-related hypercoagulability and inflammatory responses, correlations among coagulation and inflammatory markers leading to multicollinearity, and limited event counts affecting estimate stability [36]. Fibrinogen, as a coagulation substrate, may rise as a compensatory response in some contexts; in the setting of PPH, higher levels may also signal rapid consumption of coagulation capacity [37]. These observations underscore the clinical value of assessing coagulation status before delivery, rather than focusing solely on the choice of analgesia, for predicting and preventing PPH. Identifying such hematological shifts early provides clinicians with a vital window for risk stratification, allowing for more proactive management and targeted interventions during the third stage of labor.

Strengths of this study include a relatively large sample size, rigorous multivariable adjustment, and stratified analyses by delivery pathway, which together enhance the credibility of our findings. However, several limitations warrant consideration. The retrospective design carries an inherent risk of selection bias and residual confounding, as the choice of analgesia was influenced by maternal preference rather than randomized allocation. Additionally, our single-center data may limit the generalizability of the results to other clinical settings. Furthermore, the relatively small number of PPH events in this cohort may affect the stability of multivariable estimates; thus, our observations regarding EA and PPH risk should be interpreted as observational evidence rather than definitive causal proof. Moreover, our predictive model was designed for use at admission, meaning that dynamic intrapartum factors such as labor pain intensity (VAS scores), the specific degree of pain relief and the use of oxytocin were not incorporated. The absence of these variables may represent a source of confounding in the analysis of labor outcomes. Future work in multicenter prospective cohorts is warranted to externally validate and refine the prediction model and to integrate detailed process measures and peripartum interventions, which would further improve external validity and causal inference.

Conclusion

In conclusion, maternal age, parity, and pre-delivery coagulation indices are independent determinants of EA selection, and a clinically based multivariable model shows good discrimination. Although EA prolongs labor among women delivering vaginally, it does not increase the risk of PPH or other adverse maternal and neonatal outcomes, supporting its safety profile. Pre-delivery coagulation abnormalities, rather than the choice of analgesia, are the principal risk factors for PPH. These findings provide actionable evidence for individualized decisions on labor analgesia and for risk-based management of PPH.

Abbreviations

ALB

Albumin

ALT

Alanine aminotransferase

ART

Assisted reproductive technology

AST

Aspartate aminotransferase

AUC

The area under the ROC curve

BMI

Body mass index

CI

Confidence interval

DT

Decision tree

EA

Epidural analgesia

FIB

Fibrinogen

GB

Gradient boosting

HGB

Hemoglobin

LDA

Linear discriminant analysis

LR

Logistic regression

LYM

Lymphocyte count

NEU

Neutrophil count

NICU

Neonatal intensive care unit

OR

Odds ratio

PLT

Platelet count

PPH

Postpartum hemorrhage

PT

Prothrombin time

QDA

Quadratic discriminant analysis

RF

Random forest

ROC

The receiver operating characteristic

SVM

Support vector machine

WBC

White blood cell count

XGBoost

EXtreme Gradient Boosting

Authors’ contributions

Conceptualization, L.G., FC.G., and YQ.G.; methodology, L.G., Y.S., and YM.Z.; software, L.G.; validation, Y.S., YM.Z., Q.S. and W.L.; formal analysis, L.G. and YQ.G.; investigation, YN.L. and FC.G.; resources, Y.S., YM.Z., Q.S., W.L., and YN.L.; data curation, L.G. and YQ.G.; writing-original draft preparation, L.G. and YQ.G.; writing-review and editing, L.G. and YQ.G.; visualization, L.G.; supervision, FC.G. and YQ.G.; project administration, FC.G. and YQ.G.; funding acquisition, YQ.G.. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the China Postdoctoral Science Foundation (2025M782118), the Shandong Provincial Postdoctoral Science Foundation (SDZZ-ZR-202501039), the Jinan Science and Technology Plan (202328019), the Women and Children Health Care Association of Shandong (SFYXH-2023Y027), and the Jinan Municipal Health Commission (2025308027).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of Jinan Maternal and Child Health Hospital, Jinan, China on 1 September 2023 (approval No. 2023–1-038). Informed consent was obtained from all subjects involved in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.