A Clinically Practical Postpartum Depression Predictor: Machine Learning Model Based on Simplified Indicators

A Clinically Practical Postpartum Depression Predictor: Machine Learning Model Based on Simplified Indicators | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Clinically Practical Postpartum Depression Predictor: Machine Learning Model Based on Simplified Indicators Hongmei Lin, Chunfei Hu, Siqian Hu, Ying Ying, Lingling Shang, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7079798/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Purpose: Postpartum Depression (PPD) is the most prevalent psychiatric disorder following childbirth, posing significant risks to both maternal and infant health. Consequently, prevention is paramount. This study aims to develop a clinically practical prediction tool using simple, readily accessible indicators to facilitate early intervention. Methods: In this multi-center study, data from 5,011 postpartum women were collected through structured questionnaires and electronic health records (EHRs). PPD was defined as an Edinburgh Postnatal Depression Scale (EPDS) score ≥10 at 6 weeks postpartum. Predictors within the dataset encompassed sociodemographic characteristics, pregnancy factors, delivery experiences, and infant care practices. Following feature selection, 11 machine learning(ML) models were constructed and evaluated. Model interpretability was enhanced using SHapley Additive exPlanations (SHAP). Results: The prevalence of PPD was 13.3%. Nine core predictors were ultimately identified: support from spouse, parents, parents-in-law, siblings, and friends; fetal distress; birth experience rating; marital status; and marriage duration. Among the algorithms evaluated, the Random Forest model demonstrated relatively superior performance (Training set AUC: 0.725, 95% CI: 0.697-0.752; Validation set AUC: 0.612, 95% CI: 0.568-0.656). SHAP analysis identified optimizing the birth experience and strengthening the social support system as key, clinically actionable intervention targets. Conclusion: This study confirms that a ML model based on simplified indicators provides moderate-performance risk stratification for PPD. This clinically practical tool equips frontline clinicians in resource-constrained settings in safeguarding vulnerable mothers through proactive early-warning systems during their critical transition period, thus reducing severe PPD and preventing devastating consequences. Postpartum depression EPDS Machine learning Prediction model Figures Figure 1 Figure 2 Figure 3 Figure 4 Article Highlights 1. Large-scale multicenter cohort (n=5011) reveals 13.3% PPD prevalence and identifies key risk factors. 2. 11 ML models predict PPD risk using 9 easily accessible sociodemographic and delivery indicators. 3. Models confirm feasibility of using simplified indicators for PPD screening in resource-limited settings. 4. SHAP analysis identifies optimizing birth experience and strengthening social support as actionable PPD prevention targets. ‌Introduction Postpartum depression (PPD) is the most prevalent mental health disorder during the perinatal period, with a global pooled prevalence of 17.2%(Wang et al. 2021 ). Beyond impairing maternal parenting capacity, PPD adversely affects children's emotional, cognitive, and behavioral development(Lim et al. 2018 ; Rogers et al. 2020 ) and substantially elevates maternal self-harm risk(Johannsen et al. 2016 ; Hagatulah et al. 2024 ). Epidemiological studies indicate that 20% of postpartum maternal deaths result from suicide(Lindahl et al. 2005 ), underscoring the critical need to develop clinically actionable prediction tools as part of multifaceted preventive strategies(Khalifeh et al. 2016 ). Current clinical practice primarily follows the Grade B recommendation from the U.S. Preventive Services Task Force (USPSTF), utilizing self-report tools such as the Edinburgh Postnatal Depression Scale (EPDS) for routine screening(US Preventive Services Task Force et al. 2019). However, such scale-based screening carries inherent limitations in timeliness, typically detecting only individuals who already exhibit symptoms at the time of assessment(Liu et al. 2022 ). Consequently, there is a critical need to develop tools capable of predicting depression predisposition before symptom onset, enabling timely support and intervention. Machine learning(ML) technologies have recently provided novel methodological frameworks for PPD risk prediction. A meta-analysis of 17 studies encompassing 62 prediction models(Zhong et al. 2022 ) revealed that most models integrating multidimensional predictors such as sociodemographic characteristics, obstetric history, psychiatric diagnoses, and biomarkers from electronic health records (EHRs) demonstrate robust predictive performance (AUC 0.660–0.969). Nevertheless, such models face challenges in clinical application: complex architectures with over 20 predictors hinder clinical utility(Hochman et al. 2021 ; Zhang et al. 2021 ), while reliance on non-standardized data (e.g., serological or genetic markers) limits their applicability in primary care settings(Payne et al. 2020 ; Xiao et al. 2020 ). Evidence confirms significant associations between PPD risk and factors including low social support, economic stress, marital discord, and traumatic birth experiences(Söderquist et al. 2009 ; Cho et al. 2022 ; Waller et al. 2022 ). Accordingly, this study utilized ML to develop a concise PPD prediction model, prioritizing the predictive power of sociodemographic features. Designed for clinical utility, the model requires only a few readily simplified variables for risk stratification. Using a retrospective cohort design and multiple ML algorithms, we aim to optimize the balance between predictive accuracy and clinical practicality, ultimately providing primary care with an accessible and implementable PPD risk screening tool. Materials and Methods Study Population This study extends our prior research developing and validating a prenatal depression prediction model. From an established cohort, we enrolled 20,950 women in late pregnancy (≥34 weeks) who received prenatal care and completed questionnaires at the Women's Hospital of Zhejiang University School of Medicine, Shaoxing Maternity and Child Health Care Hospital, and Shengzhou Maternity and Child Health Care Hospital. Ultimately, 5,011 participants delivered at these hospitals between January 1, 2020, and April 15, 2024, and completed valid follow-up at 6 weeks postpartum, forming the analytical cohort. All participants provided written informed consent. The research protocol was developed in compliance with the Declaration of Helsinki and formally approved by the Medical Ethics Committee of Shaoxing Maternity and Child Health Care Hospital (Approval No. 2025-024-01). Data Collection Data were obtained from self-administered prenatal questionnaires, postpartum questionnaires, and obstetric electronic health records (EHRs). Questionnaire data originated from the Perinatal and Postpartum Depression Information System. The third-trimester survey covered sociodemographics, social support, lifestyle habits, family history, obstetric history, medical history, pregnancy complications, and Edinburgh Postnatal Depression Scale (EPDS) assessments. To mitigate misreporting bias in pregnancy-related conditions due to participants' limited medical knowledge, EHRs were used to extract discharge diagnoses for medical history, pregnancy complications, and delivery outcomes (delivery mode, gestational age at birth, neonatal birth weight). At 6-week postpartum follow-up, participants completed another questionnaire assessing postpartum EPDS, delivery experiences, and infant care practices. Outcome Definition The primary outcome was PPD, defined as an EPDS score ≥10 at 6 weeks postpartum. This threshold was selected to improve screening sensitivity and reduce missed diagnoses compared to the conventional cutoff of 13, as evidence suggests it is more appropriate for Asian populations. Predictor Variables Predictors derived from questionnaires and EHRs included: 1) Sociodemographics characteristics , assessed via prenatal questionnaires, encompassed depression risk factors established in our prior research: social support, sleep patterns, physical activity, pregnancy planning, obstetric history, and family history(e.g., hyperlipidemia, schizophrenia, anxiety/depression). 2) Pregnancy-related factors , extracted from EHRs, comprised medical history (e.g., anxiety/depression, chronic hypertension), pregnancy complications (e.g., preeclampsia, gestational diabetes, threatened abortion, placenta previa, fetal distress), and delivery outcomes (delivery mode, gestational age at delivery, neonatal birth weight). 3) Postpartum factors , assessed through postpartum questionnaires, included: Delivery experiences: trial of labor preceding cesarean, concordance between expected and actual delivery mode, intrapartum oxytocin administration, epidural analgesia during labor, family presence at delivery, and subjective birth experience rating. Infant care practices: neonatal congenital anomalies, postnatal hospitalization, current feeding method, and nighttime feeding frequency. Model Development and Evaluation The entire dataset was randomly split in a 7:3 ratio into a training set (n=3508) and a validation set (n=1503). The training set was used for model development, while the validation set was used for model evaluation. Utilizing variables selected via univariate logistic regression and the Boruta algorithm based on the training set, this study constructed 11 clinically prevalent prediction models, including Logistic Regression, Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost, AdaBoost, Multilayer Perceptron (MLP), Support Vector Machine (SVM), Naive Bayes (NB), and Decision Tree (DT). Model hyperparameters were optimized via grid search with manual refinement. Following optimization, all models underwent evaluation through 5-fold cross-validation using the dataset. Evaluation metrics comprised the area under the ROC curve (AUC), F1-score, sensitivity, and specificity. Comparative analysis across these metrics identified the optimal model as the final predictive model. The selected model was subjected to SHapley Additive exPlanations (SHAP) analysis for visual interpretation of predictions. Additionally, calibration curves were generated to assess the agreement between predicted probabilities and observed outcomes, facilitating detection of overfitting or underfitting. Statistical Analysis Continuous variables are presented as mean ± SD (normally distributed) or median (IQR) (non-normally distributed); categorical variables are presented as number (%). Between-group comparisons were performed as follows: for continuous variables, t -test (normally distributed with equal variance), t '-test (normally distributed with unequal variance), or Wilcoxon rank-sum test (non-normally distributed); for categorical variables, χ ² test or corrected χ ² test, or Fisher’s exact test if total sample size < 40 or minimum expected cell count < 1; for ordinal data, Wilcoxon rank-sum test. Univariate logistic regression results are reported as odds ratios ( OR ) with 95% confidence interval ( CI ). Statistical significance was defined as a two-tailed P -value < 0.05. All statistical analyses and plotting were performed using Python 3.9 software. Results Baseline Characteristics This study included 5,011 participants aged 16.00–47.00 years. The mean age was 28.77 ± 3.68 years, with a median of 28.00 years (IQR: 26.00–31.00). Participants had a mean EPDS score of 4.55 ± 4.57 and a median score of 3.00 (IQR: 1.00–7.00). PPD developed in 667 participants, representing a prevalence of 13.3%. As this study aimed to predict PPD risk using simplified data, antenatal EPDS scores were excluded as predictive variables. Eighty-one potential predictors were ultimately included. Differential analysis conducted on the training set identified 24 factors significantly associated with PPD ( P < 0.05), including marital status, sleep duration, previous delivery experience, current delivery experience, and social support (Supplementary Table 1). Univariate Analysis Table 1 presents the univariate logistic regression results for statistically significant variables. Higher annual household income, longer marital duration, positive previous delivery experience, and strong support from family and friends emerged as protective factors against PPD ( OR < 1, P 1, P < 0.05). Single and divorced women faced significantly higher PPD risk compared to married women ( OR = 2.463, 95% CI: 1.128–4.962; OR = 13.298, 95% CI: 2.588–96.128). Difficulty falling asleep increased depression risk ( OR = 1.008, 95% CI: 1.004–1.012), whereas longer sleep duration ( OR = 0.847, 95% CI: 0.788–0.910) and walking time ( OR = 0.994, 95% CI: 0.972–1.012) reduced risk. Among obstetric factors, placental malposition ( OR = 2.877, 95% CI: 1.101–6.778) and intrauterine fetal distress ( OR = 1.414, 95% CI: 1.016–1.926) significantly increased PPD risk. Unexpectedly, gestational diabetes ( OR = 0.635, 95% CI: 0.413–0.938) and hypothyroidism ( OR = 0.457, 95% CI: 0.192–0.920) were associated with reduced risk. Regarding postpartum factors, delivery methods inconsistent with expectation increased PPD risk by 1.6-fold ( OR = 1.614, 95% CI: 1.269–2.039), whereas smooth and satisfactory birth experiences substantially reduced risk ( OR = 0.183–0.335, P < 0.001). Nighttime feeding 1–2 times suggested a potential protective trend when compared to no feeding ( OR = 0.567, 95% CI: 0.260–1.420) or frequent feeding. Table 1 Univariate Logistic Regression Results No. Predictor OR (95%CI) Statistics P value 1 Prepregnancy BMI(kg/m²) 0.951 ( 0.921–0.981 ) 9.709 0.002 2 Annual household income(CNY) 500k 0.588 ( 0.293–1.172 ) 2.268 0.132 3 Marital status Married Reference Single 2.463 ( 1.128–4.962 ) 5.818 0.016 Divorced 13.298 ( 2.588–96.128 ) 8.898 0.003 4 Marriage duration Unmarried Reference < 1yr 0.374 ( 0.196–0.744 ) 8.515 0.004 1-10yrs 0.295 ( 0.157–0.580 ) 13.587 < 0.0001 ≥ 10yrs 0.222 ( 0.102–0.490 ) 14.288 < 0.0001 5 Smoking history No Reference Yes 2.166 ( 1.185–3.757 ) 6.991 0.008 6 Alcohol use history No Reference Yes 1.304 ( 1.067–1.591 ) 6.802 0.009 7 Pregnancy plan Planned Reference Unplanned 1.238 ( 1.007–1.518 ) 4.153 0.042 8 Prior induced abortion None Reference Once 1.075 ( 0.837–1.371 ) 0.334 0.563 Twice 1.330 ( 0.857–1.996 ) 1.764 0.184 Three times or more 2.115 ( 1.178–3.609 ) 6.964 0.008 9 Previous birth experience rating Primipara Reference Extremely Negative 3.013 ( 0.934–8.534 ) 4.008 0.045 Negative 0.988 ( 0.472–1.861 ) 0.001 0.972 Neutral 0.880 ( 0.666–1.149 ) 0.843 0.359 Positive 0.606 ( 0.412–0.866 ) 6.996 0.008 Extremely Positive 0.402 ( 0.156–0.850 ) 4.588 0.032 10 Support from spouse None Reference Minimal 2.635 ( 1.104–6.844 ) 4.427 0.035 Moderate 1.472 ( 0.712–3.448 ) 0.947 0.330 Strong 0.689 ( 0.340–1.593 ) 0.916 0.339 11 Support from parents None Reference Minimal 1.190 ( 0.590–2.390 ) 0.240 0.624 Moderate 0.713 ( 0.425–1.225 ) 1.585 0.208 Strong 0.388 ( 0.244–0.636 ) 15.171 < 0.0001 12 Support from parents-in-law None Reference Minimal 1.186 ( 0.696–2.022 ) 0.397 0.529 Moderate 0.813 ( 0.535–1.256 ) 0.908 0.340 Strong 0.404 ( 0.276–0.605 ) 20.548 < 0.0001 13 Support from siblings None Reference Minimal 1.568 ( 0.902–2.644 ) 2.709 0.100 Moderate 1.037 ( 0.750–1.433 ) 0.048 0.827 Strong 0.545 ( 0.422–0.709 ) 21.068 < 0.0001 14 Support from friends None Reference Minimal 1.514 ( 0.908–2.508 ) 2.579 0.108 Moderate 0.663 ( 0.459–0.967 ) 4.700 0.030 Strong 0.451 ( 0.325–0.636 ) 21.716 < 0.0001 15 Sleep latency(minutes) 1.008 ( 1.004–1.012 ) 18.046 < 0.0001 16 Sleep duration(hours) 0.847 ( 0.788–0.910 ) 20.666 < 0.0001 17 Weekly walking hours during pregnancy(hours) 0.994 ( 0.972–1.012 ) 0.387 0.534 18 GDM No Reference Yes 0.635 ( 0.413–0.938 ) 4.757 0.029 19 Placenta previa low-lying No Reference Yes 2.877 ( 1.101–6.778 ) 5.373 0.020 20 Hypothyroidism No Reference Yes 0.457 ( 0.192–0.920 ) 3.940 0.047 21 Fetal distress No Reference Yes 1.412 ( 1.016–1.926 ) 4.486 0.034 22 Delivery mode consistency with expectation No Reference Yes 1.614 ( 1.269–2.039 ) 15.658 < 0.0001 23 Birth experience rating Traumatic Reference Difficult but Worthwhile 0.335 ( 0.253–0.447 ) 56.686 < 0.0001 Smooth 0.329 ( 0.240–0.451 ) 47.679 < 0.0001 Very Satisfactory 0.183 ( 0.127–0.264 ) 82.519 < 0.0001 24 Nighttime feeding frequency None Reference Once or Twice 0.567 ( 0.260–1.420 ) 1.769 0.184 Three or four times 0.740 ( 0.342–1.846 ) 0.504 0.478 Five times or more 0.932 ( 0.389–2.491 ) 0.023 0.881 Variable Selection Using the Boruta algorithm, we screened statistically significant variables identified through univariate logistic regression analysis. As shown in Fig. 1 , nine key predictors(blue ridge) were confirmed as essential features for the PPD prediction model: support from spouse, support from parents, support from parents-in-law, support from siblings, support from friends, marital status, marriage duration, fetal distress and birth experience rating. Model Development and Evaluation With PPD occurrence as the dependent variable and the aforementioned nine features as independent variables, we constructed eleven ML models: LogisticRegression, RandomForest, GradientBoosting, XGBoost, LightGBM, CatBoost, AdaBoost, MLP, SVM, NB, DT. Model hyperparameters were optimized through 5-fold stratified cross-validation combined with randomized search, using ROC-AUC as the primary evaluation metric. Optimized hyperparameters for each model are detailed in Supplementary Table 2. As presented in Table 2 and Fig. 2 , the Random Forest model demonstrated relatively superior predictive performance. In the training set, it achieved an AUC of 0.725 (95% CI: 0.697–0.752), with F1-score, sensitivity, and specificity values of 0.699, 0.660, and 0.630 respectively. Corresponding values for the validation set were AUC = 0.612 (95% CI: 0.568–0.656), F1-score = 0.667, sensitivity = 0.520, and specificity = 0.617. Figure 3 displays calibration curves for all predictive models in both training and validation sets. For Random Forest, all points approach the ideal line, indicating satisfactory calibration. Table 2 Performance Comparison of All Models No. Model Dataset AUC AUC 95%CI F1-score Sensitivity Specificity 1 Logistic Regression Training 0.660 0.632–0.688 0.748 0.497 0.740 Validation 0.616 0.572–0.660 0.737 0.415 0.741 2 Random Forest Training 0.725 0.697–0.752 0.699 0.660 0.639 Validation 0.612 0.568–0.656 0.667 0.520 0.617 3 Gradient Boosting Training 0.711 0.684–0.739 0.720 0.632 0.674 Validation 0.616 0.572–0.659 0.697 0.485 0.665 4 XGBoost Training 0.701 0.673–0.728 0.724 0.615 0.683 Validation 0.627 0.583–0.671 0.701 0.485 0.672 5 LightGBM Training 0.717 0.690–0.745 0.784 0.538 0.789 Validation 0.619 0.575–0.663 0.760 0.405 0.778 6 CatBoost Training 0.685 0.657–0.713 0.702 0.612 0.652 Validation 0.610 0.566–0.654 0.685 0.510 0.645 7 AdaBoost Training 0.660 0.631–0.688 0.718 0.580 0.680 Validation 0.624 0.580–0.668 0.708 0.500 0.680 8 MLP Training 0.710 0.683–0.738 0.494 0.852 0.365 Validation 0.603 0.559–0.647 0.494 0.740 0.376 9 SVM Training 0.601 0.572–0.630 0.270 0.921 0.160 Validation 0.595 0.551–0.639 0.254 0.890 0.150 10 NB Training 0.653 0.624–0.681 0.672 0.604 0.610 Validation 0.582 0.538–0.626 0.654 0.505 0.600 11 DT Training 0.677 0.649–0.705 0.764 0.525 0.760 Validation 0.606 0.562–0.650 0.752 0.380 0.771 SHAP Analysis We employed SHAP analysis to interpret the Random Forest model's PPD prediction logic. Figure 4 presents complementary visualizations: the left beeswarm plot illustrates feature impact direction and magnitude through Shapley values, while the right donut chart quantifies global feature importance. Both consistently identify BirthExperienceRating, SupportFromSpouse, and SupportFromParentsInlaw as core predictors. In the beeswarm plot, the horizontal axis represents Shapley values and the vertical axis lists features, with red indicating high feature values and blue indicating low values. For BirthExperienceRating, most data points cluster near zero (blue), suggesting neutral impact at moderate ratings ("Difficult but Worthwhile"). Scattered high-value points (red, > 0.1) indicate reduced PPD risk at higher satisfaction levels. SupportFromSpouse exhibits a right-skewed distribution (blue-dominant), confirming its protective effect though weaker than BirthExperienceRating. Notably, single and divorced women show elevated risk compared to married individuals. The donut chart quantifies global contributions: BirthExperienceRating accounts for 27.3% of predictive power, followed by SupportFromParentsInlaw (19.3%) and SupportFromSpouse (18.5%). These three features collectively constitute 65% of the model's predictive capability, forming the core feature set. Clinically, these findings directly support our PPD prediction objectives by demonstrating: 1) the critical importance of optimizing birth experiences given BirthExperienceRating's dominance, and 2) the empirical foundation for family-centered interventions leveraging spousal and parental support systems. Discussion This study developed eleven ML models for PPD risk prediction using nine readily accessible sociodemographic and delivery-related indicators: support from spouses, parents, parents-in-law, siblings, and friends; fetal distress; birth experience rating; marital status; and marriage duration. Among these, the Random Forest model demonstrated relatively superior performance in the validation cohort (AUC = 0.61). While this predictive accuracy falls within the moderate range, it confirms the fundamental feasibility of using simplified clinical data for PPD risk stratification. This approach offers a practical screening solution particularly valuable for resource-limited healthcare settings. PPD poses a serious threat to maternal health with multifaceted consequences, including mood disorders, increased risk of child neglect(Slomian et al. 2019 ; US Preventive Services Task Force et al. 2019), and in severe cases, emergence of suicidal or infanticidal ideation(Johannsen et al. 2016 ; US Preventive Services Task Force et al. 2019; Johannsen et al. 2020 ). Crucially, maternal mortality review committees recognize that deaths attributable to mental health complications demonstrate substantially higher preventability compared to other pregnancy-related mortality causes(Trost et al. 2021 ). Early identification of high-risk individuals therefore represents a critical first step in establishing timely intervention pathways, creating a vital prevention window to mitigate adverse outcomes. From an implementation perspective, our approach addresses significant limitations in conventional PPD screening. Traditional methods often require specialized personnel training, hindering large-scale deployment in resource-constrained environments. In contrast, our ML framework enables rapid, cost-efficient risk assessment within primary care settings. The required predictive variables (including marital status, social support levels, and birth experience) can be efficiently captured immediately postpartum through concise questionnaires, eliminating dependence on biomarker testing or complex psychological assessments. This streamlined data collection significantly reduces clinician workload while enabling efficient risk stratification with minimal resources, ultimately improving screening coverage and facilitating appropriate referrals to prevent severe psychiatric sequelae. Although existing literature documents numerous ML models with good generalization capacity for PPD prediction(Cellini et al. 2022 ), their clinical translation faces substantial barriers(Fusar-Poli et al. 2018 ). The primary limitation lies in the tension between model complexity and clinical utility. Most existing predictive frameworks require multidimensional inputs spanning sociodemographic characteristics, obstetric history, biological markers, genetic indicators, and psychological assessments(Cellini et al. 2022 ; Zhong et al. 2022 ). For example, Clapp et al.'s robust 26-variable model (AUC = 0.750)(Clapp et al. 2025 ) proves impractical in routine clinical workflows due to its extensive data requirements. Such models inherently presuppose well-established prenatal screening infrastructures—resources frequently unavailable where prenatal coverage remains suboptimal(Gelaye et al. 2016 ). Similarly, Qi et al.(Qi et al. 2025 ) integrated multiple psychosocial scales and physiological indicators, ultimately identifying 11 predictors including prenatal depression, neuroticism, total cholesterol, and serum free triiodothyronine, with the model achieving AUC values of 0.787–0.858. However, these specialized physiological and psychological parameters fall beyond the scope of routinely collected data in primary care settings, thereby limiting the model's generalizability. A second significant limitation concerns the practical applicability of models over-reliant on mental health history(Zhang et al. 2025 ). Many studies prioritize predictive accuracy by emphasizing psychiatric antecedents, utilizing prior diagnoses and medication histories as core predictors. Wakefield et al.'s high-performance model (AUC = 0.91)(Wakefield and Frasch 2023 ), for instance, requires nine variables including pre-pregnancy depression history and early-pregnancy psychotropic medication use. However, screening individuals with established psychiatric vulnerabilities offers limited clinical value since these conditions already constitute recognized PPD risk factors. More importantly, stigma-induced information suppression presents a fundamental barrier: affected women may deliberately conceal psychiatric histories or resist psychological evaluations due to perceived shame, inevitably leading to critical data gaps and diagnostic omissions(Thorsteinsson et al. 2018 ). Beyond predictive functionality, our SHAP interpretability analysis reveals two principal modifiable factors. First, optimizing birth experience proves paramount: our quantification shows that women reporting "very satisfactory" deliveries experience an 82% reduction in PPD susceptibility, unequivocally demonstrating the protective value of patient-centered intrapartum care. Patient-reported birth trauma is a critical predictor of adverse psychological outcomes, with established links between traumatic delivery exposure and subsequent depression(Waller et al. 2022 ), including comorbid postpartum post-traumatic stress disorder(Dekel et al. 2020 ). Mechanistically, negative vaginal birth experiences frequently stem from induction procedures or instrumental deliveries(Froeliger et al. 2024 ), while cesarean-related trauma often relates to emergency surgery or inadequate postoperative analgesia(Froeliger et al. 2025 ). Collectively, this evidence confirms that negative birth perceptions—regardless of delivery mode—significantly elevate PPD risk. This necessitates evidence-based refinement of induction protocols, enhanced obstetric technical proficiency, optimized perioperative pain management, and strengthened multidisciplinary collaboration among obstetricians, midwives, and anesthesiologists(Du et al. 2025 ). Second, strengthening social support systems is essential: our analysis confirms that deficient spousal or in-law support substantially increases PPD vulnerability, highlighting the need for structured family counseling. Zhang and Jin's research(Zhang and Jin 2016 ) indicates that social support indirectly mitigates PPD through enhanced maternal self-efficacy. Indonesian cohort studies(Pebryatie et al. 2022 ) demonstrate that proactive spousal involvement during pregnancy promotes healthy behaviors and significantly reduces postpartum depressive symptoms. Furthermore, Japanese research(Terada et al. 2022 ) reveals that women experiencing deteriorating parental relationships or persistently unsatisfactory family dynamics face 2.81-fold and 2.39-fold increases in PPD susceptibility, respectively. We therefore advocate explicitly incorporating spousal support expectations in antenatal education materials and implementing community-based interventions that actively engage family participation. Additionally, our study identified other clinically relevant predictors. Fetal distress demonstrated significant predictive value, likely mediated through neonatal resuscitation or NICU admission pathways—known to elevate maternal depression prevalence to approximately 40%(Shovers et al. 2021 ). Placental positional abnormalities were also identified as risk factors, likely attributable to anxiety regarding adverse pregnancy outcomes precipitated by recurrent vaginal bleeding. Notably, gestational diabetes and hypothyroidism showed paradoxical protective associations, contradicting some prior research(Azami et al. 2019 ; Costantine et al. 2020 ). This may reflect generally milder phenotypes in our cohort, effective clinical management, or enhanced support systems via intensified prenatal monitoring. Other sociodemographic factors—including socioeconomic deprivation, marital discord, and tobacco or alcohol use—have been established as consistent risk amplifiers for PPD, aligning with prior research(Norhayati et al. 2015 ; Wang et al. 2021 ). Conversely, ensuring adequate sleep duration(Okun 2015 ), engaging in regular moderate-intensity physical activity(Nakamura et al. 2019 ), and implementing flexible breastfeeding practices(Zhu et al. 2025 ) demonstrate protective benefits against depression risk. This study developed a clinically practical PPD screening tool using only 9 simplified indicators, enabling rapid risk stratification in resource-limited settings. In clinical practice, frontline clinicians can administer 9 predetermined questions to postpartum women, enabling automated risk assessment through this instrument. This approach facilitates early risk identification, allowing timely interventions for high-risk individuals to effectively prevent adverse outcomes during the puerperium and beyond. Nevertheless, several limitations warrant acknowledgment. The model's moderate predictive performance (AUC = 0.61) falls short of optimal clinical standards. However, given its minimal variable requirements, this achievement retains practical value as a preliminary screening instrument in primary care contexts. Although rigorous internal validation was conducted through cross-validation techniques, the lack of external validation across independent cohorts remains a critical implementation constraint. Future studies should consequently evaluate its generalizability and robustness in diverse populations and geographic regions. Finally, the prediction timeframe is confined to the postpartum period; since some PPD pathogenesis may originate prenatally, subsequent research should integrate first- and second-trimester data to enable earlier risk prediction during pregnancy, thereby further advancing the prevention window. Conclusion This study successfully developed and validated a ML-based PPD screening framework using 9 simplified sociodemographic and delivery-related indicators. Its strengths include resource-efficient implementation and the identification of two actionable intervention targets: systematic enhancement of birth experiences and strategic reinforcement of social support networks. Despite moderate predictive performance, this tool demonstrates significant potential for scalable PPD screening in resource-limited regions, enabling early risk identification to reduce severe PPD incidence and prevent devastating outcomes. Declarations Ethics approval All participants provided written informed consent. The research protocol was developed in compliance with the Declaration of Helsinki and formally approved by the Medical Ethics Committee of Shaoxing Maternity and Child Health Care Hospital (Approval No. 2025-024-01). Funding This work was supported by the Shaoxing Health Science and Technology Plan (Grant number 2024SKY048). The funding agency had no role in the design of the study, the collection, analysis, and interpretation of the data, or in writing the manuscript. Conflict of interest The authors declare that they have no conflict of interest. Ethics approval All participants provided written informed consent. The research protocol was developed in compliance with the Declaration of Helsinki and formally approved by the Medical Ethics Committee of Shaoxing Maternity and Child Health Care Hospital (Approval No. 2025-024-01). Funding This work was supported by the Shaoxing Health Science and Technology Plan (Grant number 2024SKY048). The funding agency had no role in the design of the study, the collection, analysis, and interpretation of the data, or in writing the manuscript. Author Contribution 1.Hongmei Lin(first author): Conceptualization, Resources, Writing - original draft preparation2.Chunfei Hu(first author):Methodology, Formal analysis, Writing - original draft preparation3.Siqian Hu:Data curation4.Ying Ying:Data collection5.Lingling Shang:Data collection6.Xiaojing Qiu:Resources7.Qiong Luo(Corresponding author): Resources,Supervision, Writing - review and editing8.Hualin Xu(Corresponding author):Conceptualization, Funding acquisition, SupervisionAll authors reviewed the manuscript. Acknowledgments The authors are grateful to the women who participatedin and donated clinical data for this study. References Azami M, Badfar G, Soleymani A, Rahmati S (2019) The association between gestational diabetes and postpartum depression: A systematic review and meta-analysis. Diabetes Res Clin Pract 149:147-155. https://doi.org/10.1016/j.diabres.2019.01.034 Cellini P, Pigoni A, Delvecchio G, Moltrasio C, Brambilla P (2022) Machine learning in the prediction of postpartum depression: A review. J Affect Disord 309:350-357. https://doi.org/10.1016/j.jad.2022.04.093 Cho H et al. (2022) Association between social support and postpartum depression. Sci Rep 12(1):3128. https://doi.org/10.1038/s41598-022-07248-7 Clapp MA, Castro VM, Verhaak P, McCoy TH, Shook LL, Edlow AG, Perlis RH (2025) Stratifying Risk for Postpartum Depression at Time of Hospital Discharge. Am J Psychiatry :appiajp20240381. https://doi.org/10.1176/appi.ajp.20240381 Costantine MM et al. (2020) Effect of Thyroxine Therapy on Depressive Symptoms Among Women With Subclinical Hypothyroidism. Obstet Gynecol 135(4):812-820. https://doi.org/10.1097/AOG.0000000000003724 Dekel S, Ein-Dor T, Dishy GA, Mayopoulos PA (2020) Beyond postpartum depression: posttraumatic stress-depressive response following childbirth. Arch Womens Ment Health 23(4):557-564. https://doi.org/10.1007/s00737-019-01006-x Du W, Qian X, Xu Z, Liu Z (2025) The role of anesthesiologists in postpartum depression: current perspectives and future directions. Front Psychiatry 16:1511817. https://doi.org/10.3389/fpsyt.2025.1511817 Froeliger A et al. (2024) Prevalence and risk factors for postpartum depression 2 months after a vaginal delivery: a prospective multicenter study. Am J Obstet Gynecol 230(3S):S1128-S1137.e6. https://doi.org/10.1016/j.ajog.2023.08.026 Froeliger A et al. (2025) Prevalence and risk factors for postpartum depression 2 months after cesarean delivery: a prospective multicenter study. Am J Obstet Gynecol 232(5):491.e1-491.e11. https://doi.org/10.1016/j.ajog.2024.10.031 Fusar-Poli P, Hijazi Z, Stahl D, Steyerberg EW (2018) The Science of Prognosis in Psychiatry: A Review. JAMA Psychiatry 75(12):1289-1297. https://doi.org/10.1001/jamapsychiatry.2018.2530 Gelaye B, Rondon MB, Araya R, Williams MA (2016) Epidemiology of maternal depression, risk factors, and child outcomes in low-income and middle-income countries. Lancet Psychiatry 3(10):973-982. https://doi.org/10.1016/S2215-0366(16)30284-X Hagatulah N, Bränn E, Oberg AS, Valdimarsdóttir UA, Shen Q, Lu D (2024) Perinatal depression and risk of mortality: nationwide, register based study in Sweden. BMJ 384:e075462. https://doi.org/10.1136/bmj-2023-075462 Hochman E et al. (2021) Development and validation of a machine learning-based postpartum depression prediction model: A nationwide cohort study. Depress Anxiety 38(4):400-411. https://doi.org/10.1002/da.23123 Johannsen BM et al. (2020) Self-harm in women with postpartum mental disorders. Psychol Med 50(9):1563-1569. https://doi.org/10.1017/S0033291719001661 Johannsen BM, Larsen JT, Laursen TM, Bergink V, Meltzer-Brody S, Munk-Olsen T (2016) All-Cause Mortality in Women With Severe Postpartum Psychiatric Disorders. Am J Psychiatry 173(6):635-642. https://doi.org/10.1176/appi.ajp.2015.14121510 Khalifeh H, Hunt IM, Appleby L, Howard LM (2016) Suicide in perinatal and non-perinatal women in contact with psychiatric services: 15 year findings from a UK national inquiry. Lancet Psychiatry 3(3):233-242. https://doi.org/10.1016/S2215-0366(16)00003-1 Lim G, Farrell LM, Facco FL, Gold MS, Wasan AD (2018) Labor Analgesia as a Predictor for Reduced Postpartum Depression Scores: A Retrospective Observational Study. Anesth Analg 126(5):1598-1605. https://doi.org/10.1213/ANE.0000000000002720 Lindahl V, Pearson JL, Colpe L (2005) Prevalence of suicidality during pregnancy and the postpartum. Arch Womens Ment Health 8(2):77-87. https://doi.org/10.1007/s00737-005-0080-1 Liu L, Xu DR, Tong Y, Shi J, Zeng Z, Gong W (2022) Symptomatology in 1,112 women screened positive and negative using the Edinburgh postnatal depression scale (EPDS): longitudinal observations from the first trimester to 6 weeks postpartum of a Chinese cohort. J Psychosom Obstet Gynaecol 43(4):453-463. https://doi.org/10.1080/0167482X.2022.2052845 Nakamura A, van der Waerden J, Melchior M, Bolze C, El-Khoury F, Pryor L (2019) Physical activity during pregnancy and postpartum depression: Systematic review and meta-analysis. J Affect Disord 246:29-41. https://doi.org/10.1016/j.jad.2018.12.009 Norhayati MN, Hazlina NH, Asrenee AR, Emilin WM (2015) Magnitude and risk factors for postpartum symptoms: a literature review. J Affect Disord 175:34-52. https://doi.org/10.1016/j.jad.2014.12.041 Okun ML (2015) Sleep and postpartum depression. Curr Opin Psychiatry 28(6):490-496. https://doi.org/10.1097/YCO.0000000000000206 Payne JL et al. (2020) DNA methylation biomarkers prospectively predict both antenatal and postpartum depression. Psychiatry Res 285:112711. https://doi.org/10.1016/j.psychres.2019.112711 Pebryatie E, Paek SC, Sherer P, Meemon N (2022) Associations Between Spousal Relationship, Husband Involvement, and Postpartum Depression Among Postpartum Mothers in West Java, Indonesia. J Prim Care Community Health 13:21501319221088355. https://doi.org/10.1177/21501319221088355 Qi W et al. (2025) Prediction of postpartum depression in women: development and validation of multiple machine learning models. J Transl Med 23(1):291. https://doi.org/10.1186/s12967-025-06289-6 Rogers A et al. (2020) Association Between Maternal Perinatal Depression and Anxiety and Child and Adolescent Development: A Meta-analysis. JAMA Pediatr 174(11):1082-1092. https://doi.org/10.1001/jamapediatrics.2020.2910 Shovers SM, Bachman SS, Popek L, Turchi RM (2021) Maternal postpartum depression: risk factors, impacts, and interventions for the NICU and beyond. Curr Opin Pediatr 33(3):331-341. https://doi.org/10.1097/MOP.0000000000001011 Slomian J, Honvo G, Emonts P, Reginster JY, Bruyère O (2019) Consequences of maternal postpartum depression: A systematic review of maternal and infant outcomes. Womens Health (Lond) 15:1745506519844044. https://doi.org/10.1177/1745506519844044 Söderquist J, Wijma B, Thorbert G, Wijma K (2009) Risk factors in pregnancy for post-traumatic stress and depression after childbirth. BJOG 116(5):672-680. https://doi.org/10.1111/j.1471-0528.2008.02083.x Terada S et al. (2022) Relationship trajectories of pregnant women with their parents and postpartum depression: A hospital-based prospective cohort study in Japan. Front Psychiatry 13:961707. https://doi.org/10.3389/fpsyt.2022.961707 Thorsteinsson EB, Loi NM, Farr K (2018) Changes in stigma and help-seeking in relation to postpartum depression: non-clinical parenting intervention sample. PeerJ 6:e5893. https://doi.org/10.7717/peerj.5893 Trost SL et al. (2021) Preventing Pregnancy-Related Mental Health Deaths: Insights From 14 US Maternal Mortality Review Committees, 2008-17. Health Aff (Millwood) 40(10):1551-1559. https://doi.org/10.1377/hlthaff.2021.00615 US Preventive Services Task Force et al. (2019) Interventions to Prevent Perinatal Depression: US Preventive Services Task Force Recommendation Statement. JAMA 321(6):580-587. https://doi.org/10.1001/jama.2019.0007 Wakefield C, Frasch MG (2023) Predicting Patients Requiring Treatment for Depression in the Postpartum Period Using Common Electronic Medical Record Data Available Antepartum. AJPM Focus 2(3):100100. https://doi.org/10.1016/j.focus.2023.100100 Waller R et al. (2022) Clinician-reported childbirth outcomes, patient-reported childbirth trauma, and risk for postpartum depression. Arch Womens Ment Health 25(5):985-993. https://doi.org/10.1007/s00737-022-01263-3 Wang Z et al. (2021) Mapping global prevalence of depression among postpartum women. Transl Psychiatry 11(1):543. https://doi.org/10.1038/s41398-021-01663-6 Xiao M et al. (2020) Risk prediction for postpartum depression based on random forest. Zhong Nan Da Xue Xue Bao Yi Xue Ban 45(10):1215-1222. https://doi.org/10.11817/j.issn.1672-7347.2020.190655 Zhang R, Liu Y, Zhang Z, Luo R, Lv B (2025) Interpretable Machine Learning Model for Predicting Postpartum Depression: Retrospective Study. JMIR Med Inform 13:e58649. https://doi.org/10.2196/58649 Zhang Y, Jin S (2016) The impact of social support on postpartum depression: The mediator role of self-efficacy. J Health Psychol 21(5):720-726. https://doi.org/10.1177/1359105314536454 Zhang Y, Wang S, Hermann A, Joly R, Pathak J (2021) Development and validation of a machine learning algorithm for predicting the risk of postpartum depression among pregnant women. J Affect Disord 279:1-8. https://doi.org/10.1016/j.jad.2020.09.113 Zhong M, Zhang H, Yu C, Jiang J, Duan X (2022) Application of machine learning in predicting the risk of postpartum depression: A systematic review. J Affect Disord 318:364-379. https://doi.org/10.1016/j.jad.2022.08.070 Zhu Y, Xie Y, Yin X, Gong Y (2025) Feeding Patterns and Postpartum Depressive Symptoms: The Mediating Role of Parenting Self-Efficacy. Depress Anxiety 2025:2748707. https://doi.org/10.1155/da/2748707 Additional Declarations No competing interests reported. Supplementary Files TableS1.docx TableS2.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7079798","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":493188759,"identity":"2bc787e0-bc27-43bf-8f92-e7fc70b7f327","order_by":0,"name":"Hongmei Lin","email":"","orcid":"","institution":"Shaoxing Maternity and Child Health Care Hospital","correspondingAuthor":false,"prefix":"","firstName":"Hongmei","middleName":"","lastName":"Lin","suffix":""},{"id":493188761,"identity":"8f081d89-d646-4116-944b-0665d1f1730f","order_by":1,"name":"Chunfei Hu","email":"","orcid":"","institution":"Shaoxing Maternity and Child Health Care Hospital","correspondingAuthor":false,"prefix":"","firstName":"Chunfei","middleName":"","lastName":"Hu","suffix":""},{"id":493188763,"identity":"d3449372-d353-4610-be13-e1ac665a7e7e","order_by":2,"name":"Siqian Hu","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Siqian","middleName":"","lastName":"Hu","suffix":""},{"id":493188765,"identity":"1e515eec-3bdc-40c1-b72c-6c2a5ff939ee","order_by":3,"name":"Ying Ying","email":"","orcid":"","institution":"Shaoxing Maternity and Child Health Care Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Ying","suffix":""},{"id":493188766,"identity":"a08fcced-ef9b-4e45-a418-ff177b89e723","order_by":4,"name":"Lingling Shang","email":"","orcid":"","institution":"Shaoxing Maternity and Child Health Care Hospital","correspondingAuthor":false,"prefix":"","firstName":"Lingling","middleName":"","lastName":"Shang","suffix":""},{"id":493188771,"identity":"3ffa88ae-9964-441f-a1ce-aa9f81270a65","order_by":5,"name":"Xiaojing Qiu","email":"","orcid":"","institution":"Shengzhou Maternity and Child Health Care Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaojing","middleName":"","lastName":"Qiu","suffix":""},{"id":493188772,"identity":"27aea5cd-c73c-40b8-80eb-42d4d834b662","order_by":6,"name":"Qiong Luo","email":"","orcid":"","institution":"Women's Hospital, School of Medicine, Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Qiong","middleName":"","lastName":"Luo","suffix":""},{"id":493188773,"identity":"bab76125-2d2a-46f8-8630-295776b6905f","order_by":7,"name":"Hualin Xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyElEQVRIiWNgGAWjYFACHiAukGDgB1KfSdBiIMEg2cDAOJsULUB0gFgtBjdyj0n8MLDIMz5/xrC5sI1Bnl/sAH4tkjPy0iR7DCSKzW6kJTbPbGMwnDk7Ab8WfokcMwkeA4nEbTeYjz/mbWNIMLhNQAsbUIvkH6CWzf0HG5uJ0gKyRRpkywaG5IPEaZHseWNsLQPUMgPklxnnJAj7xeB4juHNNxV1if39wBArKLOR55cmoAUVMLJJkKIcDP6QrGMUjIJRMApGAAAATCA9/AtqPM0AAAAASUVORK5CYII=","orcid":"","institution":"Shaoxing Maternity and Child Health Care Hospital","correspondingAuthor":true,"prefix":"","firstName":"Hualin","middleName":"","lastName":"Xu","suffix":""}],"badges":[],"createdAt":"2025-07-09 04:53:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7079798/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7079798/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88093853,"identity":"942e5b4f-c186-43e7-842a-947c17744a0c","added_by":"auto","created_at":"2025-08-01 10:42:39","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":91464,"visible":true,"origin":"","legend":"\u003cp\u003eBoruta Ridge Plot\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7079798/v1/036d17ce9537f141a70c7573.jpg"},{"id":88093855,"identity":"d1ad3cc1-cee1-46ad-b876-4f963c7329cf","added_by":"auto","created_at":"2025-08-01 10:42:39","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":95627,"visible":true,"origin":"","legend":"\u003cp\u003eROC Curves Comparison\u003c/p\u003e\n\u003cp\u003eA ROC curves on training set, B ROC curves on validation set\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7079798/v1/bd37dabf187004e64e3f8170.jpg"},{"id":88093863,"identity":"3b074305-9280-4360-a900-490b0452f44f","added_by":"auto","created_at":"2025-08-01 10:42:39","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":101190,"visible":true,"origin":"","legend":"\u003cp\u003eModel Calibration Curves\u003c/p\u003e\n\u003cp\u003eA Calibration curves on training set, B Calibration curves on validation set\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7079798/v1/9e0a2a32d79dc2a7aa5632a9.jpg"},{"id":88093864,"identity":"085c2a37-d912-47e8-b85a-b2ef2b728cf5","added_by":"auto","created_at":"2025-08-01 10:42:39","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":84786,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP Analysis\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7079798/v1/82cd8275ec9f7156ea9f4d7a.jpg"},{"id":96708493,"identity":"2ee45c21-4551-4ced-a84a-1c4a05e2dd8d","added_by":"auto","created_at":"2025-11-25 10:03:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1511190,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7079798/v1/3dea4035-1df1-4308-8ec8-13ff029aa398.pdf"},{"id":88094759,"identity":"fa260b6c-067e-461b-84f7-4d0d422350a8","added_by":"auto","created_at":"2025-08-01 10:50:39","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":80391,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7079798/v1/ead46b04ab952ffa5d875a8d.docx"},{"id":88096690,"identity":"f813ef70-4bd0-4c13-95f4-553704ce72ad","added_by":"auto","created_at":"2025-08-01 10:58:39","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":13386,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.docx","url":"https://assets-eu.researchsquare.com/files/rs-7079798/v1/beaa0b9bd3340b157a34abc2.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Clinically Practical Postpartum Depression Predictor: Machine Learning Model Based on Simplified Indicators","fulltext":[{"header":"Article Highlights","content":"\u003cp\u003e1. Large-scale multicenter cohort (n=5011) reveals 13.3% PPD prevalence and identifies key risk factors.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e2. 11 ML models predict PPD risk using 9 easily accessible sociodemographic and delivery indicators.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e3. Models confirm feasibility of using simplified indicators for PPD screening in resource-limited settings.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e4. SHAP analysis identifies optimizing birth experience and strengthening social support as actionable PPD prevention targets.\u0026nbsp;\u003c/p\u003e"},{"header":"‌Introduction","content":"\u003cp\u003ePostpartum depression (PPD) is the most prevalent mental health disorder during the perinatal period, with a global pooled prevalence of 17.2%(Wang et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Beyond impairing maternal parenting capacity, PPD adversely affects children's emotional, cognitive, and behavioral development(Lim et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Rogers et al. \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) and substantially elevates maternal self-harm risk(Johannsen et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Hagatulah et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Epidemiological studies indicate that 20% of postpartum maternal deaths result from suicide(Lindahl et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), underscoring the critical need to develop clinically actionable prediction tools as part of multifaceted preventive strategies(Khalifeh et al. \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eCurrent clinical practice primarily follows the Grade B recommendation from the U.S. Preventive Services Task Force (USPSTF), utilizing self-report tools such as the Edinburgh Postnatal Depression Scale (EPDS) for routine screening(US Preventive Services Task Force et al. 2019). However, such scale-based screening carries inherent limitations in timeliness, typically detecting only individuals who already exhibit symptoms at the time of assessment(Liu et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Consequently, there is a critical need to develop tools capable of predicting depression predisposition before symptom onset, enabling timely support and intervention.\u003c/p\u003e\u003cp\u003eMachine learning(ML) technologies have recently provided novel methodological frameworks for PPD risk prediction. A meta-analysis of 17 studies encompassing 62 prediction models(Zhong et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) revealed that most models integrating multidimensional predictors such as sociodemographic characteristics, obstetric history, psychiatric diagnoses, and biomarkers from electronic health records (EHRs) demonstrate robust predictive performance (AUC 0.660\u0026ndash;0.969). Nevertheless, such models face challenges in clinical application: complex architectures with over 20 predictors hinder clinical utility(Hochman et al. \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Zhang et al. \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), while reliance on non-standardized data (e.g., serological or genetic markers) limits their applicability in primary care settings(Payne et al. \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Xiao et al. \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eEvidence confirms significant associations between PPD risk and factors including low social support, economic stress, marital discord, and traumatic birth experiences(S\u0026ouml;derquist et al. \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Cho et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Waller et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Accordingly, this study utilized ML to develop a concise PPD prediction model, prioritizing the predictive power of sociodemographic features. Designed for clinical utility, the model requires only a few readily simplified variables for risk stratification. Using a retrospective cohort design and multiple ML algorithms, we aim to optimize the balance between predictive accuracy and clinical practicality, ultimately providing primary care with an accessible and implementable PPD risk screening tool.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003eStudy Population\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study extends our prior research developing and validating a prenatal depression prediction model. From an established cohort, we enrolled 20,950 women in late pregnancy (≥34 weeks) who received prenatal care and completed questionnaires at the Women's Hospital of Zhejiang University School of Medicine, Shaoxing Maternity and Child Health Care Hospital, and Shengzhou Maternity and Child Health Care Hospital. Ultimately, 5,011 participants delivered at these hospitals between January 1, 2020, and April 15, 2024, and completed valid follow-up at 6 weeks postpartum, forming the analytical cohort. All participants provided written informed consent. The research protocol was developed in compliance with the Declaration of Helsinki and formally approved by the Medical Ethics Committee of Shaoxing Maternity and Child Health Care Hospital (Approval No. 2025-024-01).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData were obtained from self-administered prenatal questionnaires, postpartum questionnaires, and obstetric electronic health records (EHRs). Questionnaire data originated from the Perinatal and Postpartum Depression Information System. The third-trimester survey covered sociodemographics, social support, lifestyle habits, family history, obstetric history, medical history, pregnancy complications, and Edinburgh Postnatal Depression Scale (EPDS) assessments.\u003c/p\u003e\n\u003cp\u003eTo mitigate misreporting bias in pregnancy-related conditions due to participants' limited medical knowledge, EHRs were used to extract discharge diagnoses for medical history, pregnancy complications, and delivery outcomes (delivery mode, gestational age at birth, neonatal birth weight). At 6-week postpartum follow-up, participants completed another questionnaire assessing postpartum EPDS, delivery experiences, and infant care practices.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOutcome Definition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe primary outcome was PPD, defined as an EPDS score ≥10 at 6 weeks postpartum. This threshold was selected to improve screening sensitivity and reduce missed diagnoses compared to the conventional cutoff of 13, as evidence suggests it is more appropriate for Asian populations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePredictor Variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePredictors derived from questionnaires and EHRs included:\u003c/p\u003e\n\u003cp\u003e1)\u003cem\u003eSociodemographics characteristics\u003c/em\u003e, assessed via prenatal questionnaires, encompassed depression risk factors established in our prior research: social support, sleep patterns, physical activity, pregnancy planning, obstetric history, and family history(e.g., hyperlipidemia, schizophrenia, anxiety/depression).\u003c/p\u003e\n\u003cp\u003e2)\u003cem\u003ePregnancy-related factors\u003c/em\u003e, extracted from EHRs, comprised medical history (e.g., anxiety/depression, chronic hypertension), pregnancy complications (e.g., preeclampsia, gestational diabetes, threatened abortion, placenta previa, fetal distress), and delivery outcomes (delivery mode, gestational age at delivery, neonatal birth weight).\u003c/p\u003e\n\u003cp\u003e3)\u003cem\u003ePostpartum factors\u003c/em\u003e, assessed through postpartum questionnaires, included:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDelivery experiences: trial of labor preceding cesarean, concordance between expected and actual delivery mode, intrapartum oxytocin administration, epidural analgesia during labor, family presence at delivery, and subjective birth experience rating.\u003c/p\u003e\n\u003cp\u003eInfant care practices: neonatal congenital anomalies, postnatal hospitalization, current feeding method, and nighttime feeding frequency.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel Development and Evaluation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe entire dataset was randomly split in a 7:3 ratio into a training set (n=3508) and a validation set (n=1503). The training set was used for model development, while the validation set was used for model evaluation. Utilizing variables selected via univariate logistic regression and the Boruta algorithm based on the training set, this study constructed 11 clinically prevalent prediction models, including Logistic Regression, Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost, AdaBoost, Multilayer Perceptron (MLP), Support Vector Machine (SVM), Naive Bayes (NB), and Decision Tree (DT). Model hyperparameters were optimized via grid search with manual refinement. Following optimization, all models underwent evaluation through 5-fold cross-validation using the dataset. Evaluation metrics comprised the area under the ROC curve (AUC), F1-score, sensitivity, and specificity. Comparative analysis across these metrics identified the optimal model as the final predictive model. The selected model was subjected to SHapley Additive exPlanations (SHAP) analysis for visual interpretation of predictions. Additionally, calibration curves were generated to assess the agreement between predicted probabilities and observed outcomes, facilitating detection of overfitting or underfitting.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eContinuous variables are presented as mean ± \u003cem\u003eSD\u003c/em\u003e (normally distributed) or median (IQR) (non-normally distributed); categorical variables are presented as number (%). Between-group comparisons were performed as follows: for continuous variables, \u003cem\u003et\u003c/em\u003e-test (normally distributed with equal variance), \u003cem\u003et\u003c/em\u003e'-test (normally distributed with unequal variance), or Wilcoxon rank-sum test (non-normally distributed); for categorical variables, \u003cem\u003eχ\u003c/em\u003e² test or corrected \u003cem\u003eχ\u003c/em\u003e² test, or Fisher’s exact test if total sample size \u0026lt; 40 or minimum expected cell count \u0026lt; 1; for ordinal data, Wilcoxon rank-sum test. Univariate logistic regression results are reported as odds ratios (\u003cem\u003eOR\u003c/em\u003e) with 95% confidence interval (\u003cem\u003eCI\u003c/em\u003e). Statistical significance was defined as a two-tailed \u003cem\u003eP\u003c/em\u003e-value \u0026lt; 0.05. All statistical analyses and plotting were performed using Python 3.9 software.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eBaseline Characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study included 5,011 participants aged 16.00\u0026ndash;47.00 years. The mean age was 28.77\u0026thinsp;\u0026plusmn;\u0026thinsp;3.68 years, with a median of 28.00 years (IQR: 26.00\u0026ndash;31.00). Participants had a mean EPDS score of 4.55\u0026thinsp;\u0026plusmn;\u0026thinsp;4.57 and a median score of 3.00 (IQR: 1.00\u0026ndash;7.00). PPD developed in 667 participants, representing a prevalence of 13.3%.\u003c/p\u003e\n\u003cp\u003eAs this study aimed to predict PPD risk using simplified data, antenatal EPDS scores were excluded as predictive variables. Eighty-one potential predictors were ultimately included. Differential analysis conducted on the training set identified 24 factors significantly associated with PPD (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), including marital status, sleep duration, previous delivery experience, current delivery experience, and social support (Supplementary Table\u0026nbsp;1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUnivariate Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTable \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e presents the univariate logistic regression results for statistically significant variables. Higher annual household income, longer marital duration, positive previous delivery experience, and strong support from family and friends emerged as protective factors against PPD (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;1, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Conversely, smoking, alcohol consumption, unplanned pregnancy, and multiple induced abortions were identified as risk factors (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;1, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Single and divorced women faced significantly higher PPD risk compared to married women (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2.463, 95% CI: 1.128\u0026ndash;4.962; \u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;13.298, 95% CI: 2.588\u0026ndash;96.128). Difficulty falling asleep increased depression risk (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.008, 95% CI: 1.004\u0026ndash;1.012), whereas longer sleep duration (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.847, 95% CI: 0.788\u0026ndash;0.910) and walking time (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.994, 95% CI: 0.972\u0026ndash;1.012) reduced risk.\u003c/p\u003e\n\u003cp\u003eAmong obstetric factors, placental malposition (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;2.877, 95% CI: 1.101\u0026ndash;6.778) and intrauterine fetal distress (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.414, 95% CI: 1.016\u0026ndash;1.926) significantly increased PPD risk. Unexpectedly, gestational diabetes (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.635, 95% CI: 0.413\u0026ndash;0.938) and hypothyroidism (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.457, 95% CI: 0.192\u0026ndash;0.920) were associated with reduced risk. Regarding postpartum factors, delivery methods inconsistent with expectation increased PPD risk by 1.6-fold (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.614, 95% CI: 1.269\u0026ndash;2.039), whereas smooth and satisfactory birth experiences substantially reduced risk (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.183\u0026ndash;0.335, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Nighttime feeding 1\u0026ndash;2 times suggested a potential protective trend when compared to no feeding (\u003cem\u003eOR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.567, 95% CI: 0.260\u0026ndash;1.420) or frequent feeding.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eUnivariate Logistic Regression Results\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNo.\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePredictor\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eOR\u003c/em\u003e (95%CI)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eStatistics\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrepregnancy BMI(kg/m\u0026sup2;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.951 ( 0.921\u0026ndash;0.981 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9.709\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAnnual household income(CNY)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;30k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e30k-80k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.782 ( 0.461\u0026ndash;1.370 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.790\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.374\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e80k-150k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.566 ( 0.346\u0026ndash;0.963 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.796\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.029\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e150k-300k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.574 ( 0.352\u0026ndash;0.973 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.635\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.031\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e300k-500k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.412 ( 0.226\u0026ndash;0.764 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8.202\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.004\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026gt;\u0026thinsp;500k\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.588 ( 0.293\u0026ndash;1.172 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.268\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.132\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarital status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarried\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSingle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.463 ( 1.128\u0026ndash;4.962 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.818\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.016\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDivorced\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e13.298 ( 2.588\u0026ndash;96.128 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8.898\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.003\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarriage duration\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUnmarried\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;1yr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.374 ( 0.196\u0026ndash;0.744 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8.515\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.004\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1-10yrs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.295 ( 0.157\u0026ndash;0.580 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e13.587\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ge;\u0026thinsp;10yrs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.222 ( 0.102\u0026ndash;0.490 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.288\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmoking history\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.166 ( 1.185\u0026ndash;3.757 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.991\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.008\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAlcohol use history\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.304 ( 1.067\u0026ndash;1.591 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.802\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.009\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePregnancy plan\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePlanned\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUnplanned\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.238 ( 1.007\u0026ndash;1.518 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.153\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.042\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrior induced abortion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOnce\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.075 ( 0.837\u0026ndash;1.371 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.334\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.563\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTwice\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.330 ( 0.857\u0026ndash;1.996 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.184\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eThree times or more\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.115 ( 1.178\u0026ndash;3.609 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.964\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.008\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrevious birth experience rating\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrimipara\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eExtremely Negative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.013 ( 0.934\u0026ndash;8.534 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.008\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.045\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.988 ( 0.472\u0026ndash;1.861 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.972\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNeutral\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.880 ( 0.666\u0026ndash;1.149 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.843\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.359\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.606 ( 0.412\u0026ndash;0.866 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.996\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.008\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eExtremely Positive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.402 ( 0.156\u0026ndash;0.850 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.588\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.032\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSupport from spouse\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMinimal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.635 ( 1.104\u0026ndash;6.844 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.427\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.035\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.472 ( 0.712\u0026ndash;3.448 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.947\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.330\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStrong\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.689 ( 0.340\u0026ndash;1.593 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.916\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.339\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSupport from parents\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMinimal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.190 ( 0.590\u0026ndash;2.390 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.240\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.624\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.713 ( 0.425\u0026ndash;1.225 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.585\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.208\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStrong\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.388 ( 0.244\u0026ndash;0.636 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15.171\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSupport from parents-in-law\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMinimal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.186 ( 0.696\u0026ndash;2.022 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.397\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.529\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.813 ( 0.535\u0026ndash;1.256 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.908\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.340\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStrong\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.404 ( 0.276\u0026ndash;0.605 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20.548\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSupport from siblings\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMinimal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.568 ( 0.902\u0026ndash;2.644 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.709\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.037 ( 0.750\u0026ndash;1.433 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.048\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.827\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStrong\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.545 ( 0.422\u0026ndash;0.709 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e21.068\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSupport from friends\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMinimal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.514 ( 0.908\u0026ndash;2.508 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.579\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.108\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.663 ( 0.459\u0026ndash;0.967 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.700\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.030\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStrong\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.451 ( 0.325\u0026ndash;0.636 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e21.716\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSleep latency(minutes)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.008 ( 1.004\u0026ndash;1.012 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e18.046\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSleep duration(hours)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.847 ( 0.788\u0026ndash;0.910 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20.666\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWeekly walking hours during\u003c/p\u003e\n \u003cp\u003epregnancy(hours)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.994 ( 0.972\u0026ndash;1.012 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.387\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.534\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGDM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.635 ( 0.413\u0026ndash;0.938 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.757\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.029\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePlacenta previa low-lying\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.877 ( 1.101\u0026ndash;6.778 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.373\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.020\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypothyroidism\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.457 ( 0.192\u0026ndash;0.920 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.940\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.047\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFetal distress\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.412 ( 1.016\u0026ndash;1.926 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.486\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.034\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDelivery mode consistency with\u003c/p\u003e\n \u003cp\u003eexpectation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.614 ( 1.269\u0026ndash;2.039 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15.658\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBirth experience rating\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraumatic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDifficult but Worthwhile\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.335 ( 0.253\u0026ndash;0.447 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e56.686\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmooth\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.329 ( 0.240\u0026ndash;0.451 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e47.679\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVery Satisfactory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.183 ( 0.127\u0026ndash;0.264 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e82.519\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u0026thinsp;0.0001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNighttime feeding frequency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNone\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOnce or Twice\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.567 ( 0.260\u0026ndash;1.420 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.769\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.184\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eThree or four times\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.740 ( 0.342\u0026ndash;1.846 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.504\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.478\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFive times or more\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.932 ( 0.389\u0026ndash;2.491 )\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.023\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.881\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003eVariable Selection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUsing the Boruta algorithm, we screened statistically significant variables identified through univariate logistic regression analysis. As shown in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, nine key predictors(blue ridge) were confirmed as essential features for the PPD prediction model: support from spouse, support from parents, support from parents-in-law, support from siblings, support from friends, marital status, marriage duration, fetal distress and birth experience rating.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel Development and Evaluation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWith PPD occurrence as the dependent variable and the aforementioned nine features as independent variables, we constructed eleven ML models: LogisticRegression, RandomForest, GradientBoosting, XGBoost, LightGBM, CatBoost, AdaBoost, MLP, SVM, NB, DT. Model hyperparameters were optimized through 5-fold stratified cross-validation combined with randomized search, using ROC-AUC as the primary evaluation metric. Optimized hyperparameters for each model are detailed in Supplementary Table\u0026nbsp;2.\u003c/p\u003e\n\u003cp\u003eAs presented in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e, the Random Forest model demonstrated relatively superior predictive performance. In the training set, it achieved an AUC of 0.725 (95% CI: 0.697\u0026ndash;0.752), with F1-score, sensitivity, and specificity values of 0.699, 0.660, and 0.630 respectively. Corresponding values for the validation set were AUC\u0026thinsp;=\u0026thinsp;0.612 (95% CI: 0.568\u0026ndash;0.656), F1-score\u0026thinsp;=\u0026thinsp;0.667, sensitivity\u0026thinsp;=\u0026thinsp;0.520, and specificity\u0026thinsp;=\u0026thinsp;0.617. Figure 3 displays calibration curves for all predictive models in both training and validation sets. For Random Forest, all points approach the ideal line, indicating satisfactory calibration.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003ePerformance Comparison of All Models\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNo.\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eModel\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDataset\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC 95%CI\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eF1-score\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eLogistic\u003c/p\u003e\n \u003cp\u003eRegression\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.660\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.632\u0026ndash;0.688\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.748\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.497\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.740\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.616\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.572\u0026ndash;0.660\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.737\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.415\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.741\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eRandom\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eForest\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eTraining\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.725\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.697\u0026ndash;0.752\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.699\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.660\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.639\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eValidation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.612\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.568\u0026ndash;0.656\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.667\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.520\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.617\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eGradient\u003c/p\u003e\n \u003cp\u003eBoosting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.711\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.684\u0026ndash;0.739\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.720\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.632\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.674\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.616\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.572\u0026ndash;0.659\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.697\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.485\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.665\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.701\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.673\u0026ndash;0.728\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.724\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.615\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.683\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.627\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.583\u0026ndash;0.671\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.701\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.485\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.672\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eLightGBM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.717\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.690\u0026ndash;0.745\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.784\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.538\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.789\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.619\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.575\u0026ndash;0.663\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.760\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.405\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.778\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eCatBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.685\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.657\u0026ndash;0.713\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.702\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.612\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.652\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.610\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.566\u0026ndash;0.654\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.685\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.510\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.645\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eAdaBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.660\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.631\u0026ndash;0.688\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.718\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.580\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.680\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.624\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.580\u0026ndash;0.668\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.708\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.500\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.680\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eMLP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.710\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.683\u0026ndash;0.738\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.494\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.852\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.365\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.603\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.559\u0026ndash;0.647\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.494\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.740\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.376\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eSVM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.601\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.572\u0026ndash;0.630\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.270\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.921\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.160\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.595\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.551\u0026ndash;0.639\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.254\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.890\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.150\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eNB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.653\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.624\u0026ndash;0.681\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.672\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.604\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.610\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.582\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.538\u0026ndash;0.626\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.654\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.505\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.600\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTraining\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.677\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.649\u0026ndash;0.705\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.525\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.760\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eValidation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.606\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.562\u0026ndash;0.650\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.752\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.380\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.771\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003eSHAP Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe employed SHAP analysis to interpret the Random Forest model\u0026apos;s PPD prediction logic. Figure \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e presents complementary visualizations: the left beeswarm plot illustrates feature impact direction and magnitude through Shapley values, while the right donut chart quantifies global feature importance. Both consistently identify BirthExperienceRating, SupportFromSpouse, and SupportFromParentsInlaw as core predictors.\u003c/p\u003e\n\u003cp\u003eIn the beeswarm plot, the horizontal axis represents Shapley values and the vertical axis lists features, with red indicating high feature values and blue indicating low values. For BirthExperienceRating, most data points cluster near zero (blue), suggesting neutral impact at moderate ratings (\u0026quot;Difficult but Worthwhile\u0026quot;). Scattered high-value points (red, \u0026gt;\u0026thinsp;0.1) indicate reduced PPD risk at higher satisfaction levels. SupportFromSpouse exhibits a right-skewed distribution (blue-dominant), confirming its protective effect though weaker than BirthExperienceRating. Notably, single and divorced women show elevated risk compared to married individuals.\u003c/p\u003e\n\u003cp\u003eThe donut chart quantifies global contributions: BirthExperienceRating accounts for 27.3% of predictive power, followed by SupportFromParentsInlaw (19.3%) and SupportFromSpouse (18.5%). These three features collectively constitute 65% of the model\u0026apos;s predictive capability, forming the core feature set.\u003c/p\u003e\n\u003cp\u003eClinically, these findings directly support our PPD prediction objectives by demonstrating: 1) the critical importance of optimizing birth experiences given BirthExperienceRating\u0026apos;s dominance, and 2) the empirical foundation for family-centered interventions leveraging spousal and parental support systems.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study developed eleven ML models for PPD risk prediction using nine readily accessible sociodemographic and delivery-related indicators: support from spouses, parents, parents-in-law, siblings, and friends; fetal distress; birth experience rating; marital status; and marriage duration. Among these, the Random Forest model demonstrated relatively superior performance in the validation cohort (AUC\u0026thinsp;=\u0026thinsp;0.61). While this predictive accuracy falls within the moderate range, it confirms the fundamental feasibility of using simplified clinical data for PPD risk stratification. This approach offers a practical screening solution particularly valuable for resource-limited healthcare settings.\u003c/p\u003e\u003cp\u003ePPD poses a serious threat to maternal health with multifaceted consequences, including mood disorders, increased risk of child neglect(Slomian et al. \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; US Preventive Services Task Force et al. 2019), and in severe cases, emergence of suicidal or infanticidal ideation(Johannsen et al. \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; US Preventive Services Task Force et al. 2019; Johannsen et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Crucially, maternal mortality review committees recognize that deaths attributable to mental health complications demonstrate substantially higher preventability compared to other pregnancy-related mortality causes(Trost et al. \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Early identification of high-risk individuals therefore represents a critical first step in establishing timely intervention pathways, creating a vital prevention window to mitigate adverse outcomes.\u003c/p\u003e\u003cp\u003eFrom an implementation perspective, our approach addresses significant limitations in conventional PPD screening. Traditional methods often require specialized personnel training, hindering large-scale deployment in resource-constrained environments. In contrast, our ML framework enables rapid, cost-efficient risk assessment within primary care settings. The required predictive variables (including marital status, social support levels, and birth experience) can be efficiently captured immediately postpartum through concise questionnaires, eliminating dependence on biomarker testing or complex psychological assessments. This streamlined data collection significantly reduces clinician workload while enabling efficient risk stratification with minimal resources, ultimately improving screening coverage and facilitating appropriate referrals to prevent severe psychiatric sequelae.\u003c/p\u003e\u003cp\u003eAlthough existing literature documents numerous ML models with good generalization capacity for PPD prediction(Cellini et al. \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), their clinical translation faces substantial barriers(Fusar-Poli et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). The primary limitation lies in the tension between model complexity and clinical utility. Most existing predictive frameworks require multidimensional inputs spanning sociodemographic characteristics, obstetric history, biological markers, genetic indicators, and psychological assessments(Cellini et al. \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Zhong et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). For example, Clapp et al.'s robust 26-variable model (AUC\u0026thinsp;=\u0026thinsp;0.750)(Clapp et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) proves impractical in routine clinical workflows due to its extensive data requirements. Such models inherently presuppose well-established prenatal screening infrastructures\u0026mdash;resources frequently unavailable where prenatal coverage remains suboptimal(Gelaye et al. \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Similarly, Qi et al.(Qi et al. \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) integrated multiple psychosocial scales and physiological indicators, ultimately identifying 11 predictors including prenatal depression, neuroticism, total cholesterol, and serum free triiodothyronine, with the model achieving AUC values of 0.787\u0026ndash;0.858. However, these specialized physiological and psychological parameters fall beyond the scope of routinely collected data in primary care settings, thereby limiting the model's generalizability.\u003c/p\u003e\u003cp\u003eA second significant limitation concerns the practical applicability of models over-reliant on mental health history(Zhang et al. \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Many studies prioritize predictive accuracy by emphasizing psychiatric antecedents, utilizing prior diagnoses and medication histories as core predictors. Wakefield et al.'s high-performance model (AUC\u0026thinsp;=\u0026thinsp;0.91)(Wakefield and Frasch \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), for instance, requires nine variables including pre-pregnancy depression history and early-pregnancy psychotropic medication use. However, screening individuals with established psychiatric vulnerabilities offers limited clinical value since these conditions already constitute recognized PPD risk factors. More importantly, stigma-induced information suppression presents a fundamental barrier: affected women may deliberately conceal psychiatric histories or resist psychological evaluations due to perceived shame, inevitably leading to critical data gaps and diagnostic omissions(Thorsteinsson et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eBeyond predictive functionality, our SHAP interpretability analysis reveals two principal modifiable factors. First, optimizing birth experience proves paramount: our quantification shows that women reporting \"very satisfactory\" deliveries experience an 82% reduction in PPD susceptibility, unequivocally demonstrating the protective value of patient-centered intrapartum care. Patient-reported birth trauma is a critical predictor of adverse psychological outcomes, with established links between traumatic delivery exposure and subsequent depression(Waller et al. \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), including comorbid postpartum post-traumatic stress disorder(Dekel et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Mechanistically, negative vaginal birth experiences frequently stem from induction procedures or instrumental deliveries(Froeliger et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), while cesarean-related trauma often relates to emergency surgery or inadequate postoperative analgesia(Froeliger et al. \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Collectively, this evidence confirms that negative birth perceptions\u0026mdash;regardless of delivery mode\u0026mdash;significantly elevate PPD risk. This necessitates evidence-based refinement of induction protocols, enhanced obstetric technical proficiency, optimized perioperative pain management, and strengthened multidisciplinary collaboration among obstetricians, midwives, and anesthesiologists(Du et al. \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eSecond, strengthening social support systems is essential: our analysis confirms that deficient spousal or in-law support substantially increases PPD vulnerability, highlighting the need for structured family counseling. Zhang and Jin's research(Zhang and Jin \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) indicates that social support indirectly mitigates PPD through enhanced maternal self-efficacy. Indonesian cohort studies(Pebryatie et al. \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) demonstrate that proactive spousal involvement during pregnancy promotes healthy behaviors and significantly reduces postpartum depressive symptoms. Furthermore, Japanese research(Terada et al. \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) reveals that women experiencing deteriorating parental relationships or persistently unsatisfactory family dynamics face 2.81-fold and 2.39-fold increases in PPD susceptibility, respectively. We therefore advocate explicitly incorporating spousal support expectations in antenatal education materials and implementing community-based interventions that actively engage family participation.\u003c/p\u003e\u003cp\u003eAdditionally, our study identified other clinically relevant predictors. Fetal distress demonstrated significant predictive value, likely mediated through neonatal resuscitation or NICU admission pathways\u0026mdash;known to elevate maternal depression prevalence to approximately 40%(Shovers et al. \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Placental positional abnormalities were also identified as risk factors, likely attributable to anxiety regarding adverse pregnancy outcomes precipitated by recurrent vaginal bleeding. Notably, gestational diabetes and hypothyroidism showed paradoxical protective associations, contradicting some prior research(Azami et al. \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Costantine et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). This may reflect generally milder phenotypes in our cohort, effective clinical management, or enhanced support systems via intensified prenatal monitoring. Other sociodemographic factors\u0026mdash;including socioeconomic deprivation, marital discord, and tobacco or alcohol use\u0026mdash;have been established as consistent risk amplifiers for PPD, aligning with prior research(Norhayati et al. \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Wang et al. \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Conversely, ensuring adequate sleep duration(Okun \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), engaging in regular moderate-intensity physical activity(Nakamura et al. \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and implementing flexible breastfeeding practices(Zhu et al. \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) demonstrate protective benefits against depression risk.\u003c/p\u003e\u003cp\u003eThis study developed a clinically practical PPD screening tool using only 9 simplified indicators, enabling rapid risk stratification in resource-limited settings. In clinical practice, frontline clinicians can administer 9 predetermined questions to postpartum women, enabling automated risk assessment through this instrument. This approach facilitates early risk identification, allowing timely interventions for high-risk individuals to effectively prevent adverse outcomes during the puerperium and beyond.\u003c/p\u003e\u003cp\u003eNevertheless, several limitations warrant acknowledgment. The model's moderate predictive performance (AUC\u0026thinsp;=\u0026thinsp;0.61) falls short of optimal clinical standards. However, given its minimal variable requirements, this achievement retains practical value as a preliminary screening instrument in primary care contexts. Although rigorous internal validation was conducted through cross-validation techniques, the lack of external validation across independent cohorts remains a critical implementation constraint. Future studies should consequently evaluate its generalizability and robustness in diverse populations and geographic regions. Finally, the prediction timeframe is confined to the postpartum period; since some PPD pathogenesis may originate prenatally, subsequent research should integrate first- and second-trimester data to enable earlier risk prediction during pregnancy, thereby further advancing the prevention window.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study successfully developed and validated a ML-based PPD screening framework using 9 simplified sociodemographic and delivery-related indicators. Its strengths include resource-efficient implementation and the identification of two actionable intervention targets: systematic enhancement of birth experiences and strategic reinforcement of social support networks. Despite moderate predictive performance, this tool demonstrates significant potential for scalable PPD screening in resource-limited regions, enabling early risk identification to reduce severe PPD incidence and prevent devastating outcomes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll participants provided written informed consent. The research protocol was developed in compliance with the Declaration of Helsinki and formally approved by the Medical Ethics Committee of Shaoxing Maternity and Child Health Care Hospital (Approval No. 2025-024-01).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e This work was supported by the Shaoxing Health Science and Technology Plan (Grant number 2024SKY048). The funding agency had no role in the design of the study, the collection, analysis, and interpretation of the data, or in writing the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll participants provided written informed consent. The research protocol was developed in compliance with the Declaration of Helsinki and formally approved by the Medical Ethics Committee of Shaoxing Maternity and Child Health Care Hospital (Approval No. 2025-024-01).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Shaoxing Health Science and Technology Plan (Grant number 2024SKY048). The funding agency had no role in the design of the study, the collection, analysis, and interpretation of the data, or in writing the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e1.Hongmei Lin(first author): Conceptualization, Resources, Writing - original draft preparation2.Chunfei Hu(first author):Methodology, Formal analysis, Writing - original draft preparation3.Siqian Hu:Data curation4.Ying Ying:Data collection5.Lingling Shang:Data collection6.Xiaojing Qiu:Resources7.Qiong Luo(Corresponding author): Resources,Supervision, Writing - review and editing8.Hualin Xu(Corresponding author):Conceptualization, Funding acquisition, SupervisionAll authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors are grateful to the women who participatedin and donated clinical data for this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAzami M, Badfar G, Soleymani A, Rahmati S (2019) The association between gestational diabetes and postpartum depression: A systematic review and meta-analysis. Diabetes Res Clin Pract 149:147-155. https://doi.org/10.1016/j.diabres.2019.01.034\u003c/li\u003e\n\u003cli\u003eCellini P, Pigoni A, Delvecchio G, Moltrasio C, Brambilla P (2022) Machine learning in the prediction of postpartum depression: A review. J Affect Disord 309:350-357. https://doi.org/10.1016/j.jad.2022.04.093\u003c/li\u003e\n\u003cli\u003eCho H et al. (2022) Association between social support and postpartum depression. Sci Rep 12(1):3128. https://doi.org/10.1038/s41598-022-07248-7\u003c/li\u003e\n\u003cli\u003eClapp MA, Castro VM, Verhaak P, McCoy TH, Shook LL, Edlow AG, Perlis RH (2025) Stratifying Risk for Postpartum Depression at Time of Hospital Discharge. Am J Psychiatry :appiajp20240381. https://doi.org/10.1176/appi.ajp.20240381\u003c/li\u003e\n\u003cli\u003eCostantine MM et al. (2020) Effect of Thyroxine Therapy on Depressive Symptoms Among Women With Subclinical Hypothyroidism. Obstet Gynecol 135(4):812-820. https://doi.org/10.1097/AOG.0000000000003724\u003c/li\u003e\n\u003cli\u003eDekel S, Ein-Dor T, Dishy GA, Mayopoulos PA (2020) Beyond postpartum depression: posttraumatic stress-depressive response following childbirth. Arch Womens Ment Health 23(4):557-564. https://doi.org/10.1007/s00737-019-01006-x\u003c/li\u003e\n\u003cli\u003eDu W, Qian X, Xu Z, Liu Z (2025) The role of anesthesiologists in postpartum depression: current perspectives and future directions. Front Psychiatry 16:1511817. https://doi.org/10.3389/fpsyt.2025.1511817\u003c/li\u003e\n\u003cli\u003eFroeliger A et al. (2024) Prevalence and risk factors for postpartum depression 2 months after a vaginal delivery: a prospective multicenter study. Am J Obstet Gynecol 230(3S):S1128-S1137.e6. https://doi.org/10.1016/j.ajog.2023.08.026\u003c/li\u003e\n\u003cli\u003eFroeliger A et al. (2025) Prevalence and risk factors for postpartum depression 2 months after cesarean delivery: a prospective multicenter study. Am J Obstet Gynecol 232(5):491.e1-491.e11. https://doi.org/10.1016/j.ajog.2024.10.031\u003c/li\u003e\n\u003cli\u003eFusar-Poli P, Hijazi Z, Stahl D, Steyerberg EW (2018) The Science of Prognosis in Psychiatry: A Review. JAMA Psychiatry 75(12):1289-1297. https://doi.org/10.1001/jamapsychiatry.2018.2530\u003c/li\u003e\n\u003cli\u003eGelaye B, Rondon MB, Araya R, Williams MA (2016) Epidemiology of maternal depression, risk factors, and child outcomes in low-income and middle-income countries. Lancet Psychiatry 3(10):973-982. https://doi.org/10.1016/S2215-0366(16)30284-X\u003c/li\u003e\n\u003cli\u003eHagatulah N, Br\u0026auml;nn E, Oberg AS, Valdimarsd\u0026oacute;ttir UA, Shen Q, Lu D (2024) Perinatal depression and risk of mortality: nationwide, register based study in Sweden. BMJ 384:e075462. https://doi.org/10.1136/bmj-2023-075462\u003c/li\u003e\n\u003cli\u003eHochman E et al. (2021) Development and validation of a machine learning-based postpartum depression prediction model: A nationwide cohort study. Depress Anxiety 38(4):400-411. https://doi.org/10.1002/da.23123\u003c/li\u003e\n\u003cli\u003eJohannsen BM et al. (2020) Self-harm in women with postpartum mental disorders. Psychol Med 50(9):1563-1569. https://doi.org/10.1017/S0033291719001661\u003c/li\u003e\n\u003cli\u003eJohannsen BM, Larsen JT, Laursen TM, Bergink V, Meltzer-Brody S, Munk-Olsen T (2016) All-Cause Mortality in Women With Severe Postpartum Psychiatric Disorders. Am J Psychiatry 173(6):635-642. https://doi.org/10.1176/appi.ajp.2015.14121510\u003c/li\u003e\n\u003cli\u003eKhalifeh H, Hunt IM, Appleby L, Howard LM (2016) Suicide in perinatal and non-perinatal women in contact with psychiatric services: 15 year findings from a UK national inquiry. Lancet Psychiatry 3(3):233-242. https://doi.org/10.1016/S2215-0366(16)00003-1\u003c/li\u003e\n\u003cli\u003eLim G, Farrell LM, Facco FL, Gold MS, Wasan AD (2018) Labor Analgesia as a Predictor for Reduced Postpartum Depression Scores: A Retrospective Observational Study. Anesth Analg 126(5):1598-1605. https://doi.org/10.1213/ANE.0000000000002720\u003c/li\u003e\n\u003cli\u003eLindahl V, Pearson JL, Colpe L (2005) Prevalence of suicidality during pregnancy and the postpartum. Arch Womens Ment Health 8(2):77-87. https://doi.org/10.1007/s00737-005-0080-1\u003c/li\u003e\n\u003cli\u003eLiu L, Xu DR, Tong Y, Shi J, Zeng Z, Gong W (2022) Symptomatology in 1,112 women screened positive and negative using the Edinburgh postnatal depression scale (EPDS): longitudinal observations from the first trimester to 6 weeks postpartum of a Chinese cohort. J Psychosom Obstet Gynaecol 43(4):453-463. https://doi.org/10.1080/0167482X.2022.2052845\u003c/li\u003e\n\u003cli\u003eNakamura A, van der Waerden J, Melchior M, Bolze C, El-Khoury F, Pryor L (2019) Physical activity during pregnancy and postpartum depression: Systematic review and meta-analysis. J Affect Disord 246:29-41. https://doi.org/10.1016/j.jad.2018.12.009\u003c/li\u003e\n\u003cli\u003eNorhayati MN, Hazlina NH, Asrenee AR, Emilin WM (2015) Magnitude and risk factors for postpartum symptoms: a literature review. J Affect Disord 175:34-52. https://doi.org/10.1016/j.jad.2014.12.041\u003c/li\u003e\n\u003cli\u003eOkun ML (2015) Sleep and postpartum depression. Curr Opin Psychiatry 28(6):490-496. https://doi.org/10.1097/YCO.0000000000000206\u003c/li\u003e\n\u003cli\u003ePayne JL et al. (2020) DNA methylation biomarkers prospectively predict both antenatal and postpartum depression. Psychiatry Res 285:112711. https://doi.org/10.1016/j.psychres.2019.112711\u003c/li\u003e\n\u003cli\u003ePebryatie E, Paek SC, Sherer P, Meemon N (2022) Associations Between Spousal Relationship, Husband Involvement, and Postpartum Depression Among Postpartum Mothers in West Java, Indonesia. J Prim Care Community Health 13:21501319221088355. https://doi.org/10.1177/21501319221088355\u003c/li\u003e\n\u003cli\u003eQi W et al. (2025) Prediction of postpartum depression in women: development and validation of multiple machine learning models. J Transl Med 23(1):291. https://doi.org/10.1186/s12967-025-06289-6\u003c/li\u003e\n\u003cli\u003eRogers A et al. (2020) Association Between Maternal Perinatal Depression and Anxiety and Child and Adolescent Development: A Meta-analysis. JAMA Pediatr 174(11):1082-1092. https://doi.org/10.1001/jamapediatrics.2020.2910\u003c/li\u003e\n\u003cli\u003eShovers SM, Bachman SS, Popek L, Turchi RM (2021) Maternal postpartum depression: risk factors, impacts, and interventions for the NICU and beyond. Curr Opin Pediatr 33(3):331-341. https://doi.org/10.1097/MOP.0000000000001011\u003c/li\u003e\n\u003cli\u003eSlomian J, Honvo G, Emonts P, Reginster JY, Bruy\u0026egrave;re O (2019) Consequences of maternal postpartum depression: A systematic review of maternal and infant outcomes. Womens Health (Lond) 15:1745506519844044. https://doi.org/10.1177/1745506519844044\u003c/li\u003e\n\u003cli\u003eS\u0026ouml;derquist J, Wijma B, Thorbert G, Wijma K (2009) Risk factors in pregnancy for post-traumatic stress and depression after childbirth. BJOG 116(5):672-680. https://doi.org/10.1111/j.1471-0528.2008.02083.x\u003c/li\u003e\n\u003cli\u003eTerada S et al. (2022) Relationship trajectories of pregnant women with their parents and postpartum depression: A hospital-based prospective cohort study in Japan. Front Psychiatry 13:961707. https://doi.org/10.3389/fpsyt.2022.961707\u003c/li\u003e\n\u003cli\u003eThorsteinsson EB, Loi NM, Farr K (2018) Changes in stigma and help-seeking in relation to postpartum depression: non-clinical parenting intervention sample. PeerJ 6:e5893. https://doi.org/10.7717/peerj.5893\u003c/li\u003e\n\u003cli\u003eTrost SL et al. (2021) Preventing Pregnancy-Related Mental Health Deaths: Insights From 14 US Maternal Mortality Review Committees, 2008-17. Health Aff (Millwood) 40(10):1551-1559. https://doi.org/10.1377/hlthaff.2021.00615\u003c/li\u003e\n\u003cli\u003eUS Preventive Services Task Force et al. (2019) Interventions to Prevent Perinatal Depression: US Preventive Services Task Force Recommendation Statement. JAMA 321(6):580-587. https://doi.org/10.1001/jama.2019.0007\u003c/li\u003e\n\u003cli\u003eWakefield C, Frasch MG (2023) Predicting Patients Requiring Treatment for Depression in the Postpartum Period Using Common Electronic Medical Record Data Available Antepartum. AJPM Focus 2(3):100100. https://doi.org/10.1016/j.focus.2023.100100\u003c/li\u003e\n\u003cli\u003eWaller R et al. (2022) Clinician-reported childbirth outcomes, patient-reported childbirth trauma, and risk for postpartum depression. Arch Womens Ment Health 25(5):985-993. https://doi.org/10.1007/s00737-022-01263-3\u003c/li\u003e\n\u003cli\u003eWang Z et al. (2021) Mapping global prevalence of depression among postpartum women. Transl Psychiatry 11(1):543. https://doi.org/10.1038/s41398-021-01663-6\u003c/li\u003e\n\u003cli\u003eXiao M et al. (2020) Risk prediction for postpartum depression based on random forest. Zhong Nan Da Xue Xue Bao Yi Xue Ban 45(10):1215-1222. https://doi.org/10.11817/j.issn.1672-7347.2020.190655\u003c/li\u003e\n\u003cli\u003eZhang R, Liu Y, Zhang Z, Luo R, Lv B (2025) Interpretable Machine Learning Model for Predicting Postpartum Depression: Retrospective Study. JMIR Med Inform 13:e58649. https://doi.org/10.2196/58649\u003c/li\u003e\n\u003cli\u003eZhang Y, Jin S (2016) The impact of social support on postpartum depression: The mediator role of self-efficacy. J Health Psychol 21(5):720-726. https://doi.org/10.1177/1359105314536454\u003c/li\u003e\n\u003cli\u003eZhang Y, Wang S, Hermann A, Joly R, Pathak J (2021) Development and validation of a machine learning algorithm for predicting the risk of postpartum depression among pregnant women. J Affect Disord 279:1-8. https://doi.org/10.1016/j.jad.2020.09.113\u003c/li\u003e\n\u003cli\u003eZhong M, Zhang H, Yu C, Jiang J, Duan X (2022) Application of machine learning in predicting the risk of postpartum depression: A systematic review. J Affect Disord 318:364-379. https://doi.org/10.1016/j.jad.2022.08.070\u003c/li\u003e\n\u003cli\u003eZhu Y, Xie Y, Yin X, Gong Y (2025) Feeding Patterns and Postpartum Depressive Symptoms: The Mediating Role of Parenting Self-Efficacy. Depress Anxiety 2025:2748707. https://doi.org/10.1155/da/2748707\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Postpartum depression, EPDS, Machine learning, Prediction model","lastPublishedDoi":"10.21203/rs.3.rs-7079798/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7079798/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose:\u003c/strong\u003e Postpartum Depression (PPD) is the most prevalent psychiatric disorder following childbirth, posing significant risks to both maternal and infant health. Consequently, prevention is paramount. This study aims to develop a clinically practical prediction tool using simple, readily accessible indicators to facilitate early intervention.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e In this multi-center study, data from 5,011 postpartum women were collected through structured questionnaires and electronic health records (EHRs).\u003cstrong\u003e \u003c/strong\u003ePPD was defined as an Edinburgh Postnatal Depression Scale (EPDS) score ≥10 at 6 weeks postpartum. Predictors within the dataset encompassed sociodemographic characteristics, pregnancy factors, delivery experiences, and infant care practices. Following feature selection, 11 machine learning(ML) models were constructed and evaluated. Model interpretability was enhanced using SHapley Additive exPlanations (SHAP).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e The prevalence of PPD was 13.3%. Nine core predictors were ultimately identified: support from spouse, parents, parents-in-law, siblings, and friends; fetal distress; birth experience rating; marital status; and marriage duration. Among the algorithms evaluated, the Random Forest model demonstrated relatively superior performance (Training set AUC: 0.725, 95% CI: 0.697-0.752; Validation set AUC: 0.612, 95% CI: 0.568-0.656). SHAP analysis identified optimizing the birth experience and strengthening the social support system as key, clinically actionable intervention targets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e This study confirms that a ML model based on simplified indicators provides moderate-performance risk stratification for PPD. This clinically practical tool equips frontline clinicians in resource-constrained settings in safeguarding vulnerable mothers through proactive early-warning systems during their critical transition period, thus reducing severe PPD and preventing devastating consequences.\u003c/p\u003e","manuscriptTitle":"A Clinically Practical Postpartum Depression Predictor: Machine Learning Model Based on Simplified Indicators","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-01 10:42:34","doi":"10.21203/rs.3.rs-7079798/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"af9261ef-2bf5-4a6c-ad9c-20f9450eafdb","owner":[],"postedDate":"August 1st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-11-19T05:54:55+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-01 10:42:34","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7079798","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7079798","identity":"rs-7079798","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}