Analysis of influencing factors of postpartum depression in patients with gestational diabetes mellitus and construction of prediction model

Abstract

BACKGROUND

Gestational diabetes mellitus (GDM) requires strict dietary management and blood glucose monitoring, which may impose long-term psychological stress during pregnancy and increase the risk of postpartum depression (PPD). However, the psychological impact of gestational glucose indicators and blood glucose control, as well as their predictive value for PPD in patients with GDM, remains insufficiently explored.

AIM

To identify factors associated with PPD in patients with GDM and to construct a prediction model for PPD risk.

METHODS

A cross-sectional survey was conducted among 204 patients with GDM who underwent prenatal checkups and delivered at Suzhou Ninth People’s Hospital between February 2024 and June 2025. At 6 weeks postpartum, PPD symptoms were measured using the Edinburgh PPD Scale, and participants were divided into PPD (52 cases) and non-PPD (152 cases) groups. Group differences were analyzed, and multivariate logistic regression was used to identify factors associated with PPD in patients with GDM. A predictive model was constructed based on various influencing factors and evaluated using goodness-of-fit testing and the area under the receiver operating characteristic curve. Model performance was further validated using K-fold fold cross-validation.

RESULTS

Significant differences between the PPD and non-PPD groups were observed for 2-hour postprandial blood glucose (2hPG) at GDM diagnosis (P = 0.018), blood glucose control during pregnancy (P = 0.012), postpartum maternal-infant separation (P = 0.001), family care (P = 0.001), and social support (P = 0.007). Multivariate analysis identified high 2hPG, poor gestational blood glucose control, postpartum mother-infant separation, low family care, and low social support as independent risk factors for PPD in patients with GDM (all P < 0.05). The predictive model was defined as Logit (P) = 0.508 × 2hPG + 0.687 × gestational blood glucose control + 1.092 × postpartum mother-infant separation + 0.745 × low family care + 0.289 × low social support - 4.766. The goodness-of-fit test showed no evidence of overfitting (χ² = 1.754, P = 0.514). The model’s area under the receiver operating characteristic curve value was 0.840 (95% confidence interval: 0.757-0.912), with a sensitivity of 0.839 and a specificity of 0.825. After 100 rounds of 10-fold cross-validation, the model demonstrated good generalization performance.

CONCLUSION

PPD incidence is high in patients with GDM and is associated with high 2hPG at diagnosis, poor blood glucose control during pregnancy, postpartum mother-infant separation, low family care, and low social support. A predictive model integrating these factors can effectively evaluate PPD risk in patients with GDM.

Key Words: Gestational diabetes; Postpartum depression; Influencing factors; Predictive model; Postprandial blood glucose

Core Tip: Gestational diabetes mellitus adversely affects maternal and neonatal health and predisposes affected women to anxiety, depression, and other negative emotions. This study identifies key factors associated with postpartum depression in patients with gestational diabetes mellitus and constructs a predictive model to support clinical assessment and targeted intervention for reducing postpartum depression risk.

INTRODUCTION

Gestational diabetes mellitus (GDM) refers to diabetes characterized by normal glucose metabolism before pregnancy and abnormal glucose metabolism during pregnancy. According to the diagnostic criteria recommended by the International Association of Diabetes and Pregnancy Research Groups[1], the prevalence of GMD in China has reached 14.8%[2]. GDM not only compromises maternal and neonatal health but is also associated with increased risk postpartum depression (PPD)[3,4].

PPD refers to an emotional disorder after birth characterized by low mood, decreased interest, and lack of pleasure after childbirth, significantly affecting maternal physical and mental health and infant growth and development[5]. GDM itself is a contributing factor to PPD[6]. This may be because GDM patients must strictly manage their diet and control their blood sugar during pregnancy and after delivery, which imposes psychological stress and increases the risk of PPD. Therefore, greater attention to the PPD of women with GDM is required to enable early identification and intervention.

Previous studies have suggested that educational level, parity, and childbirth-related knowledge are independent risk factors for PPD among high-risk pregnant women[7,8]. However, for postpartum women diagnosed with GDM, these factors cannot guide the development of measures related to the management of PPD in patients with GDM. Therefore, this study aims to identify factors influencing PPD in patients with GDM and to construct a predictive model for PPD, thereby providing a clinical reference for obstetricians to evaluate PPD risk and formulate management measures.

MATERIALS AND METHODS

Research objects

A cross-sectional survey design was used, as patients with GDM who underwent prenatal checkups and delivered at Suzhou Ninth People’s Hospital between February 2024 and June 2025 were selected as the research subjects. At six weeks postpartum, participant’s PPD symptoms were assessed using the Edinburgh PPD Scale (EPDS) during routine obstetric follow-up, and were therefore divided into PPD and non-PPD groups. The EPDS scale consists of 10 items assessing fear, insomnia, coping ability, sadness, mood, pleasure, anxiety, self-blame, crying, and self-injury. Each item is scored using a four-point Likert scale (0-3), yielding a total score of 0-30. An EPDS score ≥ 9 at 6 weeks postpartum was used to indicate the presence of PPD symptoms[9].

Sample size estimation was performed using a rough calculation method. Assuming five independent variables in the predictive model and a reported PPD incidence of approximately 20% in women with GDM[10], with a two-sided α = 0.05 and β = 0.2, the required sample size was calculated as 123 cases. Accounting for an estimated 20% invalid questionnaire rate, a minimum of 148 participants were required. Ultimately, 204 cases were included, meeting the requirements for statistical analysis.

Inclusion criteria: (1) Diagnosis of GDM during glucose tolerance screening at 24-28 weeks of pregnancy using a 75-g oral glucose tolerance test (OGTT), with GDM diagnosed when any of the following criteria were met: Fasting blood glucose ≥ 5.1 mmol/L, 1-hour postprandial blood glucose ≥ 10.0 mmol/L, or 2-hour postprandial blood glucose ≥ 8.5 mmol/L[11]; (2) Age ≥ 18 years old; and (3) Singleton pregnancy with full-term delivery.

Exclusion criteria included: (1) Conception via assisted reproductive technology; (2) Diabetes was diagnosed before pregnancy; (3) Comorbid malignant tumors; (4) Diagnosis of depression before delivery; (5) Severe obstetric complications, including uterine rupture, heavy hemorrhage, amniotic fluid embolism, or postpartum shock; (6) Presence of pregnancy complications other than GDM; and (7) Inability to read and understand the survey questionnaire.

Research methods

Family care: Family functioning among patients with GDM was assessed using the Family Care Index Questionnaire[12]. This questionnaire comprises five items assessing fitness, cooperation, intimacy, growth, and emotion. Each item was scored using a 3-point Likert scale (0-2), yielding a total score of 0-10. Scores of 7-10 indicate high family care, 4-6 indicate moderate family care, and 0-3 indicate low family care.

Social support level: Perceived social support was evaluated using the Multidimensional Perceived Social Support Scale[13]. The scale includes 12 items across three dimensions: Family support (4 items), friend support (4 items), and other support (4 items). Each item is rated using a 7-point Likert scale (1-7), yielding a total score of 12-84, with higher scores indicating greater perceived social support. Scores ≥ 61 indicate high social support, scores of 37-60 indicate moderate social support, and ≤ 36 indicate low social support.

General information: A structured general information questionnaire was used to collect data on age, educational level, place of residence, reproductive history, pre-pregnancy body mass index, blood glucose indices at GDM diagnosis, including fasting plasma glucose (FPG), 1-hour postprandial OGTT glucose (1hPG), 2-hour postprandial OGTT glucose (2hPG), and fasting insulin, gestational blood glucose control methods, gestational blood glucose control status. Gestational blood glucose control was defined as good when FPG values at each prenatal check-up were < 5.1 mmol/L, 1hPG < 10.0 mmol/L, and 2hPG < 8.5 mmol/L; otherwise, it indicated poor control. Additional data included delivery mode, newborn sex, postpartum mother-infant separation (defined as neonatal admission to the neonatology department for treatment after birth), and neonatal feeding method.

Statistical analysis

Data were analyzed using SPSS version 25.0. Measurement data was first tested for normality using the Shapiro-Wilk test and are presented as mean ± SD. Between-group comparisons were performed using independent t-tests. Categorical variables are expressed as counts and percentages n (%) and were compared using the χ² test. Factors influencing PPD in patients with GDM were analyzed using multivariate logistic regression analysis. A predictive model was constructed based on the regression coefficients (β) and constant term of the identified influencing factors. Model fit was evaluated using the goodness-of-fit test, and predictive performance was evaluated using the area under the receiver operating characteristic curve. N-fold cross-validation was applied to further validate model stability and generalizability. A two-sided P value < 0.05 indicated statistically significant difference.

RESULTS

Incidence of PPD in patients with GDM and univariate analysis

A total of 204 patients with GDM were reexamined at the obstetrics clinic 6 weeks postpartum delivery. Based on the EPDS evaluation, 52 patients exhibited depressive symptoms and were classified into the PPD group, yielding an incidence rate of 25.49% (52/204), whereas 152 patients were classified into the non-PPD group. Univariate analysis showed that the 2hPG level was significantly higher in the PPD group than in the non-PPD group (t = 2.385, P = 0.018). The proportions of patients with poor gestational blood glucose control (χ² = 6.285, P = 0.012), postpartum mother-infant separation (χ² = 11.386, P = 0.001), low family care (χ² = 13.539, P = 0.001), and low social support (χ² = 10.057, P = 0.007) were also significantly higher in the PPD group than in the non-PPD group. No significant differences were observed between the two groups in age, education level, place of residence, fertility history, pre-pregnancy body mass index, FPG, 1hPG, fasting insulin, mode of delivery, neonatal sex, and neonatal feeding pattern (all P > 0.05; Table 1).

Table 1 Comparison of data between postpartum depression group and no-postpartum depression group, mean ± SD/n (%).

Project	PPD group (n = 52)	No-PPD group (n = 152)	t/χ2 value	P value
Age (years)	29.25 ± 4.23	28.75 ± 4.18	0.742	0.459
Educational level			2.346	0.126
High school and below	30 (57.69)	69 (45.39)
College degree or above	22 (42.31)	83 (54.61)
Place of residence			0.641	0.726
Countryside	14 (26.92)	34 (22.37)
Township	18 (34.62)	51 (33.55)
City	20 (38.46)	67 (44.08)
Birth history			0.292	0.589
Primipara	19 (36.54)	62 (40.79)
Multipara	33 (63.46)	90 (59.21)
Pre-pregnancy body mass index (kg/m2)	23.27 ± 2.73	22.95 ± 2.89	0.699	0.485
FPG (mmol/L)	5.84 ± 1.28	5.62 ± 1.13	1.171	0.243
1hPG (mmol/L)	9.89 ± 2.27	9.55 ± 2.27	0.910	0.364
2hPG (mmol/L)	8.68 ± 0.54	8.49 ± 0.48	2.385	0.018
FINS (mmol/L)	15.07 ± 2.31	14.68 ± 2.25	1.072	0.285
Blood glucose control methods			1.287	0.257
Subcutaneous injection of insulin	13 (25.00)	27 (17.76)
Diet and exercise therapy	39 (75.00)	125 (82.24)
Blood glucose control during pregnancy			6.285	0.012
Good	20 (38.46)	89 (58.55)
Poor	32 (61.54)	63 (41.45)
Type of delivery			1.853	0.173
Caesarean section	24 (46.15)	54 (35.53)
Vaginal delivery	28 (53.85)	98 (64.47)
Neonatal sex			0.224	0.636
Baby boy	29 (55.77)	79 (51.97)
Baby girl	23 (44.23)	73 (48.03)
Postpartum mother infant separation			11.386	0.001
Yes	14 (26.92)	13 (8.55)
No	38 (73.08)	139 (91.45)
Feeding methods for newborns			1.440	0.487
Artificial feeding	8 (15.38)	17 (11.18)
Mixed feeding	19 (36.54)	48 (31.58)
Exclusive breast feeding	25 (48.08)	87 (57.24)
Family care level			13.539	0.001
High	7 (13.46)	49 (32.24)
Medium	21 (40.39)	70 (46.05)
Low	24 (46.15)	33 (21.71)
Social support			10.057	0.007
High	9 (17.31)	53 (34.87)
Medium	22 (42.31)	68 (44.74)
Low	21 (40.38)	31 (20.39)

FPG: Fasting plasma glucose; 1hPG: 1-hour postprandial oral glucose tolerance test glucose; 2hPG: 2-hour postprandial oral glucose tolerance test glucose; FINS: Fasting insulin; PPD: Postpartum depression.

Multivariate logistic regression analysis of PPD in patients with GDM

The presence of PPD at 6 weeks postpartum (0 = no, 1 = yes) was defined as the dependent variable. The statistically significant items in the univariate analysis, including 2hPG, gestational blood glucose control, postpartum mother-infant separation, family care, and social support, were used as independent variables, with assignments shown in Table 2. Multivariate logistic regression analysis showed that high 2hPG levels, poor blood glucose control during pregnancy, postpartum mother-infant separation, low family care, and low social support were all independent risk factors for PPD in patients with GDM (P < 0.05; Table 3).

Table 2 Each variable assignment description.

Variable	Description of valuation
2hPG	Enter the actual value
Blood glucose control during pregnancy	0 = good; 1 = poor
Postpartum mother infant separation	0 = no; 1 = yes
Family care level	0 = high; 1 = medium; 2 = low
Social support	0 = high; 1 = medium; 2 = low

2hPG: 2-hour postprandial oral glucose tolerance test glucose.

Table 3 Results of multivariate logistic analysis.

Variable	β	SE	χ2	P value	OR (95%CI)
2hPG	0.508	0.251	4.096	0.043	1.662 (1.016-2.718)
Poor blood glucose control during pregnancy	0.687	0.288	5.690	0.017	1.987 (1.131-3.494)
Postpartum mother infant separation	1.092	0.302	13.075	< 0.001	2.980 (1.649-5.387)
Family care level
Medium	0.252	0.172	2.147	0.143	1.287 (0.919-1.802)
Low	0.745	0.293	6.465	0.011	2.106 (1.186-3.739)
Social support
Medium	0.289	0.168	2.959	0.085	1.335 (0.961-1.855)
Low	0.493	0.206	5.727	0.017	1.637 (1.093-2.452)
Constant term	-4.766	1.079	19.510	< 0.001

2hPG: 2-hour postprandial oral glucose tolerance test glucose; OR: Odds ratio; CI: Confidence interval.

Construction of a PPD risk prediction model for patients with GDM

A predictive model for PPD risk in patients with GDM was constructed based on the regression coefficients and constant term derived from the multivariate logistic regression analysis (Table 3). The model equation was Logit (P) = 0.508 × 2hPG (actual value) + 0.687 × gestational blood glucose control (0 = good, 1 = poor) + 1.092 × postpartum mother-infant separation (0 = no, 1 = yes) + 0.745 × low family care (0 = no; 1 = yes) + 0.289 × low social support (0 = no, 1 = yes) - 4.766. The goodness-of-fit test indicated no evidence of model overfitting (χ² = 1.754, P = 0.514). The AUC value was 0.840 (95% confidence interval: 0.757-0.912), with a sensitivity of 0.839 and a specificity of 0.825, indicating good predictive performance (Figure 1).

Figure 1 Receiver operating characteristic curve evaluation of postpartum depression risk prediction model in gestational diabetes mellitus patients. AUC: Area under the curve; CI: Confidence interval.

Deep validation of the PPD risk prediction model for patients with GDM

Validation indicators are shown in Figure 2. Model performance using the AUC/C-index (values > 0.7 indicating excellent discrimination), the coefficient of determination (R²; higher values indicating better fit), the discrimination index (D ≥ 0.4 indicating excellent discrimination), and calibration performance assessed using the U-test (P > 0.05 indicating better predictive calibration). The calibration metrics Emax and Eavg represent the maximum and average offset between the prediction and ideal models, respectively, with smaller values indicating closer alignment between both models. Using these evaluation criteria, K-fold cross-validation was conducted with 10 folds and repeated 100 times to perform a deep validation of the multifactor logistic regression model. The results showed that all values evaluated in the original data model were good and had small changes, indicating that the constructed predictive model had good diagnostic performance and strong generalization ability (Table 4).

Figure 2 Extraction of deep validation indicators for regression model modeling based on raw data. ROC: Receiver operating characteristic.

Table 4 Cross validation depth evaluation results of prediction models.

Indicator	Modeling data values	Correction value	Change value (Δ)
AUC/C-index	0.840	0.847	0.007
R2	0.638	0.646	0.008
Differentiation index (D)	0.505	0.487	-0.018
U inspection	-0.007	0.019	0.012
Brier score	0.088	0.124	0.036
Maximum offset Emax	0.073	0.235	0.162
Minimum offset Eavg	0.016	0.142	0.126

AUC: Area under the curve.

DISCUSSION

PPD endangers maternal physical and mental health and adversely affects early infant development, families and society. A meta-analysis[14] showed that GDM increases the risk of PPD, and Saeed Alqahtani et al[15] reported that the incidence of PPD in patients diagnosed with GDM is 6.5%-48.4%. This study’s results showed that the incidence of PPD in patients with GDM was 25.49%, which falls within the scope in the literature above. Identifying factors influencing PPD in patients with GDM is therefore important for the development of preventive strategies.

Under normal conditions following oral glucose tolerance screening during pregnancy, blood glucose levels peak at 1 hour and gradually return to baseline. When blood glucose levels remain high at 2 hours (i.e., elevated 2hPG), it indicates delayed insulin secretion peak, impaired pancreatic β-cell function, and insulin resistance, which is not conducive for postpartum glucose metabolism and increases the risk of PPD[16]. Consequently, many patients with GDM get worried about possible PPD, which increases psychological burden and may increase the risk of PPD[17]. Mak et al[18] reported that high 2-hour postprandial glucose levels during gestational glucose tolerance testing were significantly associated with higher EPDS scores within three months postpartum, supporting our research findings. The reason may be because hyperglycemia leads to excessive reactive oxygen species, which damage cellular membranes, proteins, and DNA and cause inflammatory responses, such as increased expression of inflammatory factors, including tumor necrosis factor-α and interleukin-6. Overexpression of inflammatory factors can cross the blood-brain barrier and affect neurotransmitter metabolism, including serotonin and dopamine pathways, leading to depressive symptoms. Additionally, insulin resistance associated with hyperglycemia may disrupt normal hypothalamic-pituitary-adrenal axis function, leading to abnormal cortisol secretion and mood disturbances in pregnant women[19]. This can result in excessive accumulation of negative emotions and stress in patients with GDM, which cannot resolve immediately after delivery and may continue into postpartum, becoming a “psychological burden” and increasing the risk of PPD. This may be a reason why patients with GDM are prone to depressive symptoms after childbirth. Therefore, effective blood glucose control during pregnancy is crucial for patients with GDM. Previous studies have shown that adverse pregnancy outcomes are associated with an increased risk of PPD[20]. Kodama et al[21] reported that improved glucose control during pregnancy reduces the risk of adverse outcomes, including cesarean section, small-for-gestational-age infants, and premature rupture of membranes, in patients with GDM. Minschart et al[22] further demonstrated a negative correlation the incidence of PPD in patients with GDM and the level of blood glucose control during pregnancy, with poorer blood glucose control associated with a higher risk of PPD. These findings are consistent with our study’s results, which identified poor blood glucose control during pregnancy as a risk factor for PPD in patients with GDM. Sustained hyperglycemia may lead to increased oxidative stress, inflammatory response, and insulin resistance, resulting in persistent metabolic dysregulation after delivery. Such disturbances may impair hypothalamic-pituitary-adrenal axis function and emotional regulation in patients with GDM after delivery, thereby increasing the risk of PPD.

Patients with GDM often experience heightened concern regarding their own health and that of their newborns. Postpartum mother-infant separation may further exacerbate the postpartum stress responses and increase the risk of PPD[23]. The present study found that postpartum mother-infant separation is a significant risk factor for PPD in patients with GDM. It may be because when pregnant women are diagnosed with GDM during pregnancy, they have concerns regarding fetal health, thereby increasing psychologically stress and emotional vulnerability affect delivery[24]. When newborns have comorbid diseases, they require transfer to neonatal units for monitoring or treatment, the resulting mother-infant separation may further exacerbate psychological stress, ultimately increasing the risk of PPD.

Family care reflects an individual’s satisfaction with family functioning and plays a critical role in postpartum recovery. Pebryatie et al[25] reported that a high level of spousal care during pregnancy, childbirth, and the postpartum period reduces the risk of PPD. Wan et al[26] further showed that low spousal participation before and after delivery, as well as low care from parents-in-law during postpartum, significantly increases PPD prevalence. These findings indirectly support this study’s findings that low family care is a risk factor for PPD in patients with GDM. Insufficient family care may reduce maternal satisfaction with family functioning; when faced with postpartum physiological changes and concerns regarding maternal and infant health, negative emotions are more likely to occur, thereby aggravating maternal psychological stress and increasing PPD risk. Social support refers to the material and spiritual assistance obtained through social networks and is essential for physical and mental health. Heh et al[27] found that higher levels of social support are associated with a lower risk of PPD, while Delavari et al[28] reported a significant negative correlation between social support and depression, supporting this study’s results. The reason may be that pregnant women with insufficient social support may lack access to relevant GDM-related knowledge from friends and relatives, increasing their concerns about fetal health and uncertainty about the future, thereby increasing the risk of PPD. Chen et al[29] found that enhancing women’s social support can reduce their depressive symptoms. Therefore, strengthening health education and psychological counseling for GDM patients with low social support, and encouraging participation in postpartum breastfeeding and maternal support groups (e.g., WeChat groups) may help increase their social support and reduce PPD risk.

There are many influencing factors of PPD in GDM patients, making it extremely important to construct an effective risk prediction model. In this study, multivariate logistic regression analysis was used to identify factors influencing PPD in patients with GDM, and a predictive model was established based on the regression coefficients and constant terms of each variable. Receiver operating characteristic curve analysis showed an AUC value of 0.840, and goodness-of-fit testing indicated no overfitting, indicating that the model has satisfactory predictive ability. Further validation using cross-validation demonstrated good diagnostic performance and generalization ability. Therefore, the model constructed based on identified risk factors may assist clinicians in assessing PPD risk in high-risk patients with GDM and in developing personalized management strategies.

This study has several limitations. Although the incidence and influencing factors of PPD in women with GDM were analyzed and a predictive model was constructed, this study was conducted in a single center with limited sample size, which may limit representativeness. Therefore, future studies should expand the sample size and include multicenter data to improve generalizability.

CONCLUSION

The risk of PPD in patients with GDM is associated with high 2hPG during pregnancy, poor blood glucose control during pregnancy, postpartum mother-infant separation, low family care, and low social support. Based on these factors, the prediction model can effectively evaluate PPD risk in patients with GDM and provides clinical guidance for implementing measures to reduce the PPD risk in patients with GDM.