Effects of sleep quality, anxiety, and depression on miscarriage among pregnant women: a multi-center prospective observational study

Abstract

This study aims to uncover the mechanisms and quantitative dose response relationships among sleep quality, anxiety, depression and miscarriage, as well as develop a comprehensive predictive model for the miscarriage rate. In this study, 1058 pregnant women in mainland China were recruited. We utilized both univariate, multivariate analyses and sensitivity analysis to investigate the relationship between sleep quality, anxiety, depression, and miscarriage. Then, we used mediation analysis and directed acyclic graph to explore how anxiety and sleep quality mediate the relationship between depression and miscarriage. We employed restricted cubic spline (RCS) to examine the dose-response relationship between these variables and constructed a nomogram model for predicting the occurrence of miscarriages. During our investigation, 16.4% of the participant had a miscarriage. Our results showed a significant association between sleep quality, anxiety, depression and miscarriage both unadjusted and multivariable multinomial logistic regression. Dose-response relationships showed that the miscarriage rate slowly increases with increasing PSQI, SAS and SDS scores at first. However, when a certain threshold is reached, even slight increases in the scores will lead to a sharp rise in the miscarriage rate. Anxiety mediated the effect of depression on miscarriage by 44% and sleep quality had a similar mediation effect (16%). The quantitative dose response relationships between PSQI, SAS, SDS, and the miscarriage rate are all positive. In the impact of depression on the miscarriage rate, anxiety and sleep quality also play significant mediating roles. By revealing high-risk pregnant women, early intervention can be provided, aiming to reduce the miscarriage rate.

Introduction

Miscarriage is generally defined as the loss of a pregnancy before viability. In recent years, with the rapid development of society, increasing life pressures, worsening environmental pollution, and the rise in the number of advanced maternal age pregnancies, the incidence of miscarriage has been on the rise. The pooled risk of miscarriage is 15.3% (95% CI 12.5–18.7%) of all recognised pregnancies [1].

Sleep is a fundamental physiological need for humans, which helps alleviate fatigue and restore energy. Sufficient sleep aids in the clearance of neurotoxic waste accumulated during wakefulness [2]. Sleep deprivation often leads to fatigue, lack of concentration, delayed reaction time, and impaired judgment [3]. Pregnant and postpartum women, as a special population, often experience significant changes in reproductive hormone levels such as estrogen and progesterone during the perinatal period. Physiological changes, such as increased frequency of urination and difficulty finding a comfortable sleeping position, as well as psychological turmoil, can lead to poor sleep quality, anxiety, and depression [4]. On average, approximately 45% of pregnant women experience sleep disturbances [4]. Poor sleep quality or sleep disorders are associated with an increased risk of adverse outcomes during pregnancy, such as hypertensive disorders of pregnancy (HDP), gestational diabetes, fetal growth restriction, and preterm birth [4]. The relationship between sleep quality and miscarriage has also been established [5, 6]. Researchers have found through prospective studies that poor sleep quality is associated with a higher risk of miscarriage [5, 6].

Studies have shown that poor sleep quality is often accompanied by feelings of anxiety and depression in patients. Complaints of poor sleep are reported by up to 90% of people with diagnosed depression [7] and up to 70% of people with anxiety [8]. Depression is the leading cause of disease-related disability in women. Epidemiological studies have shown that the lifetime prevalence of a major depressive disorder in women (21.3%) is almost twice that in men (12.7%) [9]. Therefore, the impact of postpartum depression on miscarriage is also a topic of significant interest among researchers. Liang et al. conducted a longitudinal cohort study, which revealed that a history of depression is associated with a higher risk of subsequent infertility, miscarriage, and recurrent miscarriages [10]. Research indicates that the occurrence of depression in pregnant women may be associated with miscarriage, and depression may lead to symptoms of poor sleep quality and anxiety in patients [11]. Since depression requires a professional diagnosis by a mental health expert, even though it is known that depressed pregnant women may be at a higher risk of miscarriage, it is difficult to implement preventive interventions due to the lack of a simple method to assess a pregnant woman’s depression status. In contrast, a patient’s poor sleep quality and anxiety levels are easier to determine, as individuals can self-assess whether they are experiencing sleep deprivation or anxiety. Therefore, we utilized mediation analysis to explore whether anxiety and sleep quality mediate the relationship between depression and miscarriage, thereby enabling the reduction of the impact of depression on miscarriage by controlling these mediating variables. This research has significant clinical implications.

Through a systematic review of the existing literature, our study hypothesizes that poor sleep quality, anxiety, and depression in pregnant women may influence the risk of miscarriage. We employed a prospective design to investigate the possible relationships, utilizing both univariate and multivariate models. Additionally, sensitivity analyses were conducted to verify the stability and reliability of the results.

Previous models for predicting miscarriage in pregnant women primarily focused on clinical blood markers, lifestyle factors, age, and previous history of miscarriage [12, 13]. However, the scope for altering these variables is limited, thereby restricting the potential for preventive measures against miscarriage. In contrast, sleep quality, anxiety, and depression have the characteristic of being early preventable, allowing for intervention before the onset of miscarriage symptoms, thereby enabling earlier prevention of miscarriages. Therefore, our study constructs a nomogram model for predicting the early occurrence of unexpected miscarriages in pregnant and postpartum women. This model can serve to more effectively alert pregnant and postpartum women, emphasizing the importance of improving sleep quality and reducing anxiety and depression. By doing so, it facilitates the early prevention of miscarriages and reduces the miscarriage rate.

Methods

Study design

This was a prospective cohort study conducted in the three Subsidiary Hospital of First Affiliated Hospital, Zhejiang University School of Medicine and Affiliated Jinhua Hospital Zhejiang University School of Medicine. Baseline recruitment was conducted from May 2022 to May 2023, and pregnant women who visited the outpatient clinic for the first prenatal examination at First Affiliated Hospital, Zhejiang University School of Medicine and Affiliated Jinhua Hospital Zhejiang University School of Medicine were recruited when they met the following inclusion criteria: <14 gestational weeks, singleton pregnancy, resided in Hangzhou or Jinhua during the past half year and have no plan to move out after delivery. The baseline survey includes basic information such as age, height, and weight, as well as questionnaire surveys on sleep quality, anxiety and depression. Exclusion criteria: Individuals with intellectual impairments who are unable to comprehend the questionnaire and provide accurate answers; Individuals who cannot be followed up with clinical examination results and delivery outcomes.

Follow-up for pregnancy outcomes

The follow-up of the pregnancy outcomes was carried out by local healthcare personnel. Pregnant women receive regular antenatal care and give birth in the hospital. Information on pregnancy outcomes is obtained through the hospital’s medical electronic information system, which automatically records information during each antenatal care and delivery. Miscarriage is defined as the loss of pregnancy that occurs before reaching 20 completed weeks of gestation [14]. The incidence rate of miscarriage was defined as the proportion of participants who had a miscarriage to all participants.

Covariates

Covariates were collected at the first prenatal visit, including age, educational level, occupation, history of cesarean section, history of preterm birth, history of miscarriage, history of pregnancy, history of operation, gravidity, number of children, pre-pregnancy weight, basic disease, bad habits (Smoking or drinking during pregnancy), anxiety, depression, sleep quality and so on. Pre-pregnancy BMI was calculated using weight (in kilograms) divided by the square of height (in meters). Inflammatory markers such as White Blood Cell count (WBC), Neutrophil Percentage (NEU%), NEU (Neutrophil Count), Platelet Count (PLT), Albumin (ALB) and Albumin (ABL) were obtained from the hospital information system.

Anxiety symptoms was assessed at baseline using the Zung Self-Rating Anxiety Scale developed by Zung [15]. It consists of 20 questions. Responses are scored on a four-point scale, ranging from 1 (no or very little time) to 2 (sometimes), to 3 (most of the time) and 4 (most or all of the time). The raw score is the sum of all responses. The standard score is calculated as 1.25 times the original score. The current study used an index score cutoff of ≥ 50 to diagnose anxiety.

Depression symptoms was assessed at baseline using the Zung Self-Rating Depression Scale developed by Zung [16]. It consists of 20 questions. Responses are scored on a four-point scale, ranging from 1 (no or very little time) to 2 (sometimes), to 3 (most of the time) and 4 (most or all of the time). The raw score is the sum of all responses. The standard score is calculated as 1.25 times the original score. The current study used an index score cutoff of ≥ 53 to diagnose depression.

Sleep quality was assessed at baseline using the Pittsburgh Sleep Quality Index, a self-report scale used to assess respondents’ sleep quality over a 2-week period [17]. PSQI consists mainly of 18 items with 7 components, which represent sleep quality, time to fall asleep, sleep duration, sleep efficiency, sleep disorders, hypnotic drugs, and daytime dysfunction. Each question is rated on a scale of 0–3. The sum of the 7 component scores is the total score of PSQI. The higher the score, the worse the sleep. The total score ranged from 0–21, with a score of > 5 indicating poor sleep quality (sleep disturbance) in pregnant women, with 98% high sensitivity and 90% specificity (kappa = 0.89, p < 0.01), suitable for Chinese [17].

We investigated whether participants had hypertension, diabetes, liver and kidney disease, heart disease, tumor and other diseases. Since most pregnant women did not have these diseases, we collectively referred to these diseases as “basic diseases” for subsequent analysis. We surveyed the smoking history and alcohol consumption history of pregnant women. Due to the relatively low number and proportion of individuals who smoked and drank alcohol, we have decided to consolidate these two variables into a category called “bad habits”.

Sample size estimation

The primary objective of this study was to investigate the relationship between three distinct primary factors (sleep quality, depression, and anxiety) and maternal miscarriage using multiple logistic regression, with sleep quality being the main variable of interest. Therefore, we employed the sample size calculation formula specifically designed for logistic regression [18, 19], The sample size calculation formula for logistic regression when there is only one covariate is as follows:

$$\mathrm{N}=\frac{{\left(Z_{1-\frac{\alpha }{2}}\sqrt{\frac{\stackrel{\circledR }{P}(1-\stackrel{\circledR }{P})}{R}}+Z_{1-\beta }\sqrt{P_0\left(1-P_0\right)\frac{P_1(1-P_1)(1-R)}{R}}\right)}^2}{{\left(P_0-P_1\right)}^2(1-R)}$$ 1

The calculation formula for multiple covariates in logistic regression is as follows:

$$N_m=\frac{N}{1-{\rho }^2}$$ 2

In formula (1), $\stackrel{\circledR }{P}$ represents the overall event rate, which in this study refers to the miscarriage rate among pregnant women. Based on relevant literature [1], this rate is set at 15.3%. $P_0$ is the event rate at X₁ = 0 and $P_1$ is the event rate at X₁ = 1. Percent with X = 1 is the percentage of the sample in which X = 1, which is often called the prevalence of X. In this study, P₀ is the miscarriage rate when sleep quality is good, and P₁ is the miscarriage rate when sleep quality is poor. According to the literature [20, 21], the miscarriage rate for pregnant women with poor sleep quality is 1.8 times that of pregnant women with good sleep quality. R is the proportion of the sample with X₁ = 1, and $\stackrel{\circledR }{P}=\left(1-\mathrm{R}\right)P_0+\mathrm{R}\left(P_1\right)$. In this study, R represents the miscarriage rate when sleep quality is poor. Based on the literature [21, 22], this rate ranges from 20 to 30%, and in this study, we set it at 20%. In formula (2), ρ represents the correlation between other independent variables and the variable of interest (sleep quality). Based on relevant literature [7], we assumed ρ is 0.4. α denotes the significance level, which is set at 0.05 in this study. The term 1-β represents the study’s power, set at 0.2. Using the PASS software, the calculated minimum sample size required for multiple logistic regression is 504 participants. Considering a potential 20% loss to follow-up during the study, the final sample size to be included should be at least 630 participants.

Statistical analysis

Mean (SD) values and proportions of baseline characteristics were calculated. Incidence rate (95%CI) of miscarriage was calculated. Incidence rate of miscarriage in women with different characteristics were also compared using chi-square test or Wilcoxon rank sum test. Multivariable logistic regression models were used to estimate the adjusted odds ratios (ORs) and their 95% CIs of miscarriage for women with different exposures. We combined logistic model and RCS methods to describe the stoichiometric response relationship with anxiety, depression, sleep quality and miscarriage (RCS details seen appendix pp1). We used nomogram model to predict the probability of miscarriage (nomogram details seen appendix pp1&pp2). Mediation was evaluated with linear regression with the product of coefficients method to estimate the direct and indirect effects of the relationship (mediation analysis details seen appendix pp2&pp3) and draw their directed acyclic graph (appendix pp 9), done with the R package lavaan version 0.6–12 (appendix pp2&pp3). Multiple imputation using random forest was employed to fill in the missing information (details of multiple imputation with random forests seen appendix pp3). All data were analysed with R (version 4.2.0). A p value less than 0.05 was considered significant.

Sensitivity analysis

In order to further confirm the relationship between sleep quality, anxiety, and depression with miscarriage and validate the stability of the main analytical results, we conducted the following sensitivity analyses: From a clinical perspective, understanding the gestational week in which a miscarriage occurs can contribute to a deeper understanding of its etiology and may inform treatment and prevention strategies. Therefore, incorporating gestational week as a time variable in the analysis is crucial, as it provides valuable insights into the timing of miscarriage and enables a more precise assessment of the influence of various factors on miscarriage risk. A Cox regression analysis was conducted with gestational age as the time variable and miscarriage as the outcome variable. The gestational age information was determined through ultrasound examinations. The gestational age of pregnant women who have not experienced a miscarriage is calculated based on the weeks of pregnancy at the time of delivery.

Results

Participant recruitment

A total of 1058 participants were enrolled, of whom 997 participants attended. Subjective sleep quality was measured with the PSQI questionnaire. Depression and anxiety status were measured with SDS and SAS questionnaires, respectively. 902 of 997 participants attended follow-up offering the above three questionnaire, the response rate was 90.5% (Fig. S1). Of these, 217 participants lost to follow-up or went to other hospital or did not submit the complete questionnaire, 665 participants entered the final analysis of anxiety, depression and sleep quality (Fig. S1). Among the 665 participants included in the final analysis, 444 were from the three campuses of the First Affiliated Hospital of Zhejiang University School of Medicine, and 221 were from Affiliated Jinhua Hospital Zhejiang University School of Medicine.

Basic characteristic

Among all the 665 pregnant women we followed, 10.4% (n = 69) were unemployed, 91.3% (n = 607) had bachelor or above degree, 32.8% (n = 218) were overweight or obese, 8.9% (n = 59) had basic diseases, 69.0% (n = 459) were having their first pregnancy, 28.6% (n = 190) had a history of cesarean section, 28.1% (n = 187) had a history of miscarriage, 19.7% (n = 131) had a history of operation. The mean age at baseline was 30.1 (SD: 3.4) (Table 1). The results for each variable grouped by miscarriage are shown in Table S1 (appendix pp5–pp6).

Table 1.

Demographics, anxiety, depression and sleep quality of the study participants.

	Overall
n	665
Age (mean (SD))	30.07 (3.4)
Education (%)
Bachelor’s degree	502 (75.5)
Master degree or above	105 (15.8)
Senior high school and below	58 (8.7)
Prepregnant BMI
Underweight	33 (5.0)
Normal weight	414 (62.3)
Overweight	197 (29.6)
Obese	21 (3.2)
Occupation = unemployed (%)	69 (10.4)
Basic disease = Yes (%)	59 (8.9)
Bad habits = Yes (%)	6 (0.9)
First pregnancy = Yes (%)	459 (69.0)
History of cesarean section (%)
Reject to answer	144 (21.7)
Yes	190 (28.6)
No	331 (49.8)
Operation = Yes (%)	131 (19.7)
Miscarriage = Yes (%)	109 (16.4)
History of miscarriage (%)
No	478 (71.9)
Once	137 (20.6)
Twice or more	50 (7.5)
Status
Normal production	507 (76.2)
Miscarriage	109 (16.4)
Premature delivery	49 (7.4)
WBCa (mean (SD))	9.17 (2.4)
NEU%a (mean (SD))	71.14 (7.7)
NEUa (mean (SD))	6.56 (2.3)
PLTa (mean (SD))	223.1 (63.0)
ALBa (mean (SD))	39.1 (3.2)
D-dimer (mean (SD))	1593.3 (1186.5)
CRPa (mean (SD))	28.2 (31.4)
Anemia	515 (77.4)
Iron deficiency	538 (80.9)
Gestational age of losses(mean (SD))	8.81 (2.31)
PSQIa (mean (SD))	8.7 (2.4)
Poor sleep quality (PSQI score ≥ 5) = Yes(%)	614 (92.3)
SASa (mean (SD))	39.4 (5.6)
SAS standard score (mean (SD))	49.3 (7.1)
Anxiety (SAS standard score ≥ 50) = Yes(%)	294 (44.2)
SDSa (mean (SD))	44.91 (6.1)
SDS standard score (mean (SD))	55.1 (7.6)
Depression (SDS standard score ≥ 53) = Yes(%)	462 (69.5)

^aWBC white blood cell count, NEU% neutrophil percentage, NEU neutrophil count, PLT platelet count, ALB albumin, CRP C-reactive protein, SAS self-rating anxiety scale questionnaire, SDS self-rating depression scale questionnaire, PSQI Pittsburgh sleep quality index questionnaire, SD standard deviation.

Sleep quality, anxiety and depression

The average global PSQI score was 8.7 (SD: 2.4) of the pregnant women participated in this study, 92.3% (n = 614) of them had poor sleep quality. The average SAS standard score was 49.3 (SD: 7.1), 44.2% (n = 294) were anxiety. The average SDS standard score were 55.1 (SD: 7.6), 69.5% (n = 462) were depression (Table 1). The results for each variable grouped by miscarriage are shown in Table S1 (appendix pp5–pp6).

Incidence rate of miscarriage in populations with different characteristics

Out of 665 pregnant women, 109 (16.4%, 95%CI:[13.8–19.4%]) have a miscarriage. Compared with women who did not miscarriage, participants who had a miscarriage tend to have significantly higer SAS scores (P < 0.05) (Fig. 1A). Similarly, SDS and PSQI scores tend to have higher values in those women who had a miscarriaged (Fig. 1B, C). The pre-pregnancy BMI of women in the miscarriage group was significantly lower than that of women in the normal group (Fig. 1D). While age and history of miscarriage did not have a relationship with miscarriage of participant pregnant women (Fig. 1E, F). Education, history of operation, occupation, first pregnancy, history of cesarean section and basic disease did not have relationship with miscarriage of participant pregnant women either (Fig. S2). The association between inflammatory markers and miscarriage in pregnant women can be seen from Fig. S3. There is a significant correlation between PLT and NEU% with miscarriage in pregnant women.

Fig. 1. The association between SAS scores, SDS scores, PSQI scores, prepregnant BMI, age and history of miscarriage and miscarriage.

A The distribution of SAS standard score between those participants who had a miscarriage or not. B The distribution of SDS standard score between those participants who had a miscarriage or not. C The distribution of PSQI scores between those participants who had a miscarriage or not. D The distribution of prepregnant BMI values between those participants who had a miscarriage or not. E The association between age and miscarriage. F The association between the history of miscarriage and miscarriage. The top row of each panel displays the following numbers: parameter, statistic, significance, effect size, confidence intervals, and number of observations. The bottom right corner of each panel shows the p-value correction methods used for pairwise comparisons. Any significant differences in pairwise comparisons are indicated by continuous lines above the graph, with corresponding p-values displayed. Abbreviation: SAS Self-Rating Anxiety Scale questionnaire, SDS Self-Rating Depression Scale questionnaire, PSQI Pittsburgh Sleep Quality Index questionnaire.

Association between sleep quality, anxiety, depression and miscarriage

The relationship between sleep quality and miscarriage was assessed. Participants with poor sleep quality (assessed by the Pittsburgh Sleep Quality Index) were more likely to have a miscarriage compared to participants who reported lower PSQI values (Fig. 2A). A similar association was observed between anxiety, depression and miscarriage. Participants who reported higher SAS and SDS value were more likely to have a miscarriage compared to participants who reported lower SAS and SDS values (Fig. 2A). The sensitivity analysis results indicate a correlated relationship between anxiety and depression with miscarriage, while no significant correlation was found between sleep quality and miscarriage (Fig. 2B).

Fig. 2. Association between sleep quality, anxiety, depression and miscarriage.

Both unadjusted (squares) or adjusted (circles) relative risks are shown alongside 95% CIs. A The results of multivariable multinomial logistic regression models. B The results of sensitivity analysis (cox regression models). The association was adjusted for age, BMI, basic disease, occupation, education, first pregnancy, history of cesarean section, NEU%, PLT and operation in four models. Abbreviation: SAS Self-Rating Anxiety Scale questionnaire, SDS Self-Rating Depression Scale questionnaire, PSQI Pittsburgh Sleep Quality Index questionnaire, OR Odds Rates, CI Confidence interval.

Dose-response relationships between miscarriage rate and anxiety, sleep quality, depression

Using multivariable logistic regression model combined with RCS method, the dose-response relationships between SAS score, PSQI score, SDS score, age, and miscarriage were analyzed separately (Fig. 3). The overall relationship between PSQI score and miscarriage rate indicates that higher PSQI scores are associated with higher miscarriage rates. Initially, the miscarriage rate slowly increases with increasing PSQI score. However, after PSQI ≥ 12, even slight increases in PSQI have a significant impact on the miscarriage rate. The overall relationship between SDS score and miscarriage rate is also positive. When SDS score increases, the miscarriage rate initially remains stable. However, after SDS ≥ 70, the miscarriage rate increases slowly at first and then sharply with increasing SDS score. A similar pattern is observed for SAS score. When SAS score < 53, it has no effect on the miscarriage rate. However, when SAS score ≥ 53, the miscarriage rate sharply increases with increasing SAS score. Age shows a clear negative correlation with miscarriage rate, indicating that the miscarriage rate decreases as age increases (Fig. 3). Age displays a clear negative correlation with miscarriage rate (Fig. 3).

Fig. 3. Dose-response relationships between miscarriage rate and PSQI scores, SAS scores, SDS scores, NEU%.

We combined multivariable multinomial logistic regression models and RCS methods, the association was adjusted for age, BMI, basic disease, occupation, education, first pregnancy, history of cesarean section, NEU%, PLT and operation. The red lines represent the changes in the odds ratio (OR) with respect to SAS, SDS, PSQI, and NEU%. The red shaded area represents the 95% confidence interval (CI). The black horizontal dashed line represents the reference line Y = 1. Abbreviation: SAS Self-Rating Anxiety Scale questionnaire, SDS Self-Rating Depression Scale questionnaire, PSQI Pittsburgh Sleep Quality Index questionnaire, OR Odds Rates, CI Confidence interval, RCS restricted cubic spline.

Prediction the probability of miscarriage based on nomogram model

Based on the logistic regression model, a nomogram model was constructed to predict the probability of miscarriage occurrence (Fig. 4). By combining significant variables from logistic regression: age, SAS score, SDS score, and PSQI score, the corresponding scores for each indicator are given on the point scale axis. The total score is obtained by adding the score values for each indicator. The probability of miscarriage in pregnant women is then determined based on the total score, with higher scores indicating a greater likelihood of miscarriage. The discriminative power of this nomogram prediction model is shown in Fig. S4, with an AUC of 89.9%, corrected pAUC of 78.4%, indicating excellent predictive performance. The model’s calibration curve closely aligns with the reference line, indicating strong consistency between predicted and actual probabilities (Fig. S5).

Fig. 4. Nomogram model to predict the probability.

The values below the first four graduated axes in the predictive model for miscarriage in pregnant women correspond to the actual values of the indicators. The corresponding score values for each variable are indicated by the colors in the legend on the right side. The sum of the score values for the four variables yields the score value above the final graduated axis. Each score value corresponds to a different miscarriage risk below the final graduated axis. Abbreviation: SAS Self-Rating Anxiety Scale questionnaire, SDS Self-Rating Depression Scale questionnaire, PSQI Pittsburgh Sleep Quality Index questionnaire, NEU_ Neutrophil Percentage.

Mediation analysis

Mediation analysis was done to investigate the contribution of anxiety and sleep quality in mediating the effect between depression and miscarriage (Fig. 5A). Anxiety mediated the effect of depression on miscarriage by 44% (95% CI: 28%, 56%) and sleep quality had a similar mediation effect (16% [95% CI 8%, 24%]); Fig. 5B; Table S2, appendix p9), and their directed acyclic graph are shown in Fig. S6 (appendix pp 9).

Fig. 5. The effect of anxiety or sleep quality in mediating the effect of depression on miscarriage.

A Mediation model frame: Mediation models investigated the association between depression and miscarriage to investigate whether anxiety and sleep quality could be considered mediators in the relationship. The detail results are reported in Supplementary Table S2 (appendix pp9). The pathway labelled c′ represents the direct effect of depression on miscarriage. The pathways labelled a_i represent the effect of depression on the hypothesis mediators (Anxiety and Sleep quality). Lastly, the pathways labelled b_i represent the effect of the mediators on miscarriage and are calculated whilst controlling for depression. B The results of mediation models. Mediation models were used to investigate the effects of sleep quality or anxiety, recognised causes of miscarriage, in mediating the association between depression and miscarriage.

Discussions

To our knowledge, this is the first study exploring the mediating role of anxiety and sleep quality in the relationship between depression and the miscarriage rate, and their dose response relationship and interaction relationship. Of all the participants, 92.3% (n = 614) of them had poor sleep quality, 44.2% were anxiety and 69.5% were depressed. Previous epidemiological studies have demonstrated that the lifetime prevalence of major depressive disorder in women (21.3%) is nearly twice that in men (12.7%) [9]. In our investigation, we found that 69.5% of pregnant women experienced depression, a significantly higher rate than previous research [9]. This could be due to the fact that the survey was conducted during the COVID-19 pandemic, which may have exacerbated depressive symptoms among pregnant women. Alternatively, it could be that poor sleep quality among pregnant women contributed to the increased incidence rate of depression.

During our investigation, 16.4% of the participant pregnant women had a miscarriage. Our results showed a significant association between anxiety, depression and miscarriage both unadjusted and multivariable multinomial logistic regression, as well as the sensitivity analysis. This result aligns with a prospective cohort study in UK Biobank [23]. Scholars from Australia utilized a longitudinal cohort study to investigate the relationship between depression and miscarriage in pregnant women. They found that women with a history of depression (excluding postnatal depression) were at higher risk of infertility [risk ratio (RR) = 1.34, 95% confidence interval (CI): 1.21–1.48], miscarriage (RR = 1.22, 95% CI: 1.10–1.34), and recurrent miscarriages (≥2; RR = 1.39, 95% CI: 1.17–1.64), compared to women without a history of depression [10]. These findings are consistent with those of our study (where the odds ratio for depression was 1.22 [1.18, 1.27]). Therefore, we need to pay attention not only to postpartum depression and anxiety in pregnant women but also to anxiety and depression during pregnancy, in order to reduce adverse outcomes for them. Similarly, our results showed a significant association between sleep quality and miscarriage both unadjusted and multivariable multinomial logistic regression, while sensitivity analysis did not reveal this association, which is consistent with the previous study [23]. Therefore, the quality of sleep during pregnancy also requires special attention. The sensitivity analysis in this study did not reveal a relationship between sleep quality and miscarriage, which might be due to the lower proportion of pregnant women with poor sleep quality and the narrower range of the PSQI in this study. Therefore, future research should expand the sample size and collect data on sleep quality of pregnant women from different periods.

RCS results showed that the miscarriage rate slowly increases with increasing PSQI, SAS and SDS score at first. However, when a certain threshold is reached, even slight increases in the scores of SAS, SDS, and PSQI will lead to a sharp rise in the miscarriage rate. The threshold for PSQI, SAS and SDS is 12, 53, and 70, respectively. This result intuitively demonstrates the quantitative relationships between the SAS, SDS, and PSQI scores, revealing their significant impact on miscarriage. Therefore, when pregnant women’s scores on the PSQI, SAS, and SDS exceed the threshold values, doctors, pregnant women, and their families should pay additional attention and seek professional help if necessary. We did not find a significant association between age and miscarriage in pregnant women [23], which is inconsistent with previous research findings, which may be attributed to the relatively narrow age range included in this study.

This study innovatively used anxiety, depression, and sleep quality as predictors to construct a predictive model for miscarriage in pregnant women. Compared to traditional models that incorporate clinical blood indicators, lifestyle factors, age, and history of miscarriage [12, 13, 24, 25], this approach enables earlier prevention and control of miscarriage. This study utilized a nomogram model to identify high-risk pregnant women, necessary interventions and treatments can be initiated in advance, providing personalized care and support such as psychological counseling, support, cognitive behavioral therapy, and sleep management, which can alleviate the adverse effects of insufficient sleep, anxiety, and depression on maternal health and reduce the risk of miscarriage. It can also help healthcare institutions and government departments optimize resource allocation. By identifying high-risk groups, limited resources can be prioritized for pregnant women in need of greater attention and intervention, thus improving resource efficiency. Additionally, it contributes to raising public awareness of maternal health, promoting overall societal health consciousness, and increasing happiness indices.

In this study, we found anxiety mediated the effect of depression on miscarriage and sleep quality had a similar mediation effect. This mediating effect helps to explain the complex associations between these variables, unveiling the mechanisms behind the outcomes. Moreover, since the subjective assessment of depression can be challenging [26], studying the mediating roles of anxiety and sleep quality in the relationship between depression and miscarriage recurrence helps clarify that poor anxiety and sleep quality exacerbate the impact of depression on miscarriage. This insight can guide clinicians to adopt more comprehensive care approaches. Clinicians can implement targeted interventions aimed at reducing anxiety and improving sleep quality among pregnant woman with poor anxiety and sleep quality, and develop and implement early intervention programs focused on decreasing anxiety and enhancing sleep quality.

The prospective cohort study design, controlling various risk factors related with miscarriage, and the first insight into the association and mediation effect of anxiety, sleep quality on the relationship between depression and miscarriage and uncovered the dose-relationship of them are the strengths of this study. However, there are several limitations in this study. First, quantification of sleep quality, anxiety, depression based on questionnaires relied upon participant recall and therefore could be affected by recall bias. Therefore, this study utilized internationally recognized standard questionnaires for sleep quality, depression and anxiety, and tried to use quantitative measures to minimize recall bias as much as possible. Second, selection bias could also affect the results. To reduce selection bias, we conducted the survey across three different Subsidiary Hospitals and Affiliated Jinhua Hospital Zhejiang University School of Medicine and increased the sample size as much as possible to mitigate the impact of selection bias on the results. Despite these limitations, our findings are helpful to better understand the role of anxiety, depression and sleep quality on health among pregnant women.

In conclusion, anxiety and depression and poor sleep quality were associated with a higher risk of miscarriage. Anxiety and sleep quality also play a significant role in the impact of depression on the miscarriage rate. By revealing high-risk pregnant women through the nomogram model, early intervention and personalized care and support can be provided, aiming to reduce the miscarriage rate among pregnant women. Our findings highlight the importance of healthy mental state of pregnant women.

Availability of data and materials

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

JHP and XDZ conceptualized this article and wrote the original draft of the manuscript. JHP, LX, XDZ, SYY, JJ, QL, JQL, PPZ, LYZ, XJQ, DLL, ZLX, XYL contributed to the study design, data collection, methodology, data analysis, and data interpretation. SYY, JJ, QL, PPZ, and TAJ reviewed and edited the manuscript. TAJ supervised the study group. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. All authors read and approved the final version of the manuscript.

Funding

National Key Scientific Instrument and Equipment Development Projects of China (82027803), National Natural Science Foundation of China (81971623).

Code availability

All code is available upon reasonable request.

Ethics approval and consent to participate

The studies involving human participants were reviewed and approved by First Affiliated Hospital of Zhejiang University School of Medicine and Affiliated Jinhua Hospital Zhejiang University School of Medicine. The participants provided their oral informed consent to participate in this study. Consent for publication was obtained from all participants.

Ethics

The study was approved by the institutional review boards at the First Affiliated Hospital of Zhejiang University School of Medicine (IIT20230328A) and Affiliated Jinhua Hospital Zhejiang University School of Medicine (20241030101), and all participants gave written informed consent at the enrollment. All methods were performed in accordance with the relevant guidelines and regulations.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-025-03363-x.