Depression, anxiety, and stress across the perinatal period: longitudinal trends and associations with adverse outcomes

Abstract

Background

Depression, anxiety, and stress are common during the perinatal period and may adversely affect maternal and neonatal outcomes. However, longitudinal data on their trajectories across pregnancy and the postpartum period, as well as their associations with adverse outcomes, remain limited.

Aim

To assess the prevalence and longitudinal trends of depression, anxiety, and stress from the first trimester through three months postpartum and to examine their associations with adverse perinatal outcomes, along with sociodemographic and obstetric factors.

Methods

This prospective cohort study included 253 pregnant women followed from early pregnancy to three months postpartum at five time points. Psychological symptoms were assessed using the Depression, Anxiety and Stress Scale–21 (DASS-21). Longitudinal changes were analyzed using linear mixed-effects models. Associations with perinatal outcomes (cesarean delivery, postpartum hemorrhage, gestational diabetes mellitus, gestational hypertension, and selected neonatal outcomes) were examined using univariable and multivariable Firth penalized logistic regression.

Results

The prevalence of clinically relevant depressive symptoms ranged from 4.5% to 9.9%, anxiety symptoms from 18.2% to 28.3%, and stress symptoms from 12.3% to 20.3% across the study period. Depressive symptoms remained relatively stable across the perinatal period. After correction for multiple testing, only anxiety showed a significant decline after childbirth, while no consistent significant changes were observed for stress. In multivariable analyses, cesarean delivery was associated with BMI ≥ 25 kg/m² and a history of obstetric complications, whereas no significant associations were found for postpartum hemorrhage, small for gestational age, or gestational diabetes.

Conclusions

Perinatal psychological symptoms show distinct longitudinal patterns, with a decline in anxiety after childbirth. Most associations with adverse outcomes were not statistically significant or were imprecise. Larger multicenter studies are needed to clarify these relationships.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-026-09114-0.

Introduction

Anxiety and depressive disorders account for 9.1% of all diseases and 63.1% of mental health disorders. The majority of cases and disability-adjusted life years (DALYs) are reported in countries with low and middle sociodemographic index (SDI). The prevalence of both disorders increases with age, particularly among individuals aged 10–24 years, and is consistently higher in females [1].

The prevalence of depression is significantly higher among women, who are almost twice as likely to be diagnosed with depressive disorders compared with men [2]. Anxiety symptoms frequently co-occur with depressive symptoms. This increased risk is likely attributable to a combination of biological and psychosocial factors, including hormonal changes across the lifespan and gender-related life experiences [3]. These factors become particularly pronounced during pregnancy and the postpartum period, when women demonstrate increased vulnerability to the development of these disorders [4, 5]. Stress represents a distinct but closely related dimension of perinatal mental health, arising from psychological and physiological responses to pregnancy-related demands, including physical changes, concerns about childbirth, relationships, and the health of the baby [6]. Distinguishing between depression, anxiety, and stress is important, as these conditions may influence pregnancy outcomes through different biological pathways. For example, anxiety and stress are more strongly linked to activation of the hypothalamic–pituitary–adrenal (HPA) axis and increased cortisol levels, whereas depression may involve different mechanisms. However, these constructs are often conflated in perinatal research [7].

Al-Abri et al. reported a pooled global prevalence of perinatal depression of 26.3%, with depressive symptoms affecting approximately one-quarter of women during pregnancy (28.5%) and the postpartum period (27.6%) [8]. Similarly, a meta-analysis by Dennis et al. showed that the pooled prevalence of anxiety symptoms during pregnancy was 22.9%, while postnatal anxiety symptoms were observed in 17.8% of women at 1–4 weeks postpartum and in up to 9.3% of women more than 24 weeks after delivery [9]. A literature review by Andrade et al. indicates that stress during pregnancy is highly prevalent, with reported rates ranging from 11.6% to 91.86% [10].

However, trends of anxiety, depression, and stress across pregnancy and the postpartum period remain insufficiently clarified. One study examining normative positive and negative affect during pregnancy (without postpartum follow-up) found that negative affect followed a U-shaped pattern, with the highest levels in early pregnancy, a decline in the second trimester, followed by an increase in late pregnancy [11]. A longitudinal study by Ching-Yu Cheng et al., which assessed depression, anxiety, and stress at five time points, found increasing anxiety and depressive symptoms from the 24th week of pregnancy until delivery, while perceived stress decreased during pregnancy but rose again after childbirth [12].

Unrecognized and untreated perinatal mental health problems are associated with adverse maternal and neonatal outcomes, including gestational hypertension, gestational diabetes, preeclampsia, intrauterine growth restriction, and later cognitive and psychosocial developmental difficulties in children [6, 13–15]. These associations may be mediated by the dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis and increased pro-inflammatory cytokine production, which can impair placental perfusion and trigger systemic maternal physiological changes [7]. Beyond these neuroendocrine pathways, adverse pregnancy outcomes are influenced by a broad range of genetic, epigenetic, and environmental factors acting through complex biological and psychosocial mechanisms [16–19].

In Serbia, the number of studies examining the prevalence of perinatal depression and anxiety remains limited [20–22]. To date, no studies have longitudinally assessed depressive symptoms, anxiety, and stress across all trimesters of pregnancy and the postpartum period, nor have they examined their associations with perinatal outcomes within a single analytical framework. This lack of longitudinal data limits our ability to identify specific sensitive periods during pregnancy when psychological distress may most significantly impact maternal and neonatal health.

Despite increasing research on perinatal mental health, longitudinal studies that simultaneously assess depression, anxiety, and stress across several stages of pregnancy and the postpartum period remain limited. In particular, evidence on how these symptoms evolve over time and how they relate to adverse maternal and neonatal outcomes is still insufficient. To address this gap, the present prospective cohort study followed pregnant women at five measurement points, from the first trimester to three months postpartum. The study aimed to evaluate the prevalence and trajectories of depression, anxiety, and stress using the DASS-21 scale, and to examine their associations, along with sociodemographic and obstetric factors, with adverse perinatal outcomes.

Materials and methods

Study design

This prospective cohort study was conducted at the Obstetrics and Gynecology Clinic “Narodni front” in Belgrade, Serbia, from July 2023 to April 2025. Participants were recruited using a consecutive sampling strategy during their first-trimester aneuploidy screening visit between July 2023 and January 2024. All pregnant women attending the clinic during this period who met the eligibility criteria were invited to participate. Participants were followed throughout pregnancy and up to three months postpartum.

Participants and selection criteria

The study included 253 pregnant women attending routine prenatal visits at the “Narodni front” Clinic, a tertiary care institution where all pregnant women are offered first-trimester aneuploidy screening. Enrollment was performed by clinical researchers during the initial first-trimester screening window.

Eligible participants were women with a singleton, physiological pregnancy defined as a low-risk pregnancy without known pre-existing chronic medical conditions (such as chronic hypertension, pre-gestational diabetes, or autoimmune diseases) and without clinically identified pregnancy complications at the time of recruitment. Participants were recruited between 11 and 14 gestational weeks during their first-trimester screening visit. This visit included an ultrasound examination and blood sampling for the measurement of placental hormones, which are used to assess the biochemical risk of aneuploidy. Participants were also offered the option to continue pregnancy follow-up at the same institution.

Inclusion criteria were: age ≥ 18 years, singleton physiological pregnancy, and signed informed consent.

Exclusion criteria were: multiple pregnancy; a documented diagnosis of a severe psychiatric disorder (including schizophrenia, bipolar disorder, or other psychotic disorders), based on medical records or specialist psychiatric reports available at the time of recruitment, that were present or clinically active at study inclusion; or insufficient knowledge of the Serbian language.

Follow-up

Participants were followed throughout pregnancy and up to three months postpartum. Data were collected at five time points: first trimester (T1; 11–14 gestational weeks, mean gestational age 12.9 ± 0.8 weeks), second trimester (T2; 20–24 gestational weeks, mean 23.3 ± 2.2 weeks), third trimester (T3; 30–36 gestational weeks, mean 30.8 ± 1.8 weeks), one month postpartum (PP1; 3–4 weeks after delivery, mean 3.5 ± 0.4 weeks), and three months postpartum (PP3; 12–13 weeks after delivery, mean 12.4 ± 0.3 weeks).

Participants completed study questionnaires either in paper format during routine visits or electronically via a secure online link. Data collection was supervised by trained clinical researchers and medical staff at each time point to ensure consistency and to address any participant queries.

Ethical considerations

The study protocol was approved by the Ethics Committee of the Obstetrics and Gynecology Clinic “Narodni front” (approval No. 22008-2023-009349). All participants provided written informed consent after being informed about the study objectives. Participation was voluntary, and women could withdraw from the study at any time without consequences.

Data collection

Data were collected using the following instruments:

General Questionnaire – collected sociodemographic characteristics such as age, occupation, educational level, marital status, and place of residence.

Pregnancy Risk Assessment Questionnaire, adapted from the Pregnancy Risk Assessment Monitoring System Phase Eight (PRAMS-8), developed by the U.S. Centers for Disease Control and Prevention (CDC) (2016–2022) [23]. The PRAMS-8 collects information on women’s experiences before, during, and after pregnancy to better understand factors influencing maternal and neonatal health. For this study, selected items were used, and some were partially modified (e.g., addition of response options related to pregnancy and delivery complications) to align with the study objectives and follow-up periods.

The Depression, Anxiety, and Stress Scale-21 (DASS-21) is a validated self-report instrument designed to measure symptoms of depression, anxiety, and stress over the previous week. It consists of 21 items divided into three subscales, each containing seven items. The DASS-21 subscale scores were calculated by summing the scores for the relevant items and multiplying them by two to ensure comparability with the full-length DASS-42 scale, in accordance with standard scoring instructions.

Symptom severity was categorized using the recommended cut-off values for the DASS scale (depression: normal 0–9, mild 10–13, moderate 14–20, severe 21–27, extremely severe ≥ 28; anxiety: normal 0–7, mild 8–9, moderate 10–14, severe 15–19, extremely severe ≥ 20; stress: normal 0–14, mild 15–18, moderate 19–25, severe 26–33, extremely severe ≥ 34) [24]. For the purposes of prevalence estimation, each domain—depression, anxiety, and stress—was dichotomized into binary variables: normal (combining normal and mild categories) and presence of symptoms (combining moderate, severe, and extremely severe categories). This categorization was applied to focus on clinically relevant symptom levels rather than subclinical manifestations.

The DASS-21 was chosen for its ability to simultaneously assess depression, anxiety, and stress as related but distinct dimensions of psychological distress, allowing a broader evaluation of maternal mental health. This was particularly relevant for the longitudinal design of the study, enabling the assessment of multiple domains of psychological distress across pregnancy and the postpartum period [25]. In addition, a validated Serbian version of the DASS-21 is available and has been widely used in population and clinical studies [26].

The questionnaire administered one month after delivery included items on delivery-related complications and newborn characteristics (gestational age at birth and birth weight). Low birth weight (LBW) and small for gestational age (SGA) were defined according to standard obstetric criteria as birth weight < 2500 g and birth weight below the 10th percentile for sex and gestational age, respectively, based on national reference growth charts used in routine clinical practice at our institution [27].

In addition, data on the following perinatal complications were extracted from medical records (patient histories and discharge summaries): preterm birth (gestational age ≥ 28 and < 37 weeks), miscarriage, gestational hypertension (defined as blood pressure ≥ 140/90 mmHg after 20 weeks of gestation), gestational diabetes (diagnosed via a 75 g OGTT at 24–28 weeks, with T2 psychological assessments conducted prior to screening), and postpartum hemorrhage (defined as blood loss ≥ 500 mL for vaginal or ≥ 1000 mL for cesarean delivery).

Outcomes

The outcomes of interest were selected based on their clinical relevance and included cesarean section, gestational diabetes mellitus (GDM), gestational hypertension, postpartum hemorrhage (PPH), preterm birth, and birth weight–related outcomes (macrosomia, low birth weight, and small for gestational age).

Statistical analysis

Sample size

Sample size was calculated based on previously reported prevalence of depression and anxiety during pregnancy and the postpartum period [8, 9]. Assuming a precision of 0.06, a 95% confidence level, and an estimated prevalence of depression of 26.3% in the Serbian population, the minimum required sample size was 207 participants. Using the same parameters and an estimated anxiety prevalence of 22.9%, the required sample size was 189 participants. The larger estimate was adopted, resulting in a final minimum sample size of 207 participants. Given the longitudinal design of the study and repeated follow-up assessments, the baseline sample size was increased by approximately 20% to account for potential loss to follow-up and to ensure adequate statistical power for longitudinal and regression analyses, resulting in the inclusion of 253 participants.

Data analysis

Descriptive and inferential statistical analyses were performed. Categorical variables were presented as absolute and relative frequencies. Continuous variables were summarized as means with standard deviations or as medians with ranges (minimum–maximum), depending on their distribution.

Participant flow and the number of completed assessments at each time point are presented in the study flow chart (Fig. 1). Missingness across study waves was examined descriptively by reporting the number of available observations at each assessment (T1, T2, T3, PP1, PP3). All participants attending a given assessment completed the questionnaires; therefore, there was no item-level missingness, and missing data occurred only due to attrition between study waves.

Fig. 1.

Flowchart of participants followed from the first trimester of pregnancy to 3 months postpartum

To assess potential selection bias due to loss to follow-up, baseline sociodemographic and psychological characteristics were compared between participants who completed all follow-up assessments (N = 220) and those lost to follow-up (N = 33). Differences between groups were assessed using χ² or Fisher’s exact tests for categorical variables and t-tests or Mann–Whitney U tests for continuous variables, as appropriate.

Longitudinal changes in depression, anxiety, and stress scores across five time points were analyzed using linear mixed-effects models, with time treated as a categorical fixed effect (T1–PP3) and subject included as a random intercept to account for within-person correlation of repeated measures. This approach allowed the inclusion of all available observations and provided valid estimates under the missing-at-random assumption.

To account for multiple testing in the analysis of longitudinal changes in psychological outcomes, false discovery rate (FDR) correction was applied to pairwise comparisons across time points within each psychological domain.

Logistic regression analyses were used to examine associations between sociodemographic, pre-pregnancy health, obstetric and pregnancy-related factors, as well as depression, anxiety, and stress during pregnancy, and perinatal complications. For these analyses, a subset of outcomes was pre-specified as primary (gestational diabetes mellitus, gestational hypertension, postpartum hemorrhage, and small for gestational age), while analyses involving other outcomes were considered exploratory. Given the pre-specification of primary outcomes and the exploratory nature of these analyses, false discovery rate (FDR) correction was not applied to the regression models and unadjusted p-values are presented.

Psychological distress was assessed at predefined time points during pregnancy (first, second, and third trimester). For each outcome, only exposure variables that clearly preceded the expected timing of outcome onset or diagnosis were considered in regression models, ensuring appropriate temporal ordering (as illustrated in Fig. 2). Therefore, second- and third-trimester depression, anxiety, and stress scores, as well as GWG, were not considered in models for GDM and gestational hypertension, as they did not clearly precede these outcomes.

Fig. 2.

Timeline of measurements and outcomes. Only variables that preceded outcome diagnosis were included in regression models

Covariates were selected based on clinical relevance and prior evidence. Gestational weight gain was considered as a pregnancy-related factor potentially associated with both maternal psychological status and obstetric outcomes; however, it was included only in analyses where temporal ordering could be reasonably assumed.

Multivariable analyses were performed using Firth penalized logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs), in order to reduce small-sample bias in models with limited numbers of events. Candidate predictors were selected based on clinical relevance and univariable associations, while considering events-per-variable to limit overfitting.

Complete-case analysis was applied. In this study, missing data arose solely from attrition between study waves, as all participants who attended a given assessment completed the questionnaires. To assess the potential impact of missing data, a sensitivity analysis was performed by comparing baseline characteristics between participants who completed follow-up assessments and those lost to follow-up. The groups were largely comparable, suggesting that the risk of substantial bias due to missing data is limited. Therefore, the use of complete-case analysis was considered appropriate, although the results should be interpreted with caution.

Statistical significance was set at p < 0.05, and results were presented as odds ratios with 95% confidence intervals. Given the limited number of events and multiple analyses across outcomes and time points, the association analyses were interpreted as exploratory.

All statistical analyses were performed using IBM SPSS Statistics for Windows, version 23.0 (IBM Corp., Armonk, NY, USA) and R software (version 4.3.1, R Foundation for Statistical Computing, Vienna, Austria).

Results

Of the 253 participants included in the study, 220 completed the entire follow-up period from pregnancy to three months postpartum, while 33 were lost to follow-up. This corresponds to an attrition rate of 13%, which is within the range reported in comparable longitudinal studies of perinatal mental health, where loss to follow-up has been reported to vary from approximately 2.6% to 46.8% [12, 28–30].

Comparisons between participants who completed follow-up and those lost to follow-up were conducted to assess potential selection bias due to attrition. Participants lost to follow-up were significantly more often unmarried (p = 0.035). No statistically significant differences were observed between the two groups with respect to other sociodemographic characteristics. Baseline psychological distress also did not differ significantly between participants who completed follow-up and those lost to follow-up (Table 1).

Table 1.

Baseline sociodemographic and psychological characteristics of participants who completed follow-up and those lost to follow-up

Demographic characteristics	Total sample (T1) N = 253	Lost to follow- up N = 33	Completed follow-up (PP3) N = 220	p value
Age (years), n (%)
≤ 35	196 (77.5)	25 (75.8)	171 (77.7)	0.801
> 35	57 (22.5)	8 (24.2)	49 (22.3)
Education, n (%)
High school or Lower	81 (32.0)	9 (27.3)	72 (32.7)	0.531
University degree	172 (68.0)	24 (72.7)	148 (67.3)
Employment, n (%)
Unemployed	28 (11.1)	3 (9.1)	25 (11.4)	1.000
Employed	225 (88.9)	30 (90.9)	195 (88.6)
Marital status, n (%)
Unmarried	96 (37.9)	18 (54.5)	78 (35.5)	0.035
Married	157 (62.1)	15 (45.5)	142 (64.5)
Place of residence, n (%)
Belgrade	215 (85.0)	28 (84.8)	187 (85.0)	0.982
Other	38 (15.0)	5 (15.2)	33 (15.0)
Lives with husband/partner, n (%)				1.000
No	14 (5.5)	1 (3.0)	13 (5.9)
Yes	239 (94.5)	32 (97.0)	207 (94.1)
Monthly household income per household member (€), n (%)
≤ €800	170 (67.2)	23 (69.7)	147 (66.8)	0.743
> €800	83 (32.8)	10 (30.3)	73 (33.2)
Overall health, n (%)
Bad/Average	23 (9.1)	5 (15.2)	18 (8.2)
Good	113 (44.7)	13 (39.4)	100 (45.5)	0.412
Very good	117 (46.2)	15 (45.5)	102 (46.4)
Depression score, median (min-max)	2 (0–38)	2 (0–38)	2 (0–34)	0.550
Anxiety score, median (min-max)	4 (0–36)	6 (0–36)	4 (0–30)	0.302
Stress score, median (min-max)	10 (0–42)	12 (0–42)	10 (0–42)	0.626

p values represent comparisons between participants who completed follow-up and those lost to follow-up. Categorical variables were compared using the χ² test or Fisher’s exact test, as appropriate, and continuous variables using the Mann–Whitney U test. T1 – first trimester; PP3 – three months postpartum

Most women were nulliparous, with over half having no previous deliveries (51.8%). A minority reported a history of spontaneous (22.1%) or induced abortion (7.9%), and only a small proportion had experienced stillbirth (2.4%). Among women with at least one prior delivery, complications were reported in 32.0% of previous pregnancies and 28.7% of prior deliveries. The majority of current pregnancies were planned (89.3%), and 4.7% of women conceived following infertility treatment (Table 2).

Table 2.

Participants’ obstetric history and pre-pregnancy health status (N = 253)

Pre-pregnancy health status	N = 253 n (%)
Parity (number of previous deliveries)
0 (nulliparous)	131 (51.8)
1	78 (30.8)
2+	44 (17.4)
Reproductive history
History of spontaneous abortion	56 (22.1)
History of induced abortion	20 (7.9)
History of stillbirth	6 (2.4)
Type of previous delivery*
Vaginal	98 (80.3)
Caesarean section	22 (18.0)
Both vaginal and caesarean	2 (1.7)
Previous pregnancy-related complications*
Complications during previous pregnancy	39 (32.0)
Complications during previous delivery	35 (28.7)
Current pregnancy
Planned current pregnancy	226 (89.3)
Infertility treatment	12 (4.7)

*Percentages for type of previous delivery and complications during previous pregnancy /delivery are calculated among women with at least one previous delivery (N = 122)

Smoking and alcohol consumption during pregnancy remained relatively low and demonstrated a slight decline over time. At T1, according to BMI categories, 21.7% of women were classified as overweight and 9.1% as obese. Gestational weight gain (GWG) was within recommended ranges in 60% of women, while 22% experienced excessive and 17.6% insufficient GWG (Table 3).

Table 3.

Health-related characteristics of women across pregnancy (T1–T3)

Health-related characteristics	T1 N = 253 n (%)	T2 N = 237 n (%)	T3 N = 224 n (%)
Smoking during pregnancy	55 (21.7)	44 (17.4)	43 (17.0)
Alcohol consumption during pregnancy	23 (9.1)	19 (7.5)	19 (7.5)
BMI
Underweight/Normal weight	175 (69.2)	-	-
Overweight	55 (21.7)	-	-
Obesity	23 (9.1)	-	-
Gestational weight gain (GWG)
Insufficient	-	-	39 (17.6)
Adequate	-	-	132 (59.7)
Excessive	-	-	50 (22.6)

T1 - First trimester; T2 - Second trimester; T3 - Third trimester; BMI (BMI) was assessed at T1 only. Gestational weight gain (GWG) was assessed at the end of pregnancy. Differences in N reflect loss to follow-up and missing data

Regarding pregnancy outcomes, 1.6% of infants were born preterm, and 26.9% of deliveries were by caesarean section. Small-for-gestational-age newborns accounted for 14.2% of cases, whereas low birth weight and macrosomia were less common, occurring in 1.6% and 8.7%, respectively. Gestational hypertension and gestational diabetes were observed in 8.7% and 7.1% of cases, respectively (Table 4).

Table 4.

Pregnancy outcomes (N = 221)

Pregnancy outcomes	N = 221 n (%)
Preterm birth	4 (1.6)
LBW	4 (1.6)
SGA	36 (14.2)
Macrosomia	22 (8.7)
Caesarean section	68 (26.9)
Postpartum hemorrhage	15 (5.9)
Gestational hypertension	22 (8.7)
Gestational diabetes	18 (7.1)

LBW Low birth weight, SGA Small for gestational age

Mean DASS-21 scores were generally low across all assessment points. The mean depression score ranged from 3.09 (SD = 5.27) to 4.25 (SD = 6.81), the mean anxiety score from 4.32 (SD = 6.11) to 6.48 (SD = 6.58), and the mean stress score from 10.27 (SD = 8.36) to 11.69 (SD = 9.57) across the five time points from early pregnancy to three months postpartum.

The detailed distribution of depression, anxiety, and stress severity categories at each assessment is presented in Supplementary Table 1. Most participants were classified within the normal severity category for all three domains at each assessment. The proportion of women with clinically relevant depressive symptoms (moderate to extremely severe) ranged from 4.5% to 9.9% across the study period. The prevalence of anxiety symptoms ranged from 18.2% to 28.3%, while stress symptoms ranged from 12.3% to 20.3% (Table 5). Overall, anxiety symptoms were more frequently reported than depressive symptoms, whereas stress symptoms showed intermediate prevalence.

Table 5.

Mean score and prevalence of depression, anxiety, and stress symptoms according to DASS-21 across pregnancy and postpartum

	T1N = 253	T2 N = 237	T3 N = 224	PP1 N = 221	PP3 N = 220
Depression score, mean (SD)	4.25 (6.81)	3.28 (5.19)	3.87 (6.45)	3.61 (6.48)	3.09 (5.27)
Anxiety score, mean (SD)	6.37 (6.57)	6.48 (6.28)	6.48 (6.58)	4.89 (6.53)	4.32 (6.11)
Stress score, mean (SD)	11.69 (9.57)	10.33 (7.65)	10.86 (8.77)	11.31 (8.81)	10.27 (8.36)
Prevalence of symptoms (moderate – extremely severe)
Depression	25 (9.9)	15 (6.3)	16 (7.1)	21 (9.5)	10 (4.5)
Anxiety	64 (25.3)	67 (28.3)	62 (27.7)	41 (18.6)	40 (18.2)
Stress	48 (19.0)	30 (12.7)	36 (16.1)	45 (20.3)	27 (12.3)
Comorbid depression, anxiety and stress	16 (6.3)	7 (3.0)	13 (5.8)	14 (6.3)	7 (3.2)

T1 - First trimester; T2 - Second trimester; T3 - Third trimester; PP1 – One month postpartum; PP3 -Three months postpartum. Differences in N reflect loss to follow-up and missing data

Prevalence represents the proportion of participants with moderate, severe, or extremely severe symptoms. Comorbid depression, anxiety, and stress refers to participants who simultaneously reported symptoms of all three conditions

Longitudinal changes in depression, anxiety, and stress scores across the five time points were analyzed using linear mixed-effects models with random intercepts for participants. The first trimester (T1) was used as the reference category. The estimated coefficients (β ± SE) and corresponding unadjusted p-values are presented in Table 6.

Table 6.

Linear mixed-effects model estimates (β ± SE) and p-values for depression, anxiety, and stress scores across five time points. Random intercepts for participants were included

	Depression		Anxiety		Stress
Time point	β (SE)	p	β (SE)	p	β (SE)	p
T1 (ref)	4.25 (0.39)	< 0.001	6.37 (0.41)	< 0.001	11.69 (0.55)	< 0.001
T2	-0.88 (0.40)	0.027	0.15 (0.37)	0.684	-1.30 (0.51)	0.011
T3	-0.24 (0.41)	0.563	0.19 (0.38)	0.617	-0.76 (0.52)	0.144
PP1	-0.40 (0.41)	0.329	-1.32 (0.38)	0.001	-0.19 (0.52)	0.710
PP3	-0.92 (0.41)	0.024	-1.89 (0.38)	< 0.001	-1.26 (0.52)	0.016

T1 (ref) the first trimester was a reference value, SE standard error 

Significant p-values are shown in bold

Figure 3 shows the estimated marginal means (EMM) and 95% confidence intervals for depression, anxiety, and stress scores across the five time points. Depression scores (red line) remained relatively stable across pregnancy and decreased slightly postpartum. Anxiety scores (blue line) decreased notably after delivery, while stress scores (green line) showed reduction in the second trimester and at three months postpartum.

Fig. 3.

Estimated marginal means (EMM) and 95% confidence intervals for depression, anxiety, and stress scores across the five time points

Pairwise comparisons between time points with false discovery rate (FDR) adjustment are presented in Table 7. After correction for multiple testing, statistically significant differences were observed only for anxiety scores. Anxiety levels were significantly lower at postpartum assessments compared to pregnancy. Compared with the first trimester, anxiety scores differed at PP1 (β = 1.32, q < 0.001) and PP3 (β = 1.89, q < 0.001), indicating higher anxiety levels during pregnancy than postpartum. Similar differences were observed when comparing second- and third-trimester measurements with postpartum time points (all comparisons q < 0.001). No statistically significant pairwise differences were observed for depression or stress scores after FDR correction, although some time points differed from the reference in the mixed-effects model.

Table 7.

Pairwise comparisons of depression, stress, and anxiety scores across time points (FDR-adjusted)

	Depression		Anxiety		Stress
Time point	β (SE)	q	β (SE)	q	β (SE)	q
T1-T2	0.88 (0.40)	0.136	-0.15 (0.37)	0.760	1.30 (0.51)	0.081
T1-T3	0.24 (0.41)	0.704	-0.19 (0.38)	0.760	0.76 (0.52)	>0.289
T1-PP1	0.40 (0.41)	0.470	1.32 (0.38)	< 0.001	0.19 (0.52)	0.789
T1-PP3	0.92 (0.41)	0.136	1.89 (0.38)	< 0.001	1.26 (0.52)	0.081
T2-T3	-0.65 (0.41)	0.286	-0.04 (0.38)	0.920	-0.54 (0.52)	0.425
T2-PP1	-0.48 (0.41)	0.400	1.47 (0.38)	< 0.001	-1.11 (0.52)	0.113
T2-PP3	0.04 (0.41)	0.925	2.04 (0.38)	< 0.001	-0.04 (0.52)	0.933
T3-PP1	0.16 (0.41)	0.771	1.51 (0.39)	< 0.001	-0.56 (0.53)	0.425
T3-PP3	0.69 (0.42)	0.286	2.08 (0.39)	< 0.001	0.50 (0.53)	0.431
PP1-PP3	0.52 (0.42)	0.400	0.58 (0.39)	0.194	1.06 (0.53)	0.113

T1 - First trimester; T2 - Second trimester; T3 - Third trimester; PP1 – One month postpartum; PP3 -Three months postpartum. β coefficients represent differences calculated as the first time point minus the second 

SE standard error

P-values were adjusted for multiple comparisons using the Benjamini–Hochberg procedure (q-values); Significant p-values are shown in bold

Multivariable analysis

Detailed results of the univariable analyses are presented in Supplementary Table 2. Statistically significant findings from the univariable analysis, together with the corresponding multivariable regression analysis, are presented in Table 8.

Table 8.

Multivariable Firth penalized logistic regression models for predicting caesarian section, postpartum hemorrhage, small for gestational age, and gestational diabetes

Caesarian section (N = 68)
Patients’ characteristics	Univariable analysis		Multivariable analysis
	OR (95%CI)	p	OR (95%CI)	p
Employment (Employed)	3.64 (1.05–12.60)	0.042	1.33 (0.26–9.30)	0.739
Complications during previous delivery	12.12 (4.32–34.02)	< 0.001	12.47 (4.46–40.34)	< 0.001
Smoking during pregnancy	2.75 (1.41–5.34)	0.003	1.23 (0.34–4.15)	0.740
BMI ≥ 25 kg/m2 T1	2.17 (1.19–3.97)	0.012	4.03 (1.36–13.40)	0.011

Postpartum hemorrhage (N = 15)
Patients’ characteristics	Univariable analysis		Multivariable analysis
	OR (95%CI)	p	OR (95%CI)	p
Anxiety T1	3.04 (1.05–8.84)	0.041	1.69 (0.49–5.51)	0.394
Depression T3	7.13 (1.92–26.39)	0.003	3.88 (0.90-15.37)	0.068
Anxiety T3	3.38 (1.17–9.79)	0.024	2.03 (0.60–6.67)	0.248

Small for gestational age (N = 36)
Patients’ characteristics	Univariable analysis		Multivariable analysis
	OR (95%CI)	p	OR (95%CI)	p
BMI ≥ 25 kg/m2 T1	0.31 (0.12–0.84)	0.022	0.44 (0.15–1.13)	0.090
Excessive GWG	0.27 (0.07–0.92)	0.037	0.44 (0.11–1.35)	0.162

Gestational diabetes (N = 18)
Patients’ characteristics	Univariable analysis		Multivariable analysis
	OR (95%CI)	p	OR (95%CI)	p
Overall health (Good and very good)	0.27 (0.08–0.94)	0.039	0.36 (0.11–1.36)	0.126
BMI ≥ 25 kg/m2 T1	3.11 (1.17–8.28)	0.023	2.63 (0.97–7.18)	0.056

T1 - First trimester; T2 - Second trimester; T3 - Third trimester

OR Odds ratio, 95% CI 95% Confidence interval, BMI Body mass index, SGA Small for Gestational Age, GWG Gestational Weight Gain

Significant p-values in multivariable models are shown in bold

Delivery complications were strongly associated with cesarean section (OR = 12.47, 95% CI 4.46–40.34, p < 0.001), and higher body mass index also increased the odds of cesarean delivery (OR = 4.03, 95% CI 1.36–13.40, p = 0.011). In contrast, smoking during the first trimester (OR = 1.23, 95% CI 0.34–4.15, p = 0.740) and employment status (OR = 1.33, 95% CI 0.26–9.30, p = 0.739) were not significantly associated with cesarean section. Data on delivery complications were missing for 131 participants. Among the 122 women with available data, all had a history of previous childbirth, resulting in no variability in the variable “Any previous delivery”. Therefore, this variable could not be included in the multivariable Firth penalized logistic regression model.

In the multivariable Firth penalized logistic regression model, none of the examined predictors were significantly associated with postpartum hemorrhage: first-trimester anxiety (OR = 1.69, 95% CI 0.49–5.51, p = 0.394), third-trimester depression (OR = 3.88, 95% CI 0.90–15.37, p = 0.068), and third-trimester anxiety (OR = 2.03, 95% CI 0.60–6.67, p = 0.248). In addition, sensitivity analyses were performed using depression and anxiety scores as continuous and ordinal variables. Although some associations were observed in univariable analyses, these did not remain statistically significant after adjustment in multivariable models, confirming the main findings. The results of these analyses are presented in Supplementary Table 3.

In univariable analyses, both pre-pregnancy BMI ≥ 25 kg/m² and excessive gestational weight gain were associated with lower odds of SGA. However, after mutual adjustment, these associations were attenuated and no longer statistically significant: BMI ≥ 25 kg/m² T1 (OR = 0.44, 95%CI 0.15–1.13, p = 0.090), and excessive GWG (OR = 0.44, 95%CI 0.11–1.35, p = 0.162).

Overall health - Good and very good (OR = 0.36, 95%CI 0.11–1.36, p = 0.126) and BMI ≥ 25 kg/m² T1 (OR = 2.63, 95%CI 0.97–7.18, p = 0.056) were not significantly associated with gestational diabetes.

Discussion

In this prospective longitudinal clinical cohort, we examined trajectories of depression, anxiety, and stress across pregnancy and the postpartum period, as well as their associations with adverse perinatal outcomes. Depressive symptoms remained relatively stable across the perinatal period. After correction for multiple testing, only anxiety showed a statistically significant decline after childbirth, whereas changes in stress scores did not reach statistical significance. Most associations between the examined factors and adverse outcomes were not statistically significant or were estimated with limited precision.

In this predominantly low-risk clinical cohort, the prevalence of clinically relevant depressive symptoms ranged from 4.5% to 9.9%, anxiety symptoms from 18.2% to 28.3%, and stress symptoms from 12.3% to 20.3% across the study period, with anxiety symptoms being the most frequent. While descriptive trends suggested some fluctuations in depression and stress across time, these changes were not statistically significant after adjustment for multiple testing. In contrast, anxiety scores showed a clear and statistically significant decline in the postpartum period, particularly at one and three months after delivery.

These findings are in line with previous research indicating a transient increase in stress and a decrease in anxiety symptoms immediately after childbirth [12]. The observed temporal fluctuations likely reflect a complex interplay of biological, psychosocial, and situational factors. A slight increase in stress observed during the first postpartum month may be associated with sleep disturbance, adaptation to the parental role, and the demands of newborn care [31], as well as pronounced hormonal changes occurring in the early postpartum period [32]. Conversely, the decline in anxiety and stress later in the postpartum period may reflect reduced anticipatory fear and uncertainty related to labor and potential obstetric complications. However, some studies have reported increased or new-onset postpartum anxiety [33], possibly due to the substantial adaptive challenges encountered during the early weeks of motherhood.

The relative stability of depressive symptoms in our sample aligns with the findings by Song et al. [30], who reported stable depressive trajectories during pregnancy among women with similar sociodemographic characteristics. One possible explanation is that perinatal depression often follows a more chronic course [34], and may be less sensitive to short-term biological and situational changes than anxiety and stress, which tend to be more responsive to acute fluctuations across the perinatal period. While our findings differ from studies reporting increases in depressive symptoms in late pregnancy or early postpartum [12], such discrepancies may reflect differences in assessment instruments, cultural context, or timing of measurement. Taken together, these findings underscore the importance of continuous mental health screening throughout the entire perinatal period rather than reliance on a single assessment point.

Previous studies conducted in Serbia have primarily focused on the prevalence of perinatal depression, less frequently on anxiety, and none have examined mental health trajectories across the entire duration of pregnancy and the postpartum period. Reported prevalence of depression in the third trimester ranged from 10.6% to 26.3% [21, 22], whereas in our study, it was 7.1%. This discrepancy may partly reflect differences in assessment instruments and case definitions. While earlier studies predominantly used the Edinburgh Postnatal Depression Scale (EPDS), which is designed to detect even mild depressive symptoms, our study applied the Depression, Anxiety and Stress Scale–21 (DASS-21). Furthermore, only moderate-to-severe symptom levels were considered clinically relevant when estimating prevalence, which may have resulted in lower prevalence estimates compared with studies using lower EPDS thresholds. Differences in timing of assessment, sociodemographic characteristics, and socioeconomic context may also have contributed to variability in prevalence estimates. In addition, because the DASS-21 simultaneously captures depression, anxiety, and stress as related but distinct dimensions, it provides a broader perspective on perinatal psychological distress than instruments focused exclusively on depressive symptoms. Overall, our findings provide novel longitudinal insights into perinatal mental health in Serbia and highlight the need for greater methodological standardization.

In the present study, several sociodemographic, obstetric, and psychological factors were associated with adverse perinatal outcomes. Cesarean delivery was more frequent among women older than 35 years, smokers, employed women, those with a history of pregnancy complications, and women with an early pregnancy BMI ≥ 25 kg/m². These findings are consistent with previous studies identifying advanced maternal age, elevated BMI, and smoking as important risk factors for operative delivery [35–38]. A prospective cohort study reported that pre-pregnancy overweight and obesity independently increased the odds of C-section (aOR 1.82–1.85), after adjustment for relevant confounders [39]. However, in our study only early pregnancy BMI ≥ 25 kg/m² and a history of complications in previous pregnancies remained independently associated with cesarean delivery after adjustment for potential confounding variables. Although some studies have not confirmed BMI was independently associated with cesarean birth [40], discrepancies may arise from differences in sample size, population characteristics, or clinical indications for operative delivery.

In univariate analyses, postpartum hemorrhage (PPH) was associated with anxiety in the first trimester and with depressive and anxiety symptoms in the third trimester. Similar associations have been reported previously, and have been hypothesized to reflect potential biological pathways linking psychological distress in late pregnancy with vascular function, inflammatory responses, and uterine contractility [7, 41, 42]. However, after adjustment for potential confounding variables, these associations were no longer statistically significant. This finding suggests that the associations observed in univariate analyses may be partly explained by confounding from other maternal or obstetric factors.

With respect to neonatal outcomes, in univariate analyses the likelihood of delivering a small-for-gestational-age (SGA) infant was lower among women with a BMI ≥ 25 kg/m² at T1 and among those with excessive gestational weight gain, suggesting a potential protective effect. This finding is consistent with studies demonstrating the influence of maternal nutritional status on fetal growth [43–45]. However, these associations were no longer present in the multivariable model. The attenuation of the associations between baseline BMI and excessive gestational weight gain with SGA in the multivariable model may be explained by their interrelated nature and the limited number of outcome events. In addition, gestational weight gain may lie on the causal pathway between maternal characteristics and fetal growth, and its inclusion in the model could have resulted in partial overadjustment. Although some studies emphasize pregestational BMI and gestational weight gain as key predictors of SGA and macrosomia [38], others have reported inconsistent findings [46], suggesting that population characteristics, residual confounding, and limited statistical power may contribute to heterogeneity across studies.

In univariate analyses, gestational diabetes mellitus (GDM) was more common among women with poorer pregestational health and those with a BMI ≥ 25 kg/m². However, these associations were no longer statistically significant after adjustment for potential confounding variables. Although higher pregestational BMI is widely recognized as an important risk factor for gestational diabetes mellitus (GDM) [47, 48], in the present study, pregestational BMI ≥ 25 kg/m² showed only a borderline association (p = 0.056) in the multivariable model. This may be attributed to the relatively low number of GDM cases in our cohort, which likely limited the statistical power to detect independent associations. A similar pattern was observed for self-rated health, suggesting that the lack of statistical significance in adjusted analyses may reflect limited power and overlapping effects of related variables rather than the absence of a true association.

Taken together, these findings suggest that although several sociodemographic, obstetric, and psychological factors were associated with adverse perinatal outcomes in univariate analyses, most did not remain significant after adjustment for potential confounders. This finding highlights the multifactorial etiology of perinatal outcomes and underscores the importance of a comprehensive approach to maternal assessment that integrates obstetric history, physical health, and psychosocial well-being throughout pregnancy. The relatively low incidence of preterm birth and low birth weight may partly reflect characteristics of the study population, which consisted of women with physiological singleton pregnancies attending routine prenatal care at a tertiary obstetric center. As a result, the cohort likely represented a relatively low-risk population.

Strengths and limitations

Despite providing detailed insight into perinatal mental and physical health, several limitations should be acknowledged. First, the study was conducted at a single tertiary care center with a relatively modest sample size, predominantly including primiparous women with physiological pregnancies. Furthermore, while the attrition rate was relatively low (13%), participants lost to follow-up were more likely to be unmarried, which may introduce a minor degree of selection bias, although other baseline characteristics, including psychological scores, were comparable between groups. These factors may limit the external validity and generalizability of the findings to more diverse populations, particularly those with higher obstetric risk.

Second, mental health was assessed using the self-reported Depression, Anxiety and Stress Scale–21 (DASS-21) rather than a perinatal-specific instrument such as the Edinburgh Postnatal Depression Scale (EPDS). Although the DASS-21 allows simultaneous assessment of depression, anxiety, and stress, it was not specifically designed for the perinatal period. In addition, potential overlap between psychological symptoms and somatic experiences related to pregnancy may influence score interpretation. This may lead to overestimation of symptom prevalence or affect observed associations with outcomes. This should be considered when comparing our findings with studies using perinatal-specific screening instruments such as the EPDS. Furthermore, the use of dichotomized DASS-21 variables in regression analyses may have led to loss of information and reduced statistical power, potentially influencing the detection and magnitude of associations between psychological distress and outcomes.

Third, the relatively small number of events for certain outcomes, such as postpartum hemorrhage (PPH) and gestational diabetes mellitus (GDM), may contribute to model instability and wide confidence intervals. To address potential small-sample bias, multivariable analyses were performed using Firth penalized logistic regression; however, the results should still be interpreted cautiously and considered hypothesis-generating. In addition, for outcomes diagnosed during pregnancy, temporal overlap between exposure assessment and outcome onset may limit causal interpretation, and reverse causation cannot be fully excluded.

Fourth, multiple regression models were explored across several outcomes and time points. This multiplicity increases the risk of Type I error and potential false-positive findings.

Finally, longitudinal trajectories of depression, anxiety, and stress were analyzed using linear mixed-effects models, which appropriately account for within-subject correlations across repeated measurements; however, the relatively modest sample size may still limit the precision of estimated trajectory parameters.

Despite these limitations, a key strength of this study is its prospective longitudinal design, with repeated assessments at five time points from early pregnancy through the postpartum period, which enabled a comprehensive evaluation of trajectories of depression, anxiety, and stress across the perinatal period. The integration of psychological assessments with detailed obstetric and metabolic data further allowed exploration of potential associations between maternal mental health and adverse perinatal outcomes in a population that has been relatively understudied in longitudinal research.

Conclusion

This prospective longitudinal cohort study provides insight into trajectories of depression, anxiety, and stress across pregnancy and the early postpartum period in a relatively low-risk population. While depressive symptoms remained relatively stable, only anxiety showed a statistically significant decline after childbirth, whereas no consistent significant changes were observed for stress after correction for multiple testing.

Most associations between psychological factors and adverse perinatal outcomes were not statistically significant or were estimated with limited precision, reflecting the small sample size and low number of events. The observed associations, including those related to cesarean delivery, where early pregnancy BMI ≥ 25 kg/m² and a history of obstetric complications were independently associated with the outcome, should therefore be interpreted with caution.

Given the single-center design and the relatively homogeneous, low-risk cohort, the generalizability of these findings is limited. Overall, the results should be considered exploratory and hypothesis-generating. Future research should focus on larger, multicenter longitudinal studies including more diverse populations and incorporating more comprehensive mental health assessments, combining self-report instruments with structured clinical interviews.

Abbreviations

BMI

Body mass index

CDC

Centers for Disease Control and Prevention

CI

Confidence interval

DALYs

Disability-adjusted life years

DASS-21

Depression Anxiety Stress Scale – 21

EPDS

Edinburgh Postnatal Depression Scale

EPV

Events per variable

GDM

Gestational diabetes mellitus

GWG

Gestational weight gain

LBW

Low birth weight

OR

Odds ratio

aOR

Adjusted odds ratio

PPH

Postpartum hemorrhage

PRAMS-8

Pregnancy Risk Assessment Monitoring System, Phase 8

SDI

Sociodemographic Index

SGA

Small for gestational age

T1

First trimester

T2

Second trimester

T3

Third trimester

PP1

One month postpartum

PP3

Three months postpartum

Authors' contributions

MŽ contributed to the conception of the study and drafted the original version of the manuscript. VMM and ZP contributed to the study design and methodology and critically reviewed the manuscript for important intellectual content. MM, ODžS, DM, and JVM contributed to participant recruitment, data collection, longitudinal follow-up, and quality control of the collected data. JR was responsible for data management, data cleaning, and statistical analysis. NM contributed to the interpretation of the results and critical revision. AN conceived and supervised the study, contributed to the study methodology and statistical analysis, and participated in manuscript writing and critical revision. All authors contributed to the interpretation of the data and approved the final version of the manuscript.

Funding

The authors received no specific funding for this work.

Data availability

The datasets generated and analyzed during the current study are not publicly available due to ethical and privacy considerations but are available from the authors upon reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Ethics Committee of the Obstetrics and Gynecology Clinic “Narodni front” (approval No. 22008-2023-009349). All participants provided written informed consent after being informed about the study objectives. Participation was voluntary, and women could withdraw from the study at any time without consequences.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.