Study design

This is a prospective cohort study comparing two cohorts separated by a 20-year interval, respectively, in two municipalities with similar socioeconomic characteristics in the state of São Paulo, Brazil: Jundiaí Cohort Study (USP-MatStress): From an initial sample of 1182 women with gestational age ≤ 16 weeks who received prenatal care between September 1997 and August 2000 in 12 health units and five hospitals in the Municipality of Jundiaí, São Paulo, Brazil, 865 were followed quarterly in a cohort study until the birth of their children in one of the five hospitals in Jundiaí [4,5]. The Araraquara Cohort Study: The sample included women with a gestational age ≤ 19 weeks who received prenatal care in the 37 Basic Health Units and the Special Health Service (SESA) of the municipality of Araraquara, São Paulo, Brazil. The pregnant women participating in the mental health assessment, part of the Araraquara Cohort Study, were followed quarterly throughout prenatal care until the birth of their children from 05 January 2017 to 30 December 2024 [32].

The study conducted in Araraquara was approved by the Research Ethics Committee with Human Subjects at the School of Public Health, University of São Paulo, prior to data collection, under the protocol number CAEE: 59787216.2.0000.5421. The study conducted in Jundiaí was also approved 289/98, and informed written consent was obtained from all participants. The Ethical Committees of the School of Public Health, University of São Paulo, and the Health Secretariat of Jundiaí, SP approved the protocol.

Mental health changes during pregnancy

Three standardized questionnaires were used to assess the mental health of pregnant women, measured at three visits during pregnancy (gestational age ≤ 16 or 19 weeks, 20–26, and 30–36 weeks). These included the general health questionnaire (GHQ), which screens for non-psychotic psychological distress [33,34], the State-Trait Anxiety Inventory (STAI), composed of the State Anxiety Inventory (SAI) and the Trait Anxiety Inventory (TAI), which assess both transient (state) and enduring (trait) anxiety [4,5], and the Perceived Stress Scale (PSS), which measures the individual’s perception of stress in daily life [35].

Predictors

Several factors were considered as predictors for the study, including socioeconomic and demographic characteristics such as age, race (white, black, brown, yellow), marital status (single, separated/widowed, stable union), monthly family income in Brazilian minimum wages, per capita income, occupation, working status during pregnancy, the reason for not working (unemployed, maternity leave, others), the time spent working outside the home, and the hours worked both outside and inside the home. Additionally, the number of people per room and the level of education were considered.

Housing conditions included the type of house (owned, owned but not yet paid off, granted, other condition), the material of the house walls (brick), the presence of sewage, the number of rooms, and the possession of items such as a refrigerator, car, motorcycle, and access to piped water.

Obstetric history included the number of previous pregnancies, the time since the last delivery, the occurrence of previous abortions, previous stillbirths, and previous neonatal deaths. Pregnancy risk factors considered included morbidities such as hypertension, diabetes, rubella, urinary infection and pyelonephritis, syphilis, gonorrhea, cervicitis, vaginitis, tuberculosis, AIDS, and hepatitis. These variables were measured at three visits during pregnancy.

Statistical analysis

Descriptive analyses included calculating means, standard deviations, medians, interquartile ranges, frequencies, and percentages of the studied variables distributed between the Araraquara and Jundiaí groups. Line graphs were used to visualize the evolution of mental changes over the visits. Due to strong asymmetry, the variables age, number of people per room, family income, and per capita income were included in the model on a logarithmic scale.

Models for unbalanced data with random effects in the panel were used [36]. as detailed below, including the effects of significant interactions between location and visit:

where: is the observed score for individual , in location (1 = Araraquara, 2 = Jundiaí), at visit , ; is the observed value for covariate l , for individual , in location e visita ; : constant. : is the parameter relative to the effect of covariate ; is the effect of location ( : is the effect of visit ( is the interaction effect between location and visit ( e equal to zero); is the random error.

To accommodate the longitudinal design and repeated measures structure, we applied random-effects panel models (mixed models), allowing individual-specific intercepts and controlling for time-invariant unobserved heterogeneity. This approach suits unbalanced panels where not all individuals are observed in all visits. The dependent variables (GHQ, SAI, TAI, and PSS) were modelled as continuous outcomes to preserve variability and avoid loss of information that can result from categorization. This decision aligns with recommendations in epidemiological modelling. [36]. The model fit was performed following a step-by-step strategy, an iterative method that selects and removes independent variables, keeping those that presented a significance level of p < 0.02. All analyses were performed in Stata, version 18 (College Station, TX: Stata Corp LLC).

Descriptive statistics of mental health scales

Table 1 presents the descriptive statistics of the mental health scales (GHQ, SAI, TAI, and PSS) across the three prenatal visits. At baseline (Visit 1), GHQ scores were higher in Araraquara (mean = 4.00) than in Jundiaí (mean = 2.78), indicating worse mental health status in the former. Similarly, SAI, TAI, and PSS scores differed between the cohorts. In both cities, scores decreased in subsequent visits. By Visit 2, GHQ scores dropped to 3.09 in Araraquara and 2.31 in Jundiaí. At Visit 3, this downward trend persisted, with GHQ scores reducing further to 2.89 in Araraquara and 2.46 in Jundiaí.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics of mental health assessment scales in the Jundiaí (1997-2000) and Araraquara (2017-2020) cohorts.

https://doi.org/10.1371/journal.pone.0330921.t001

Evolution of average mental health scores

Fig 1. Displays the mean scores for each mental health instrument (GHQ, SAI, TAI, and PSS) across the three visits. Table 2 shows the results of model adjustments for location and visit effects. All scales showed significant decreases over time: scores were significantly lower in Visit 2 and Visit 3 compared to Visit 1 (p < 0.001 for all instruments). Additionally, scores were consistently lower in Jundiaí than in Araraquara (p < 0.05), suggesting better mental health in the former. For PSS, the location difference was only significant in Visit 3, where Jundiaí had higher scores than Araraquara (p < 0.05). Significant interaction effects between location and visit were observed, especially in Visit 3. These include GHQ (Coef. = 0.729), SAI (Coef. = 1.667), TAI (Coef. = 0.878), and PSS (Coef. = 1.940), all with p-values ≤ 0.001, indicating that the decline in scores over time was not uniform between cities.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Parameter estimates of mental health assessment models in the Jundiaí (1997-2000) and Araraquara (2017-2020) cohorts.

https://doi.org/10.1371/journal.pone.0330921.t002

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Profiles (mean and standard deviation) of the scales GHQ, SAI, TAI, and PSS over three visits in the Jundiaí and (1997-2000) and Araraquara (2017-2020) cohorts.

https://doi.org/10.1371/journal.pone.0330921.g001

Factors associated with mental health changes

Table 3 presents the raw estimates from the mixed model, while Table 4 shows the adjusted effects considering all covariates. Several factors were associated with worse mental health. Being single was associated with higher GHQ (Coef. = 0.404, p = 0.019), SAI (Coef. = 1.623, p = 0.004), and TAI (Coef. = 1.427, p = 0.022) scores. Separated or widowed women also had higher scores across these scales, with notable effects in GHQ (Coef. = 1.120), SAI (Coef. = 3.961), and TAI (Coef. = 3.225). A higher number of people per room, measured on a logarithmic scale, was positively associated with GHQ (Coef. = 0.281, p = 0.042), SAI (Coef. = 1.138, p = 0.013), and PSS (Coef. = 0.946, p = 0.012) scores. Higher education (high school or more) was associated with lower anxiety scores: SAI (Coef. = −1.416, p = 0.016) and TAI (Coef. = −1.910, p = 0.006).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Parameter estimates of model (1) for factors associated with mental health changes (marginal models).

https://doi.org/10.1371/journal.pone.0330921.t003

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Multivariable panel data models for factors associated with mental health changes in the Jundiaí (1997–2000) and Araraquara (2017–2020) Cohorts.

https://doi.org/10.1371/journal.pone.0330921.t004

Pre-gestational weight was associated with better mental health, particularly lower SAI (Coef. = −0.0897, p = 0.025) and TAI (Coef. = −0.0938, p = 0.023) scores, while current weight showed a borderline positive association with anxiety (Coef. = 0.0671, p = 0.087). Age, also log-transformed, was associated with lower scores in TAI (Coef. = −2.652, p = 0.031) and PSS (Coef. = −2.066, p = 0.020). Obstetric history also mattered: the number of previous pregnancies was positively associated with higher scores in GHQ (Coef. = 0.235, p = 0.002), SAI (Coef. = 0.754, p = 0.002), TAI (Coef. = 0.985, p < 0.001), and PSS (Coef. = 0.453, p = 0.029).

Regarding maternal morbidities, hypertension was linked to worse GHQ scores (Coef. = 0.393, p = 0.032). Urinary infections and pyelonephritis were associated with increased scores in GHQ (Coef. = 0.286), TAI (Coef. = 0.862), and PSS (Coef. = 0.931). Cervicitis and vaginitis during pregnancy were linked to higher GHQ (Coef. = 0.290), SAI (Coef. = 0.677), and TAI (Coef. = 0.754) scores. Tuberculosis showed a strong association with increased GHQ (Coef. = 2.837, p = 0.010) and TAI (Coef. = 6.989, p = 0.033) scores.

After adjusting for all covariates, mental health scores remained significantly lower in Visits 2 and 3 across all instruments compared to Visit 1. For GHQ, the coefficients were −0.875 (p < 0.001) in Visit 2 and −1.053 (p < 0.001) in Visit 3. For SAI, the coefficients were −2.265 and −2.607, both with p < 0.001. For TAI, the reductions were −2.045 and −3.825, and for PSS, −1.014 and −1.959, all with p < 0.001. Jundiaí continued to show significantly lower scores compared to Araraquara in GHQ (Coef. = −1.357, p < 0.001), SAI (Coef. = −4.466, p < 0.001), TAI (Coef. = −2.540, p < 0.001), and PSS (Coef. = −0.983, p = 0.045).

Socioeconomic conditions and austerity

Socioeconomic factors such as lower income, lower educational level, and higher household density were significantly associated with higher levels of psychological distress symptoms, anxiety, and perceived stress in pregnant women from Araraquara. These results are consistent with previous studies that identified low income and lack of social support as critical predictors of mental disorders during pregnancy [7,8,23,25,30,37,41]. The presence of a partner and a robust social network are important protective factors against these disorders. FARIAS et al (2021) also observed that maternal mental health is closely linked to socioeconomic conditions, corroborating our findings [14]. Brazil has changed over the past twenty years and may have increased the vulnerability of low-income pregnant women. Rising inequality, austerity, and limited social policies have created structural barriers to mental well-being [16,42]. Second, the effects of accelerated urbanisation, such as overcrowded housing, reduced green spaces, and fragmented social networks, may increase the risk of common mental disorders [42,41]. Third, the underdevelopment of early mental health screening programs and psychosocial support within prenatal services may have led to delayed diagnoses and care [43,44].

Urbanization, isolation, and loneliness

Urbanisation and globalization, driven by liberal economic policies, have transformed Brazilian metropolises in recent decades, resulting in significant growth. However, these changes have also led to an increase in loneliness and social isolation, which are critical factors profoundly affecting mental health. Recent studies have highlighted that rapid urbanization and the social changes resulting from globalization contribute to loneliness and stress, exacerbating mental health problems in large cities [42]. Loneliness is a significant risk factor for various mental illnesses, including symptoms of depression and anxiety, and is exacerbated by the urban environment, which often promotes isolation [45].

Climate change, COVID-19 pandemic, and perinatal morbidity

Health complications during pregnancy, such as urinary infection, pyelonephritis, cervicitis, vaginitis, tuberculosis, and hypertension, were associated with worse mental health scores. These results are in line with existing literature showing that health complications can exacerbate stress and anxiety during pregnancy [8,20,26,41]. Moreover, gestational hypertension was associated with higher GHQ scores, highlighting the interaction between physical and mental health. GOMES et al (2023), also reported that pregnant women with health complications are more likely to develop mental disorders [15]. The COVID-19 pandemic exacerbated mental health problems globally, with increased stress, insomnia, anxiety, and depression. The COVID-19 pandemic further exacerbated mental distress, and emerging evidence suggests its psychological effects may persist in the post-pandemic era [46]. Additionally, i recent years, the planet has experienced climate changes that have significant effects on mental health. Extreme climatic events such as floods and storms, and chronic stresses such as extreme heat and drought affect mental health, leading to anxiety disorders, depression, and post-traumatic stress disorder [47].

The findings of this study underscore that the integration of validated instruments GHQ, STAI, and PSS facilitates a multidimensional assessment of psychological well-being. While each instrument evaluates specific domains, they have been shown to complement one another in perinatal mental health research, providing a more robust picture of emotional distress [48,49], and the urgent need for targeted, evidence-based interventions aimed at pregnant women in vulnerable conditions. Public policies should prioritise economic security, adequate housing, and equitable access to quality education as foundational determinants of maternal mental health. Moreover, it is imperative to institutionalize the integration of mental health screening and care into routine prenatal services, ensuring early identification and timely treatment of psychological distress and comorbid conditions. Building upon international experiences, Kinser et al (2018) highlight the effectiveness of structured screening tools and intervention protocols for perinatal distress, stress, and anxiety symptoms which could be adapted and scaled within the Brazilian Unified Health System. Such measures not only improve maternal well-being but also enhance fetal and neonatal health outcomes, reinforcing the need for a comprehensive, intersectoral response to maternal mental health [27].

This study has some limitations. The analysis was based on data from two specific cohorts, which may limit the generalizability of the findings to other populations. Additionally, unmeasured variables, such as specific forms of social support and exposure to stressful life events, may have influenced the outcomes. Another important methodological limitation is the use of self-reported screening tools (GHQ, STAI, and PSS) to assess mental health symptoms. These instruments are designed to detect nonspecific psychological distress and do not replace clinical diagnosis or psychiatric evaluation. Therefore, the results should be interpreted with caution regarding the identification of mental disorders. Future studies should consider incorporating validated diagnostic assessments, as well as additional psychosocial and environmental variables, to enable a more comprehensive understanding of mental health during pregnancy. Existing literature, including the work of LANCASTER et al. (2010), emphasizes that a more holistic approach can provide deeper insights into the complexity of maternal mental health [25].