Treatment gap and barriers to access mental healthcare among women with postpartum depression symptoms in Pakistan

Study area

Punjab is the most populous province in Pakistan with a total population of about 127 million. In terms of land area, it ranks as the second-largest province. Lahore is the capital of Punjab and other major cities are Faisalabad, Rawalpindi, Multan, and Gujranwala. The healthcare structure of Punjab comprises primary, secondary, and tertiary/teaching care hospitals (THOS). Primary healthcare institutes consist of 2,500 Basic Health Units (BHUs) and 316 Rural Health Centers (RHCs). RHCs have 6,118 total working beds. Secondary healthcare institutes include 127 Tehsil Headquarter (THQ) hospitals with 7,170 functioning beds, and 26 District Headquarter (DHQ) hospitals have 7,416 total beds whereas 45 THOS are specialized care institutes and comprise 26,293 beds. In 2021, on average there were 321, 351, 118, 50, and 16 deliveries per month per THOS, DHQ, THQ, RHCs, and BHUs, respectively.

Participants

The study included women aged 18–45, those within 6 months of postpartum, those who signed informed consent, and those with an EPDS score of more than or equal to ≥13. Women were excluded from the study if they experienced fetal demise or declined participation.

During the screening phase to identify women with PPD symptoms, data collectors approached a total of 3,340 individuals. While screening 3,340 individuals, 41 refused to further participate, of the rest of the 3,299 individuals, 79 out of 3,299 women didn’t answer all items of the EPDS questionnaire. 3,220 women entirely completed the EPDS and 1,503 out of 3,220 had 13 or above scores. Those women (n = 1,503) who had positive PPD symptoms (13 or above scores, now met the inclusion criteria) were study participants and they were interviewed about their sociodemographic and obstetric characteristics, perceived social support, help-seeking behavior, their perceived need for care in the past six months, and barriers they face to access care.

The duration of the interview was between 18 to 20 min. Compensation was not provided to participants. This decision was made to maintain integrity by minimizing potential biases that could be introduced by offering compensation. Although the interviews proceeded smoothly, occasional interruptions occurred due to background noise and, in some cases, the presence of respondents’ family members caused disturbances.

Data collection instruments

In this study, socioeconomic and obstetric characteristics were considered basic demographic characteristics and independent variables, while social support, barriers to access to mental health care, and treatment gap were considered intervening variables.

The data collection instruments consisted of the following sections:

General help-seeking questionnaire (GHSQ)

Help-seeking behavior was determined with the help of GHSQ (Wilson et al., 2005). The purpose of this scale is to measure the help-seeking intentions of individuals. The scale provides information regarding the help-seeking behavior of participants by formal and non-formal means. This is a nine-item 7-point Likert scale and categorizes help sources into two main categories: formal and informal (partners, friends, family, mental health professionals, physical health professionals, phone helpline, teachers, community, and no one) and problem type (personal-emotional and suicidal ideation) (Wilson et al., 2005). The GHSQ, when assessed in English, has demonstrated satisfactory psychometric characteristics, with internal consistency ranging from 0.67 to 0.90 and a test-retest reliability of 0.86.

To enhance the usefulness of the GHSQ, developers have suggested researchers to adapt the questionnaire by integrating specific types of sources of help relevant to their particular study (Wilson et al., 2005). In this study, the scale was adapted to assess actual help-seeking behavior rather than intended behavior, and the response categories were reduced to yes/no. Upon recommendations of researchers and to keep it aligned with the Pakistani cultural setting, this scale was adapted as follows: formal help-seeking sources were

1. General health professionals, 2. mental health specialists (e.g., psychiatrists, psychologists), 3. others (traditional healers, and religious scholars/influencers), whereas non-formal sources for help-seeking involved parents, spouses, friends, and other relatives. Reliability of GHSQ was tested and a cronbach’s alpha value of 0.72 was obtained.

Translation of questionnaires

For the translation of questionnaires, we followed the cross-cultural adaptation process (Beaton et al., 2000). We gathered questionnaires in English and translated them into Urdu.

First, a pharmacy post-graduate student and an individual with no medical background, both native Urdu speakers who are also fluent in English translated questionnaires into Urdu. Each of them produced a translation along with a report (mentioned uncertainties and an explanation of their choices) in written form. An independent researcher then joined and resolved any inconsistencies to produce a single forward-translated Urdu version. After this, a professional translator specializing in medical translation and a pharmacy post-graduate not previously involved in the initial translation translated this Urdu version back into English (one translation by each). Once again, an independent researcher compiled these back-translated versions and produced a pre-final version.

Questionnaires validation

Procedure

A questionnaire-based multicenter cross-sectional study was carried out among women aged 18–45 years, during the first 6 months postpartum, who were willing to participate and provide informed consent, and had a PPD screening score of ≥13 on the EPDS. EPDS has a sensitivity of more than 80% in various cultural settings (Atuhaire et al., 2023; Cox, Holden & Sagovsky, 1987; Husain et al., 2014; Shoaee et al., 2019). This demonstrates that it is an acceptable way to correctly identify women having PPD. The study was carried out from January to June 2023. The screening and data collection process was conducted by a team consisting of two post-graduate pharmacy students and one post-graduate psychology student. Using stratified random sampling, the sample was chosen. First, we selected five major cities in Punjab, Pakistan: Faisalabad, Rawalpindi, Lahore, Multan, and Gujranwala. Next, we randomly selected seven public or private hospitals in each city, and finally, a total of 3,220 were screened for PPD symptoms from five cities.

We obtained the approval of the study from the Institutional Review Board, Government College University Faisalabad, Pakistan before initiating the study (Approval number: GCUF/ERC/196). The study met the guidelines outlined in the Helsinki Declaration.  All patients provided informed verbal consent instead of written because the study did not involve any clinical intervention the patient’s participation was clearly at minimum risk. Face-to-face administration of survey questionnaires was conducted in the native language, Urdu, using validated tools.

The Statistical Package for Social Sciences (SPSS, version 25.0) was used for data analysis. Descriptive statistical methods including frequency, percentages and Mean ± S.D were used to analyze the demographic and obstetric features of the study participants. The Mann–Whitney U and Kruskal–Wallis test were used to compare the mean score of treatment gap, social support and barriers across different demographic and obstetric characteristics. Scatter plot matrix with fitted linear regression line was used to present the relationships between treatment gap, stigma barriers, instrumental barriers, social support, and intensity of depression. Finally, the classification and regression tree (CRT) based decision tree method was used to classify the importance of different variables for the intensity of PPD.

Sociodemographic variables

The sociodemographic characteristics of the women in Table 1 demonstrated that approximately 49% of women with PPD symptoms were between the ages of 26 and 35. The majority of the women (89%) resided in urban areas. Education levels varied among the women, with 36% having attended high school, and 14% having no formal education. Employment status indicated that the majority of women were not employed (90%). Approximately 53% of women had more female babies. Regarding marriage duration, 60% of the women had been married for more than 6 years. Household income analysis revealed that approximately 60% of the women reported a monthly income between 26,000 PKR (approx. 90$) to 50,000 PKR (about 180$). In 2023, Pakistan had a GDP per capita of $1,568 (Hussain, 2023). Living arrangements indicated that 71% of the women were living with their spouses’ extended family.

Obstetric variables

The frequency and corresponding percentages of obstetric variables are presented in Table 2. Among the women, 77% had a gestational age of 9 months, indicating a full-term pregnancy. Approximately 61% of the newborns were female. The majority of the women (72%) gave birth in a government hospital. Only 17% of the women reported that the current pregnancy was their first. Additionally, 19% of the women had experienced a previous miscarriage.

Social support

The mean score for perceived social support was calculated by doing a sum across all 12 items on MSPSS and dividing by 12. Any mean score ranging from 3 to 5 is considered as moderate social support. The mean score for social support was calculated to be 4.376 which falls under the category of moderate social support. Table 3 represents the comparison of mean score of social support, depression, treatment gap, stigma barriers, attitudinal barriers, and instrumental barriers across all categories of sociodemographic and obstetric variables. However, the distribution of mean score for social support was found to vary significantly across different categories of age groups, educational status, gender of children, gender of the newborn, and first delivery status (p < 0.05), suggesting that there is a significant association between age groups, educational status, gender of children, gender of the newborn, and first delivery status and mean social support scores.

Depression

The distribution of mean depression scores was not the same across various categories of sociodemographic and obstetric factors including education level (p < 0.001), the gender of children (p < 0.001), living arrangements (p = 0.002), gender of the newborn (p = 0.03), type of delivery (p = 0.012), and birth location of the newborn (p = 0.004) as shown in Table 3. These findings suggest the association of these sociodemographic and obstetric factors (education level, the gender of children, living arrangements, gender of the newborn, type of delivery, and birth location of the newborn) with depression scores.

Treatment gap

The analysis revealed a significant treatment gap in accessing mental healthcare services among women with PPD symptoms. Table 4 elaborated that out of 1,503, only a small proportion of women sought medical attention for their PPD symptoms, with 33 women (2%) seeking specialized care from a mental health provider and 75 women (4%) were seeking help from a general physician. Even though a significant majority of the women 1,215 (81%) reported a perceived need for any type of mental health care, only 108 out of 1,215 were those who sought treatment either from general health professionals or mental health professionals. 91.23% of women with potentially high PPD symptoms who perceived the need to get treated could not seek any care from healthcare providers reflecting a high treatment gap. Moreover, of those who didn’t perceive the need for any type of mental health care (21%), none of them used care from any healthcare professionals.

Table 3 demonstrates that the distribution of the treatment gap varied across various groups of education (p = 0.001), the gender of children (p = 0.035), living arrangements (p < 0.001), gender of the newborn (p < 0.001), type of delivery (p = 0.012), birth location of the newborn (p = 0.027) and first delivery status (p < 0.001).

Barriers to access to mental healthcare

The distribution of stigma barriers, attitudinal barriers, and instrumental barriers across various sociodemographic and obstetric factors was assessed. The distribution of these barriers was not the same across all categories of age groups, educational status, and birth location of the newborn (p < 0.05). Variables, including age, education, and birth location of the newborn, were significantly associated with stigma, attitudinal and instrumental barriers scores.

Gender of children was significantly associated with mean attitudinal and instrumental barrier scores. Moreover, living arrangements were significantly associated with mean instrumental barrier scores.

Figure 1 depicts a decision tree for classifying PPD scores. Out of 1,503 women (node 0), 13% of women who had mean instrumental barrier scores >14.5 had the highest mean depression scores (22.354) whereas 87% of women who had mean instrumental barriers scores ≤14.5 had comparatively lower mean depression scores; however, 26.5% who had social support scores ≤4.25 (node 3), and instrumental barrier scores ≤14.5 had mean depression scores higher (19.401) than the other group (18.076) having mean social support scores > 4.25 (node 4).

Out of these 26%, 8% (node 6) of the women who had stigma barrier scores ≤14.5, social support scores ≤4.25; however, instrumental barrier scores >19.5 had mean depression scores of 20.575 which are higher than 18.66% of the women who had stigma barrier scores ≤19.5.

Importance of each variable (instrumental barriers, stigma barriers, attitudinal barriers, social support, treatment gap, gender of children, and age, in relation to depression) identified by CART analysis is presented in Fig. 2. Among these variables, instrumental barriers were considered the most important, based on their highest normalized importance value. The instrumental barrier scores are a highly crucial feature, suggesting that women with more barrier scores tend to have higher mean depression scores.

Figure 3 illustrates a scatter plot matrix to visualize the strength and direction of the relationship between variables. A negative relationship between social support and depression scores can be seen, as social support increases, mean depression scores decrease. There is also a negative relationship between social support and barrier (stigma, instrumental, attitudinal) scores, as mean social support scores increase, barrier scores decrease. A strong positive association between stigma and attitudinal barrier scores underscores the relationship between elevated stigma and higher levels of attitudinal barriers. Moreover, a moderate positive association between stigma and instrumental barrier scores can be visualized in the scatterplot matrix, which reinforces that increased stigma scores are linked to increased instrumental barrier scores. A weak positive relationship can be visualized between barriers (stigma, instrumental, attitudinal) and depression.