Study design and setting

A cross-sectional online study was conducted in Khulna, Chittagong, Mymensing, Rajshahi, Barisal, Rangpur, Sylhet, and Dhaka divisions in Bangladesh from April 1–30, 2020, corresponding to the mandatory lockdown period imposed in different parts of the country.

Bangladesh is a developing nation in South-East Asia that became a separate political and economic entity only 50 years ago. Originally part of the British Raj and then pre-Independence India, Bangladesh became an eastern province of Pakistan and gained independence in 1947 following its ‘Great Liberation War’ and was named Bangladesh [20]. Bangladesh covers 147,570 square kilometres and has a population of roughly163 million, with India and Myanmar at its borders and majority are Muslims (89%). About 38.2% of the population residing in urban areas, including the capital Dhaka, which has an estimated population of 21 million people and growing [21]. The country has about 220 psychiatrists and 50 trained clinical psychologists serving the whole nation [20] and has absorbed close to a million refugees from Myanmar in recent years.

Sampling

Invitations to participate were sent out online through social media platforms (e.g., Facebook, Google Plus, LinkedIn, and Twitter) because traditional face-to-face interviews were not possible due to the lockdown. The use of social media platforms also enabled the researchers to reach target respondents living in different parts of Bangladesh. Residents were also asked to share the e-link with their friends and networks to increase the reach. A snow-ball sampling technique was used for collecting information from participants.

To be eligible for participation in this study, the respondents had to reside in Bangladesh, be able to provide online informed consent, and be 18 years and over at the time of data collection. Informed consent was obtained through an online preamble before the respondents began the questionnaire. The participants were assured of the anonymity and confidentiality of their responses and information provided as well as their freedom of choice of participation. The study was approved by the Institutional Review Board of Dr. Wazed Research and Training Institute, Begum Rokeya University, Rangpur (#BRUR/DWRTI/a.n.004). All procedures were in accordance with the 1964 Helsinki Declaration of Helsinki as revised in Fortaleza.

Outcome Measures

The mental health symptoms of depression, anxiety, and stress were the outcome measures. The Depression, Anxiety and Stress Scale-21 Items (DASS-21) is a set of three self-report scales designed to measure the emotional states of depression, anxiety, and stress and were calculated based on a previous study [22]. These instruments are often used to assess subjective complaints and do not directly imply that respondents have a discrete diagnosis as classified in the diagnostic and statistical manual of mental disorders (DSM) and international classification of diseases and related health problems (ICD) [23]. However, their use in this study was predominantly aimed at assessing the perceived severity of symptoms related to depression, anxiety, and stress [24].

Depression subscale

Responses to each item were rated from 0 (never) to 3 (almost always). Items 3, 5, 10, 13, 16, 17, and 21 (see supplementary Table 1) were classified as the depression subscale, and the total depression subscale score was multiplied by 2 to calculate the final score and then divided into normal (0–9), mild depression (10–13), moderate depression (14–20), severe depression (21+). The validity of the tool was tested, and the Cronbach’s alpha coefficient ranged from 0.75 to 0.77 among the depression subscale, indicating a satisfactory level of reliability.

Anxiety subscale

Items 2, 4, 7, 9, 15, 19, and 20 (see supplementary Table 1) formed the anxiety subscale, and the total anxiety subscale score was multiplied by 2 to calculate the final score and then divided into normal (0–7), mild anxiety (8–9), moderate anxiety (10–14), severe anxiety (15+) [18]. The Cronbach’s alpha coefficient testing the validity of the anxiety subscale showed scores ranging from 0.65 to 0.77, indicating acceptable internal consistency.

Stress subscale

Items 1, 6, 8, 11, 12, 14, and 18 (see supplementary Table 1) were classified as the stress subscale. The total stress subscale score was multiplied by 2 to calculate the final score and then divided into normal (0–14), mild stress (15–18), moderate stress (19–25), severe stress (26+), and extremely severe stress (35–42) [22]. The Cronbach’s alpha coefficient of stress subscale scores ranged from 0.78 to 0.81, indicating an acceptable level of internal consistency.

Statistical analysis

Data were analysed using the STATA/MP Version.14.1 (Stata Corp 2015, College Station, TX, USA). Descriptive statistics using frequency tabulations were used to present the sample characteristics. The prevalence and 95% confidence intervals (CIs) for normal, mild, moderate, and severe levels of depression, anxiety, and stress were calculated. The association was further tested by odds ratios (OR) using univariate and multiple multinomial logistic regression analyses to identify factors associated with symptoms of depression, anxiety, and levels of stress. In the multiple multinomial logistic regression analyses, four-stage modeling was employed. In the first stage, the demographic factors were entered into the model. We conducted a manually executed elimination method to determine factors associated with symptoms of depression, anxiety, and levels of stress. The significant factors in the first stage were added to the household factors in the second stage model; this was then followed by the elimination procedure. We used a similar statistical approach for compliance with public health and health condition/factors in the third and fourth stages, respectively. Associations were presented as unadjusted OR for all explanatory variables and then adjusted OR (AOR) with their 95% CI for the variables retained in the final step. The level of statistical significance was set at p < 0.05.

Spatial analysis

We conducted spatial distribution for the three mental health subscales. A series of maps were prepared using ArcGIS Desktop 10.8 [25]. The average level of depression, anxiety, and stress for the first-level administrative unit of Bangladesh (division) was calculated based on the identification of factors through map comparisons and regression analysis for mild, moderate, and severe levels of the mental health variables. In the maps, the adjusted odds ratios (AOR) for each level was categorized into five quantiles and presented using graduated colour symbols.

Sociodemographic characteristics of the study sample

Table 1 shows the study sample characteristics. Of the 10,900 participants who completed the questionnaire, data for 10,660 adult respondents from the 8 divisions in Bangladesh who provided responses for the outcome measures of interest were included in the analysis. There were 5,238 males (49.1%), mostly young (18–37 years: 8,466, 79.4%), married people (5,332, 50.3%), who had a university degree or higher (6,233, 58.5%), and many lived in Dhaka division (5,260, 49.6%) with their families (8,533, 80.4%) or with up to 3 people (9,328, 87.9%) in a house and were working full time or part-time (5,974, 56.0%) during the pandemic.

There were also 343 individuals (3.2%) with confirmed cases of COVID-19 and 832 individuals (7.8%) who tested negative to COVID-19. Although majority of the respondents reported that they enforced measures in their homes to protect their families (9,705, 91.5%) and had quarantine experience (8,454, 79.7%), more than two thirds (~78% each) did not comply with the public health advice asking people to avoid public transport, handshaking, large gatherings, and to always wear facemasks when going out. These were put in place to contain the spread of the disease. More than two thirds of the participants had at least one symptom of COVID-19 a couple of weeks before data collection, especially persistent fever and cough (9,727, 91.7%) and difficulty breathing (9,207, 86.8%). Additional demographic and epidemic-related characteristics are presented in Table 1.

Prevalence of depression, anxiety, and stress symptoms

The prevalence and 95% CIs for the different levels of the three mental health conditions that were examined in this study are shown in Figure 1. The prevalence was higher for anxiety 86.9% (9,267 participants total, including 1,440 participants [13.5%, 95%CI, 12.4–13.7%] with mild anxiety and 6,310 participants [59.2%, 95%CI, 58.3–60.1%] with severe anxiety), followed by depression 65.0% (6,822 participants total, including 1,563 participants [14.7%, 95%CI, 14.0–15.4%] with mild depression and 1,552 participants [14.6%, 95%CI, 13.9-15.3%] with severe depression), while 50.5% (95% CI, 28.8%–29.6%) reported stress symptoms including severe stress (1,684 participants, 15.8% [95% CI, 15.1–16.5%]).

Factors associated with symptoms of depression

The results of the unadjusted analysis of demographic and COVID-19 related variables are presented in supplementary Tables 1–3, for depression, anxiety, and stress, respectively. Table 2 presents the multivariate analysis of factors associated with depression in this cohort. In the multivariable analysis, being married; having lower than a postgraduate degree; living alone or living in shared accommodation; living in the Sylhet, Dhaka, Chittagong, Rajshahi, and Rangpur divisions; or having experienced any COVID-19 related health symptoms in the couple of weeks before data collection were associated with symptoms of depression at all levels. Figure 2 presents the adjusted odds ratios for depression symptoms among the respondents by region. The map in Figure 2 indicates that significantly high odds of depression occurred in all eight regions of Bangladesh, but residents in Rangpur, Sylhet, and Chittagong (darkest brown colour) reported the highest odds of severe depression. Individuals who were tested for COVID-19 also had remarkably higher levels of mild–severe symptoms of depression compared to those that had not been tested. In addition, female participants, those who were divorced or separated, residents of Barisal or Mymensing divisions, and those who traveled by public transport displayed a higher odds of moderate–severe symptoms of depression. Individuals with confirmed COVID-19 had at least a 50% higher odds of moderate–severe depression symptoms compared to those not tested for the disease.

Factors associated with symptoms of anxiety

Table 3 presents the multivariate analysis of factors associated with anxiety. Females, those who were divorced or widowed, those who lived in shared accommodation during the lockdown, individuals who experienced self-quarantine, and those who experienced fever and cough a couple of weeks before data collection were more likely to experience mild-severe symptoms of anxiety, while older participants experienced a lower risk of anxiety at all levels. The distribution by region (Figure 3) revealed greater than a threefold increase in the odds of feeling severely anxious among respondents residing in four regions. Compared to those from Khulna division, participants who lived in other divisions, particularly Chittagong [aOR, 6.26, 95%CI, 4.40–8.90], Mymensing [aOR 5.26, 95%CI, 3.04–9.12], and Rangpur [aOR, 5.86, 95%CI, 3.64–9.42] were more likely to feel severely anxious (see Figure 3) as well as those who lived alone [aOR, 1.70, 95%CI, 1.35–2.16] during the lockdown. Those with confirmed or suspected cases of COVID-19 were about two times more likely to experience severe symptoms of anxiety compared to those who were not tested. Other symptoms of COVID-19 were also significantly associated with some degree of anxiety in this study.

Factors associated with symptoms of stress

The factors associated with stress in the multivariate analysis are shown in Table 4, while the regional distribution is shown in Figure 4. Those who lived in Dhaka, Rangpur, and Silhet experienced severe levels of stress during the pandemic (Figure 4). Females, those living outside of Khulna division, participants with lower than postgraduate education, those older than 27years, were married, lived alone, and had experienced self-quarantine and any of the health symptoms in this study (except for fever), were more likely to experience symptoms of stress at all levels. Participants who were divorced/widowed [aOR, 1.54, 95%CI, 1.06–2.24], those who lived in shared accommodations [aOR, 2.26, 95%CI, 1.69–3.02], as well as individuals who failed to comply with the precautionary measures advising people to avoid traveling by public transport and/or shake hands had a remarkably higher odds of moderate (aOR, 4.02, 95%CI, 2.74–5.91) and severe stress symptoms (aOR, 1.63, 95%CI, 1.15–2.30], while severe stress symptoms were also found among those who failed to wear face masks when going out [aOR, 1.64, 95%CI, 1.19–2.25].

Limitations and strengths

This study had some limitations. First, it was limited in scope. Many of the participants (50%) were from the capital city of Dhaka division, limiting the generalization of our findings to rural regions. Second, the study was carried out during the COVID-19 lockdown period and lacked longitudinal follow-up. The analysis of the periodic state of individuals may not reflect the psychological state, which changes with time and with alterations in one’s surrounding environment. Because of the increasingly arduous situation and the fear of the second wave, the mental health symptoms of residents could become more severe. Thus, the long-term psychological implications of this population are worth further investigation. Third, due to ethical requirements on anonymity and confidentiality, we were not allowed to collect contact details and personal information from the respondents. As a result, we could not conduct a prospective study that would provide a concrete finding to support the need for a focused public health initiative. Fourth, the study was not able to distinguish between pre-existing mental health symptoms and new symptoms. Fifth, there was an oversampling of a particular network of peers (e.g., students), which may lead to selection bias. However, a large percentage of respondents who were within 40 years old consisted of students. They were exposed to higher mental health symptoms due to the temporary shutdown of educational institutions, disbandment of social gatherings, and pressure from having to attend classes online [54]. Sixth, the self-reported levels of psychological impact, anxiety, depression, and stress may not always align with assessment by mental health professionals.

Despite these limitations, this study has several strengths. The study provides additional global data to inform understanding of the impact of the COVID-19 pandemic in developing countries. The findings also advance the knowledge of mental health impact during the COVID-19 pandemic in such a region with a struggling health care system by providing potentially helpful data to inform possible public health management strategizing. The findings also shed added light on the impact of COVID-19 on the mental health conditions of the educated people in Bangladesh who are the potential future workforce. An impressive strength of our study is the data volume (>10,000 survey respondents), generally exceeding that of similar studies. In addition, this study used impressive GIS maps to highlight the most affected areas in Bangladesh.