A cross-sectional study was conducted among house-bound residents in Bangladesh from different cities (including Dhaka, Narayangonj, Sylhet, Chittagong, Mymensingh and Cox's Bazar) between July and September 2020. The data were collected utilizing an online survey tool (Google Forms) with a convenience sampling design. An online survey was used to avoid the possibility of infection and to maintain spatial distancing. After incorporating all questions in Google Forms, a shareable link was generated. The survey link was disseminated via online platforms including Facebook, Messenger, and WhatsApp in order to get a rapid response and cover a geographically diverse area of Bangladesh. Initially, 5057 participants submitted a response. After removing incomplete surveys, 4,020 were kept for final analysis. All surveys were completed in the Bengali language. The inclusion criteria included being (i) able to read Bangla, (ii) aged 18 years or over, and (iii) able to complete the entire survey. All respondents gave virtual informed consent.

The sample size was calculated using the following Equation (1):

Here, n = number of samples

z = 1.96 (95% confidence level)

p = prevalence estimate (50% or 0.5)

q = (1-p)

d = precision limit or proportion of sampling error (0.05).

We hypothesized that the prevalence estimate (p) for the present study would be 50%. Assuming a 10% non-response rate, a sample size of 423.5 ≈ 424 participants was estimated. Our sample size exceeded this estimate.

A self-reported questionnaire including questions concerning socio-demographics and financial difficulties along with a psychometric scale (i.e., DASS-21—see below) was used to collect data from participants.

Descriptive analysis including frequencies and percentages were performed for categorical variables, while means, standard deviations, etc. were calculated for continuous variables. We present prevalence estimates of depression, anxiety and stress using the predefined thresholds above. All items of the DASS-21 yielded Skewness and Kurtosis values within the ± 2.0 range, indicating that they were normally distributed (43). Chi-square tests were performed to determine the significant relationship of depression, anxiety, and stress with all examined variables. In addition, binary logistic regression analysis was performed to determine the candidates for multivariable logistic regression analysis. The variables that were statistically significant (p < 0.05) in the binary logistic regression analysis of depression, anxiety, and stress were included in the multivariable regression models. The multicollinearity was checked using tolerance (>0.1) and variance inflation factor (VIF; <10) before performing multivariable logistic regression. Analyses were performed using Statistical Package for Social Science (SPSS) version 25.0.

The survey was conducted according to the guidelines of the Helsinki Declaration 1975. The study protocol was reviewed and approved by the Ethical Review Committee of the Faculty of Biological Science and Technology, Jessore University of Science & Technology, Jessore-7408, Bangladesh [Ref: ERC/FBS/JUST/2020-43(a)]. All respondents were informed about the purpose of the study, the procedure, and the right to withdraw their data. Each participant gave virtual informed consent prior to completing the study. Participants were informed that all their information would be kept anonymous and confidential, and they were provided with information about the nature and purpose of the study.

A total of 4,020 participants (male: 53.6%) were included in the final analysis. Most were younger, aged between 18 and 25 years (86%). The majority were educated up to university level (78.9%), unmarried (87.8%), resided in urban areas (76.9%), had up to 4 members in their immediate family (54%) and a family monthly income ranging from 20,000–50,000 Bangladeshi Taka [BDT; 84.87 BDT = 1 US$] (47.6%). 7.3% reported poor self-perceived health status.

Almost 6% reported that COVID-19 had affected their family members, while the majority reported that COVID-19 had affected people in their residential area (72.3%). Over two-thirds of participants reported their income had decreased during the COVID-19 pandemic (66.7%), while 72.7% reported the possibility of decreased income in the future, and 35.3% reported the possibility of losing their job during the pandemic. In addition, 17.2% reported current food scarcity, with almost half of participants thinking food scarcity was possible in the future (45.3%).

Figure 1 presents the patterns of mental health concerns among Bangladeshi residents during the COVID-19 pandemic. The prevalence estimates of moderate to severe depression, anxiety, and stress were 71.1, 62.3, and 56.7%, respectively. As per as bivariate analysis (Chi-square tests), most variables were significantly associated with depression, anxiety, and stress (Table 1).

Table 2 shows the results of binary logistic regression analysis by depression, anxiety, and stress. The variables found to be significant in the binary logistic regression analysis were included in the adjusted models. As per as the multivariable logistic analysis (Table 3), females were more likely to have depression, anxiety, and stress compared to males (AOR_Depression: 1.34; 95% CI: 1.15–1.55, p < 0.001, AOR_Anxiety: 1.62; 95% CI: 1.41–1.86, p < 0.001, and AOR_Stress: 1.79; 95% CI: 1.57–2.05, p < 0.001). Similarly, the higher odds of depression, anxiety, and stress were found among participants who reported high family income, poor self-perceived health status, COVID-19 affected people in the local area, food scarcity, and probability of food scarcity in future.

Unmarried participants were 2 times more likely to have depression and stress than married participants (AOR_Depression: 1.87; 95% CI: 1.41–2.49, p < 0.001, and AOR_Stress: 1.72; 95% CI: 1.41–2.11, p < 0.001). Participants who reported smaller families and the possibility of decreasing income in the future had greater odds of depression and stress. In addition, the probability of losing employment was associated with higher odds of depression and anxiety; whereas COVID-19 affected people in the family and income status change during the COVID-19 were associated with anxiety and stress, respectively.

The results of the study were compared to earlier research focused on (i) depression, anxiety and stress [using Patient Health Questionnaire (PHQ-9), Depression, Anxiety, Stress Scale (DASS-21), Generalized Anxiety Disorder (GAD-7), Hospital Anxiety and Depression Scale (HADS), Health Anxiety Inventory (HAI), Geriatric Depression Scale-15], and (ii) studies with a particular focus on financial difficulties during COVID-19.

The results of this study showed that depression, anxiety, and stress were comparatively higher in females than males, in line with other epidemiological research (4, 46), but in disagreement with previous studies in Bangladesh using similar (47) and different instruments (48). It is likely that women may assume additional caretaking duties of their families during the pandemic which may account for their increased mental health problems. In the present study, depression and stress were significantly higher among those who reported being unmarried and with a lower number of family members. Being alone during lockdown can often cause feelings of loneliness and affect mental well-being leading to depression and stress (49, 50), and the present study supports and such findings. A previous study has reported that being married (compared to being single, widowed or separated) can be associated with depression (51), however this utilized data collected prior to the pandemic.

A significant association between high income family (monthly income 20,000–50,000 BDT) and depression, anxiety, and stress emerged in our results, inconsistent with prior Bangladeshi studies (4, 33, 34) which did not find a significant association between income and mental health concerns. More work is needed to ascertain why this is the case. Having family members and other people living their residential area infected with COVID-19 were associated with higher depression, anxiety and stress, consistent with previous studies which have suggested this may be due to fear of infection, enforced quarantine and feeling stigmatized (52). People with poor self-reported health status also exhibited higher odds of depression, anxiety and stress in the present study, consistent with previous Bangladeshi reports (32, 34).

In the present study, financial difficulties due to a reduction in family income, food scarcity, and potential unemployment were found to be statistically significant associated factors of depression, anxiety, and stress. COVID-19 is a highly contagious disease with catastrophic repercussions for humanity and the global economy, as well as for mental health causing psychosocial distress (53). Major victims of the COVID-19 outbreak include micro, small, medium-sized enterprises and also the general population (54). Prolonged country-wide lockdowns and a global economic recession disrupted normal supply chains and created economic hardship (55). A previous study conducted in Bangladesh indicated that the COVID-19 pandemic and partial lockdown had socio-economic impacts including on poorer communities, leading to price increases of basic essentials, disruptions to formal education, and the possibility of a severe socio-economic and health crisis (56). The data we present are a year old as it takes time to analyze data and publish the results. It is possible that these may have changed in the last year and further work may be needed to ascertain any temporal changes.

This study is not without limitations which should be taken into account when interpreting the results. The study was cross-sectional so causality cannot be determined. In this respect, a prospective study will be helpful. Respondents needed to have access to the internet to participate, indicating that they had a higher socio-economic standing than the general population, although research assistants collected data from friends and family members who were not digitally literate. Therefore, the possibility of sampling bias should be considered. Self-reported health status was measured using a binary response and that may have limited the range of responses. Finally, this study relied on self-reported responses regarding experience during home-quarantine stay which may not align with the clinical diagnosis of mental health professionals.