This cross-sectional study was conducted in three city corporations (Dhaka, Chattogram, and Gazipur) in Bangladesh from January to June 2023. Participants of both sexes, between the ages of 18 and 60 years, living in selected areas of Dhaka, Chattogram, and Gazipur city corporations, were included in this research. However, Participants who were injured, in a rehabilitation stage, or unwilling to participate were excluded from the study.

The sample size was calculated using the formula of Cochran’s (n = ((z² × p (1-p))/e²)). With a 5% margin of error (e), considering the mostly prevalent perceived barriers to a healthy diet (p = 66%) as reported in a previous study (25), and the standard normal variate of 1.96 (z), the required sample size was 358. However, the study team reached a large sample of 565. A total of 78 participants declined to participate in the study due to time constraints, workloads, or other personal reasons. Of the remaining participants, 487 completed the interview, yielding a response rate of 86.20%. Additionally, 20 participants were excluded as they were injured or in rehabilitation. During data cleaning, 67 more cases were excluded due to incomplete interviews, missing values and extreme outliers. Finally, a total of 400 respondents were included in the final analysis.

A multistage sampling technique was used to determine the study participants. The details of the sampling procedure have been presented in Figure 1.

In the first stage, 20 wards (sub-division of City Corporation) from three city corporations were randomly selected. At the final stage, study participants were conveniently selected for data collection from each municipal ward. Assessed cities were also represented with map (Supplementary Figure S1).

Five trained surveyor’s/data collectors were used in the data collection process who, in turn, placed in the central business district, shopping malls, and in the educational institutions to capture diverse study population. To recruit data collectors, a circular was issued among recent graduates of the Public Health Department, North South University. Applicants were shortlisted based on qualifications and interviewed. Successful candidates with strong communication skills, research ethics understanding, and fieldwork potential were selected. The principal investigator and a senior research team member provided a three-day training covering study objectives, ethical protocols, consent procedures, questionnaire content, interview techniques, handling sensitive topics, and field protocols for effective and ethical data collection. Participants were approached and a brief description of the study was conveyed. Once participants provided consent to participate in the study, data was collected through face-to-face interviews using a semi-structured and pretested questionnaire. We first formed all of the questionnaires in English, including the questions about barriers to healthy diet. Then, a professional Bengali translator translated them into Bengali. The participants had the choice of using Bengali or English questionnaire. The questionnaire includes socio-demographics, perceived barriers to healthy diet, and mental health-related questions.

People aged 18 to 60 years living in the Dhaka, Chattogram, and Gazipur city corporations were invited to participate after we approached as much as possible amount of people living in this area. The 18–60 age group is crucial for studying perceived barriers to a healthy diet and their association with mental health. This age range represents a stage in life where individuals typically experience multiple responsibilities, including career development, family obligations, and societal expectations.

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving research study participants were approved by the North South University Ethics Review Committee (REF: 2022/OR-NSU/IRB/1003). Written informed consent was obtained from all subjects/patients. Willing respondents participated voluntarily where no financial incentives or gifts were provided to this research due to funding constraints.

The STATA (V16 Stata Corp LP, TX, United States) software was used for the analyses. Outliers, identified as extreme values, and missing data were removed to ensure the accuracy of our analysis. Outliers were identified using the interquartile range (IQR) method, where values beyond 1.5 times the IQR from the first or third quartile were considered extreme and removed to minimize skewness in the dataset. For missing data, we used list wise deletion, removing cases with incomplete responses to maintain data integrity and ensure consistent sample size across analyses. Frequencies and percentages were used to narrate the baseline characteristics of the respondents. The Shapiro–Wilk test and a histogram checked the normality of outcome variables. Wilcoxon rank sum test and Kruskal-Wallis test were applied to assess the bivariate analysis as we found our outcome measurements as non-normally distributed. In the multivariate modeling, we adjusted all of the explanatory variables irrespective of their significance in the bi-variate modeling. Quantile regression was used due to the non-normality of our data, offering robust estimates even with violations of normality. We ensured the key assumptions were met, including the absence of multicollinearity, linearity, homoscedasticity, and the correct specification of the model. These diagnostics were conducted following guidelines and practices used in similar studies (33–35).

In the regression modeling, we adjusted for study location, gender, age, current marital status, family type, educational level, field of study, occupation, working hours in a day, family monthly income, BMI, Depression score, Anxiety score, and stress score. Quantile regression analyses were used to figure out how each covariate affected the perceived healthy diet barrier scores on average. A linear regression analysis was also accompanied for comparison purposes. Five quantiles, namely, the 10th, 25th, 50th, 75th, and 90th were used. The hypothesis tests were two-sided, and the p-values less than 0.05 were considered significant.

Table 1 lists the sociodemographic details of the respondents who were chosen from the city corporations of Gazipur (20.5%), Chattogram (29.0%), and Dhaka (50.5%), respectively. There were statistically significant differences (p = 0.012) in perceived barriers to a healthy diet between the study locations of Dhaka (23.95 ± 7.90), Chattogram (26.82 ± 9.08), and Gazipur (25.67 ± 7.29). The participants’ mean age was 31.10 ± 10.11, and 66.75% were between 18 and 30 years. The majority of participants were men (68.50%). Around 68.0% of the study population comprised nuclear families, and half of the sample was married.

The perceived barriers to a healthy diet score varied by gender: males scored 25.9 ± 7.65, and females scored 25.42 ± 10.07. That suggests a marginal significance (p = 0.052). Different family types, such as nuclear (25.87 ± 8.31), joint (24.0 ± 8.04), and life apart home (30.03 ± 9.71), revealed statistically significant differences in perceived barriers (p = 0.002). About 48.75% of participants had graduation degrees or higher in their educational backgrounds. Only 30.0% of the participants studied biological sciences. While 34.25% of participants held jobs, the remaining two-thirds of respondents were businesspeople and other professionals (homemakers, unemployed, and students). One-third of respondents’ working hours were 6–8; while 41.50% had no fixed working hours. Using the WHO classification, samples were divided into four BMI groups and later merged overweight and obese; which reports n = 28 (7.0%) being the underweight group, n = 294 (73.50%) normal weight group; n = 78 (19.50%) overweight and obese groups, respectively.

Figure 2 illustrates the mental health status of the 400 participants which was obtained by the DASS-21 scale. The Mean ± SD scores for were 8.00 ± 7.22 depression, 8.62 ± 7.48 for anxiety, and 16.0 ± 7.89 for stress. The results indicated that a significant portion of the participants had mild to extremely severe levels of depression, anxiety, and stress. Notably, 32.0% of the participants reported mild to extremely severe levels of depression, 47.0% reported similar levels of anxiety, and 42.5% reported similar levels of stress.

Figure 3 illustrates the results regarding the percentage of agreement and non-agreement of perceived barriers to a healthy diet among the participants for each of the 12 items. The results show that 20.0% participants agreed that healthy foods were only sometimes available in their homes. Additionally, few participants (09.25%) reported that their family does not support their efforts to change their diet. A small proportion of the participants, approximately one in eight (12%), expressed the need for more knowledge about healthy foods. On the other hand, a significant number of participants (56.25%) reported consuming junk or rich food as a reward or treat. The results also showed that it is difficult for some participants to control their eating habits during an outing with friends (56.00%). More than one-quarter of the participants (25.50%) claimed that changing their diet was too complicated. Furthermore, 44.50% of individuals reported that healthful foods were more expensive than they could afford, while 22.0% of participants said that their taste was different or unpleasant. The results also showed that it took more work for some participants (46.50%) to bring healthy foods to their work setting.

Results of quantile regression, and the Ordinary Least Square Regression (OLS) models are presented in Table 2. A multivariable quantile regression was fitted on each of the 10th, 25th, 50th, 75th, and 90th quantiles of the scores for perceived barriers to healthy diet to show a complete picture of the association between the explanatory variables and perceived healthy diet barrier scores. Pseudo R² values (ranging from 0.0867 to 0.2325) to reflect the model’s explanatory power across quantiles in the manuscript. These values indicate varying degrees of fit, with stronger fits observed at higher quantiles. The model estimate suggests that study location was associated with the perceived barriers to a healthy diet as people living in Chattogram were less likely to have a barrier score than those living in Dhaka city in the 25th (β = −3.14, 95% CI: −5.48 to −0.80) and 50th (β = −2.34, 95% CI: −4.33 to −0.34) quantile. The analysis also predicted that female gender was significantly associated with lower perceived barriers to healthy diet scores at the 50th quantile (β = −2.05, 95% CI: −3.88 to −0.21) after adjusting for other covariates. Respondents who lived apart from home tended to have higher scores on perceived barriers to a healthy diet compared to the nuclear family members at the 10th, 25th, and 50th quantiles. Participants who lived in a joint family were observed to be 3.17 points lower at the 10th quantile (β = −3.17, 95% CI: −5.45 to −0.89) perceived barriers score compared to those who lived in a nuclear family. Looking at the working hours in a day, the 90th (β = 12.44, 95% CI: 3.92 to 20.95) quantile was found to be a statistically significant predictor for the higher perceived barriers to healthy diet scores, particularly for individuals who worked for 6 h compared to the reference group.

Those whose family monthly income was (30000–60,000) BDT per month had experienced higher perceived barriers scores to healthy diet score in the 10th and 75th compared to the highest-income (> 60,000 BDT) group. The perceived barriers to a healthy diet increased by 0.34 points, 0.59 points, and 0.79 points at their 50th, 75th, and 90th quantiles, respectively, when their depression scores increased by one unit. Anxiety showed a positive and statistically significant association at the 50th quantile (β = 0.14, 95% CI = 0.01 to 0.28). Stress was also positively associated with perceived barrier scores at the 10th (β = 0.18, 95% CI = 0.09 to 0.27) and 25th (β = 0.12, 95% CI = 0.03 to 0.21) quantiles.

To address the identified barriers to a healthy diet among adults, policymakers, healthcare professionals, public health organizations, and stakeholders should take note of the study’s findings and initiate targeted interventions. Based on the findings, suggesting targeted interventions that address both mental health and dietary barriers could be a practical application for improving public health in Bangladesh. Advocacy for policies supporting mental health services and stress management programs is crucial, as they significantly facilitate healthier food choices. Education campaigns aimed at raising awareness about the importance of a balanced diet and its connection to mental health could play a pivotal role. These campaigns should be designed to reach diverse populations, including underserved communities, to ensure equitable access to information. Workplace interventions, such as the incorporation of healthy meal programs, stress management workshops, and access to mental health resources, are also recommended. Further research on innovative approaches like healthy food labeling systems and food technology-based interventions is necessary to enhance the effectiveness of interventions addressing barriers to a healthy diet. Future recommendations for research could include examining the effectiveness of these targeted interventions in addressing the perceived barriers to healthy diet and identifying additional factors that may influence healthy eating behavior among adults. Incorporating qualitative methods, such as interviews or focus groups, can offer deeper insights into cultural and contextual factors shaping dietary behavior, complementing quantitative findings to refine interventions.

This study evaluated participants’ mental health status and its association with perceptions of barriers to adopting a healthy diet. It also explores a specific context, providing insights into barriers in a representative study setting. Furthermore, this study looks into the relationship between these barriers and mental health, contributing a unique perspective on the interplay between diet and mental health. These findings contribute to the growing understanding of how barriers to healthy diets can impact overall well-being. Since the outcome variable was not linearized and not normally distributed, we used a robust statistical technique, quantile regression, to determine the association between barriers to healthy diet and other covariates. Nonetheless, this study has a few limitations. Given that it was a cross-sectional study; it was not possible to determine if certain factors caused the reported barriers to a healthy diet. The reliance on self-reported data to assess dietary habits introduces the risk of recall bias, as participants may have difficulty accurately recalling their food intake or may alter their responses. This could affect the reliability of the dietary data. Additionally, potential participant error is another concern, as misunderstandings or socially desirable responses may influence the accuracy of the reported information. Furthermore, the study’s sample was drawn from three large cities in Bangladesh, which limits its generalizability, especially to rural areas where dietary habits and health behaviors may differ significantly. Finally, the use of a non-validated questionnaire to assess dietary barriers is another limitation, as it may not accurately capture the relevant factors affecting participants’ diets.