Mental health care seeking behavior in Bangladesh: determinants and treatment gaps | BMC Psychiatry

Mental health disorders are increasing in Bangladesh with a growing health, social, and economic burden [1]. According to the recent National Mental Health Survey 2019 [2], about 19% of the adult population suffers from mental health disorders, compared to the 16.1% prevalence of mental illness reported in the first National Mental Health Survey 2003–2005 [3]. Rapid urbanization and demographic and lifestyle changes are associated with the rise in mental health disorders [4]. However, due to limited investment in mental health service provision and health systems, services often cannot adequately meet the demand for care, leading to a large treatment gap in Bangladesh. The national budget of Bangladesh for the health sector was approximately US$ 2.3 billion, of which only 0.44% was allocated for mental health, mainly to spend through tertiary mental health hospitals (35.59% of allocation) [5]. There is also a chronic shortage of mental health workforce at all service delivery levels. There are 1.17 mental health workers per 100,000 population, most of whom worked at a tertiary care setup situated in large cities [5].

The National Mental Health Survey of Bangladesh 2019 shows that only 10% of the patients having mental disorders sought mental healthcare services [2]. The treatment gap for major depressive episodes was reported to be around 74.8% in low and middle-income countries (LMIC) [6]. The treatment gap for depression was found to be over 85% in India [7, 8]. The treatment gap reported in Nepal for depressive disorder was found to be 94.9% in 2013 [9] and 88.2% in 2016/17 [10]. Many socio-economic factors can affect the decision of the treatment-seeking behavior of an individual with mental illness. Individuals with higher average education from developed countries are more likely to seek mental healthcare services than individuals from the least developed countries, where average education and awareness levels are lower. In Cambodia, existing literature shows that patient’s education can influence the treatment-seeking decision, and traditional and religious medicines are the primary sources of treatment for mentally disordered patients [11]. Social stigma also influences mental health care-seeking behavior [2, 12], and evidence suggests that stigma is higher among males, older, Asia–Pacific Islanders, and students with lower socioeconomic backgrounds [12].Mental health literature also emphasizes some other factors such as location, gender, age, and previous experience of mental disorders that affect mental health care-seeking behavior. People living in rural areas have generally less access to mental health care services than the urban population. A study conducted on mental health disparities found that a greater number of rural Americans do not have access to mental health services [13]. A similar finding is also reported in Northeast India [14]. Studies found that males are relatively reluctant to seek mental health care treatment compared to their female counterparts [13, 15, 16]. As age increases, the care-seeking tendency generally increases among people [17]. In addition, cultural beliefs, social support, and perceived social stigma also affect mental health care-seeking behavior [14, 15, 18, 19]. Moreover, religion and religious education also play important roles in seeking mental health care [14].

Despite the availability of treatments, a large number of people living with mental illness do not receive treatment for mental disorders in LMICs, including Bangladesh. Though several studies explored the treatment-seeking behavior of people with mental illness with specific mental illness on a regional basis, limited studies have been conducted on treatment-seeking behavior and treatment gap by different types of mental illness and their influencing factors in LMICs using national = level data [7, 8, 10]. This study aimed to identify the pattern of treatment-seeking behavior for mental health among the adult population of Bangladesh and factors that affect the patient’s decision to seek mental health care. To the best knowledge of the authors, this is the first study that examines the pattern of mental healthcare treatment gap across the types of mental health disorders and mental healthcare-seeking behavior among the adult population using the latest national survey data on mental health. However, few studies on mental health used the same survey data or the published report based on this dataset [20,21,22], though they mainly focused on the prevalence of mental healthcare disorders using the published report. Therefore, this study contributes to the literature on mental health by identifying the predictive factors associated with mental healthcare-seeking behavior.

Data, variables, and method estimation

Data

The study used secondary data from the National Mental Health Survey of Bangladesh 2019, conducted by the National Institute of Mental Health (NIMH), Dhaka. The National Institute of Mental Health (NIMH), carried out the survey nationwide with the technical support of the World Health Organisation (WHO). The survey was conducted during the time period of 1 October 2018 to June 2019 [2]. The survey used two-stage random sampling methods across the 64 districts from 8 divisions of the country. In the first stage, primary sampling units (PSUs) or enumerated areas (EUs) were selected, while in the second stage, households were selected using the design effect (deff) of the study. The following definition of design effect was used in the survey: \(deff=1+\left(m-1\right)ICC\), where \(ICC\) indicates intra-class correlation (ICC), and \(m\) takes different values on a trial-and-error basis to find the best combination of household numbers and PSUs satisfying the study design. A total of 62 PSUs were independently selected from each division (31 from rural and 31 from urban). It gives a total of 496 PSUs, where 248 PSUs were selected from rural and 248 from urban areas. Finally, 18 households were selected from each of the PSUs, yielding a total of 8928 households in the survey. Out of the 8928 households, a total of 7270 adults and 2270 children from 7270 households were finally interviewed for the survey.^{Footnote 1} For this study, we have used the data of the adult population (18 years and above) to analyze their healthcare-seeking behavior and other statistical analyses. The adult population has a mean age of 41 years, ranging from 18 to 99 years. The survey covered males and females in almost equal numbers and included both the rural and urban samples equally.

Variables

We have described the current healthcare-seeking behavior for 12 major mental health disorders, including addictive disorders, depressive disorders, personality disorders, anxiety disorders, somatic symptoms disorders, neurocognitive disorders, sleep–wake disorders, obsessive–compulsive disorders, disruptive disorders, schizophrenia-spectrum disorders, sexual dysfunction disorders, and bipolar-related disorders. The treatment-seeking status (whether sought or not, a dichotomous variable) of any mental health disorder has been considered the dependent variable for the logit and probit model. The probability of healthcare-seeking behavior has been estimated over selected socioeconomic variables, including gender, education, religion, age, residence, number of family members, marital status, and the existence of other mental patients in the family. To get a better understanding of the variables for healthcare-seeking behavior in Bangladesh, a detailed description of the variables is presented in Table 7 in the appendix. Besides, ‘treatment gap’ and ‘wealth index’ are defined as follows:

• Treatment gap: The treatment gap is generally defined as the difference between the actual prevalence of mental health disorders and the treated fraction of the population affected by mental health disorders. In other words, the proportion of individuals who do not receive treatment but need mental health care [23].

• Developing wealth index: Principal component analysis (PCA) was used to develop the wealth index. Selected household fixed assets are considered to develop the index, including electricity, flush toilet, land phone, mobile phone, television, refrigerator, private car, moped/scooter/motorcycle/autorickshaw, washing machine, bicycle, sewing machine, wardrobe, table, bed, chair/bench, clock, computer/laptop/tablet computer, domestic animal (cow/ox/goat, etc.), shallow machine/power tiller/tractor), rickshaw, farming land, pond for fish cultivation and roof type of the main house [2]. Each component was assigned a score to make the index, ranging from 0 to 261,512, with an average of 15,392. The lower the value of the wealth index, the poorer the household is.

Method estimation

The probability and marginal effect of the dichotomous or binary variable (status of receiving mental health care treatment) have been investigated in our study using the logit and probit model. To estimate the factors or determinants, the use of logit or probit models is common practice across various research fields [14, 17]. Although Linear Probability Model (LPM) can be used to identify the probability and marginal effect of the dichotomous variable to identify the determinants of mental healthcare-seeking behavior in Bangladesh, the logit or probit models is more frequently used to serve the purpose due to some limitations of the Linear Probability Model (LPM) [24]. The linear probability model was as follows:

where, \(z\) is the dichotomous variable and presents the respondent’s treatment-seeking status (0 for not seeking treatment and 1 for seeking treatment). The dichotomous regressand, \(z\), portrays the probability of changing the status from’does not seek mental healthcare services’to’seek mental healthcare services’. The vector of independent variables, \(x\), delineates the factors that govern the probability of making a decision for mental healthcare services (gender, education, religion, age, residential status, household size, existence of mental health patients, and marital status). It is expected that the probability value will fall in the range of 0 to 1. However, the probability value may fall outside the range of 0 to 1 under LPM [13, 14]. Besides, non-normality, heteroscedasticity of the error term, and lower value of \({R}^{2}\) of LPM lead to the selection of the cumulative distribution function (CDF) [24]. The cumulative logistic distribution is one of the CDF distributions. According to the logistic cumulative distribution function (CDF),

The cumulative distribution function captures the non-linear effect of the explanatory variable on the deponent variable. However, the most common practice is to address the normal distribution function that leads to the probit model due to the standard normal distribution function. Incorporating the normal cumulative distribution function (CDF), the probit model can be modeled as follows,

where \(z\) is the binary variable representing 1 when a respondent sought mental healthcare service and 0 otherwise. P is the probability of seeking mental health services, given the independent variables. The independent variables are explained by the \(x\) vector. The \(x\) vector contains the socioeconomic variables and \(N\) represents normal distribution with mean 0 and variance \({\sigma }^{2}\), i.e., \(N\sim N(0,{\sigma }^{2})\) [24, 25].

The marginal effect depicts the probability of the dependent variable in response to the change in the explanatory variables. The probability of mental health care seeking due to change in the independent variables is delineated through the marginal effects. The marginal effect can be obtained through the derivative with respect to the independent variables.

where, \(f(\alpha +\beta x)\) represents the standard normal probability density function assessed at \(\alpha +\beta x\) [13, 14], normally distributed with a mean of 0 and variance of 1. In our study, we wielded both logit and probit models.