Type 2 diabetes mellitus in Bangladesh: a prevalence based cost-of-illness study

Background The economic burden of type 2 diabetes has not been adequately investigated in many low- and lower middle-income countries, including Bangladesh. The aim of this study was to estimate the cost-of-illness of type 2 diabetes and to find its determinants in Bangladesh. Methods A cross-sectional study was conducted in 2017 to recruit 1253 participants with type 2 diabetes from six diabetes hospitals, providing primary to tertiary health care services, located in the northern and central regions of Bangladesh. A structured questionnaire was used for face-to-face interviewing to collect non-clinical data. Patients’ medical records were reviewed for clinical data and hospital records were reviewed for hospitalisation data. Cost was calculated from the patient’s perspective using a bottom-up methodology. The direct costs for each patient and indirect costs for each patient and their attendants were calculated. The micro-costing approach was used to calculate direct cost and the human capital approach was used to calculate indirect cost. Median regression analysis was performed to identify the determinants of average annual cost. Results Among the participants, 54% were male. The mean (±SD) age was 55.1 ± 12.5 years and duration of diabetes was 10.7 ± 7.7 years. The average annual cost was US$864.7 per patient. Medicine cost accounted for 60.7% of the direct cost followed by a hospitalisation cost of 27.7%. The average annual cost for patients with hospitalisation was 4.2 times higher compared to those without hospitalisation. Being females, use of insulin, longer duration of diabetes, and presence of diabetes complications were significantly related to the average annual cost per patient. Conclusions The cost of diabetes care is considerably high in Bangladesh, and it is primarily driven by the medicine and hospitalisation costs. Optimisation of diabetes management by positive lifestyle changes is urgently required for prevention of comorbidities and complications, which in turn will reduce the cost. Electronic supplementary material The online version of this article (10.1186/s12913-019-4440-3) contains supplementary material, which is available to authorized users.


Background
Diabetes is one of the most prevalent non-communicable diseases globally and, currently, the disease is a major public health issue in developing countries because of its chronic nature, rapidly increasing prevalence, related complications, and the requirement of long-term care. The higher prevalence of diabetes is related to an increased prevalence of obesity, population ageing, population growth, urbanisation and physical inactivity [1]. The International Diabetes Federation (IDF) estimated that, worldwide, approximately 425 million people had diabetes in 2017; the number is projected to be 629 million by 2045. For treating and preventing diabetes and its related complications, an estimated US$727 billion was spent in 2017, which represented an 8% increase from that estimated for 2015. The cost has been projected to be US$776 billion by 2045 [2]. The annual cost for people with diabetes is mainly related to direct (e.g. cost for medicine, hospital care, laboratory tests, etc.) and indirect costs (e.g. productivity loss from disability, premature mortality, etc.) [3].
Compared to people living in high-income countries, people in low-and lower middle-income countries (LMICs) have a lack of access to health insurance or publicly available medical services. Thus, they pay a larger share of out-of-pocket (OOP) health expenditures. Furthermore, in some LMICs, people with diabetes and their families bear almost all of the expenditure related to diabetes care [4]. The prevalence of diabetes has escalated more rapidly in South East Asia than in any other large region in the world [2]. Literature showed that about 90 to 95% of all diagnosed diabetes cases of this region are type 2 diabetes [5,6]. In Bangladesh, the estimated prevalence of diabetes among adults was 9.7% in 2011 [7] and the number is projected to be 13.7 million by 2045 [2]. According to the Bangladesh National Health Accounts, in 2010, Bangladesh spent US$2.3 billion on health (or US$16.20 per person per year) and 64% of this cost came from OOP payments [8]. However, according to the World Health Organization (WHO), in 2014, Bangladesh spent US$88 per person per year on health [9]. It has been observed that, on average, a household spent 7.5% of its total income on receiving health care, with the poorest 20% of the households spending approximately 13.5% of their income on it [8]. The per capita gross domestic product (GDP) of Bangladesh was US$1677 in 2018 [10], and nearly onethird (31.5%) of the population in the country was below the poverty line [11]. Hence, the OOP health care expenditure posed a notable economic burden on the Bangladeshi population.
In high-income nations, such as the USA [12][13][14][15] and in some European [16] and upper middle-income countries [17,18], the economic burden of diabetes is wellacknowledged and investigated. Most of these studies have estimated the economic burden in terms of cost, while others [14,15,17] investigated the factors (e.g. patients' demographics, complications, payment methods and health care utilisation) correlated with the total cost. Low-and lower middle-income countries represent 80% of the global diabetic population [2]; however, researchbased evidence on diabetes management-related cost is limited for most of these countries, including Bangladesh. A study [19] in Bangladesh that addressed the cost and its determinants recruited a relatively small sample from a single hospital located in the capital city, and thus covered mostly urban residents. Furthermore, the cost was calculated from the outpatient department only, which may underestimate the average annual cost.
Thus, the aim of this study was to estimate the average annual cost and to find its determinants, where the cost data included both outpatients and hospitalisation. This study's findings will provide the most up-to-date information on the economic burden incurred by people with type 2 diabetes mellitus (T2DM) in Bangladesh, which will be useful as an important aid in the planning of health care needs and allocation of scarce resources.

Study design and study population
A cross-sectional study was conducted with a prevalence-based approach [20]. Data was collected from six hospitals (specialising in diabetes) where patients' records were available for the previous years, located in the northern and central regions of the country. Two of these hospitals are from the central region, providing primary to tertiary health care, particularly to urban residents. The remaining hospitals are from the northern region, two of them providing primary and secondary health care and the other two providing primary to tertiary health care to people residing in semi-urban and rural locations. Patients attending hospitals providing primary and secondary but in need of tertiary care are usually referred to the tertiary care hospitals. Due to a similar social and economic status for people living in the northern and southern regions, the patients from the hospitals of the northern region are comparable to those living in the southern region of the country. Thus, no hospitals were selected from the southern region. The Diabetic Association of Bangladesh (BADAS), a not-forprofit but mostly self-sustaining social welfare organisation, directly or indirectly (through affiliated local associations) owns all the selected hospitals. BADAS, the highest diabetes care provider, has 75 diabetic centres/ hospitals which cover all 64 districts (second highest level tire of regional administration) across the country. Due to lack of adequate services related to diabetes in public hospitals, particularly in peripheral areas, majority of the people with diabetes are treated and managed by the hospitals under BADAS. The hospitals were purposively selected to ensure that the study included patients from rural-urban as well as professionally mixed populations (e.g. service holders, businessmen, farmers, day labours, housewives, etc.) attending various levels of health care services. Between April and September 2017, 1253 participants were recruited using systematic random sampling and probability proportional to size (PPS) methods (Fig. 1). The target population comprised registered adults of either gender with a minimum one-year duration of T2DM. People with other types of diabetes or who were pregnant at the time of data collection were excluded as those people may have some additional expenses other than T2DM. A team of trained data collectors was involved in the data collection. At the beginning of the interview, the data collectors provided an explanatory statement to each participant and, upon his/her agreement to participate by signing the consent form, participants were interviewed face-to-face. Participants were recruited from the outpatient department of the selected hospitals and those who were referred to hospital admission were followed-up to collect their hospitalisation information from the hospital inpatient department.

Data collection instrument
A structured questionnaire (Additional file 1) was developed and used in a secured web-based application, Research Electronic Data Capture (REDCap), for data collection [21]. Prior to the main survey, the content validity of the questionnaire was evaluated and pretested, using a pilot study conducted in a tertiary care hospital. The questionnaire gathered each patient's details and demographics, diabetes-related information, cost-related information and the patient's lifestyle behaviour. Patients' medical records (guide books) were also reviewed to retrieve information pertaining to clinical status and the number of medical services received during the previous year. To obtain information about the types and quantities of currently prescribed medicine, a copy of the prescription was kept in the REDcap application as an image for further use.

Calculation of costs
The total cost of T2DM was calculated from the patient's perspective for the year 2017, considering direct and indirect costs as the major components. Direct cost was estimated using a bottom-up approach for primary data collection [22] and divided into the two following sub-categories: direct medical costs, which comprise the costs of hospitalisation, outpatient visits, medicine, laboratory tests, and other service utilisation (including the use of self-monitoring blood glucose and consumables); and direct non-medical costs, which comprise the cost of transportation and meals en-route to the hospital.
The micro-costing approach was used to identify cost items in as much detail as possible for calculating direct cost [23]. Cost per daily dose [24] of each medicine and therapy was defined. Costs related to medicine, consultancy and laboratory tests that patients paid OPP were collected from a tertiary level hospital located in the central region. It should be noted that the health insurances support is literally non-exists in Bangladesh, thus all payments met by OPP. All hospitals under the BADAS are homogeneous; thus, there is a negligible cost variation between the selected hospitals. For other components of direct medical cost and direct non-medical cost, each participant's responses to the questionnaire were considered as a reference. Each component of direct cost was calculated by multiplying the unit cost with the quantities of medical services received during the previous year. The cost of hospitalisation (including hospital stay, medicine and laboratory tests during the stay) was retrieved for each patient from their hospital record, which was provided by the accounts department of the hospital. The total direct cost was calculated by adding up all components of direct medical and direct nonmedical costs.
The indirect cost was calculated for both patients and their attendants' en-route to the hospital. The productive time lost to attend outpatient visits and during hospital admission was recorded based on the information provided by the patients and their attendants. The human capital approach [25] was used to calculate the indirect cost for those who were productive and in the formal workforce or housewives, but not for people who were unable to work (retired or ill health) or who chose not to work. The productivity loss of housewives was calculated using the minimum wage rate of Bangladesh (US$224/annum) [26] as well as the median income of the participants who were in the formal workforce. The total cost was calculated by adding up total direct and total indirect costs.
All costs were calculated in Bangladeshi currency, Taka (BDT) and then, to add an international perspective, they The number of people approached to participate in the study n=1404 63 participants were unable to participate due to their critical health condition The total number of participants participated in the study n=1253 The number of participants those were eligible to participate in the study n=1316 89 participants were excluded due to having diabetes other than type 2 were converted into US$ using the mid-year currency conversion rate for the year 2017 (US$1 = BDT80).

Ethical approval
The study has been approved by the Monash University Human Research Ethics Committee (ID: 1469), the Ethical Review Committee of the Bangladesh University of Health Sciences (BUHS) and the Ethical Review Committee of the BADAS.

Statistical analysis
Descriptive statistics includes mean with standard deviation for numerical data and frequency with percentage for categorical data. A normality test of cost data was performed using histogram, Q-Q plot and the Shapiro-Wilk test. Cost data was right skewed; hence, mean and median with percentiles was used for reporting it. A median regression was run to determine the factors related to average annual cost. A bootstrapping method was used to calculate the 95% confidence interval (CI) of regression coefficients [27]. A one-way sensitivity analysis was performed to evaluate the assumption that the use of minimum wage to calculate the indirect cost of housewives may give the lowest estimate. A two-way sensitivity analyses were undertaken to explore the change in average annual cost with assumptions of a 25% (+/−) change in the prevalence of insulin use and 25% (+/−) change in the prevalence of T2DM related complications. The statistical software package STATA SE version 15.0 was used for data analysis and a p-value of 0.05 or less was considered to be statistically significant.

General characteristics of the study participants
The general characteristics of the study participants are presented in Table 1. Among the 1253 participants with T2DM, 681 (54.3%) were male. The mean age of patients was 55.1 ± 12.5 years. Approximately 45% of participants   ).

Cost-of-illness by socio demographic and clinical characteristics
Cost-of-illness (total cost) by socio-demographic and clinical characteristics is presented in Annual cost-of-illness (COI) of diabetes care

Sensitivity analysis
The result of the one-way sensitivity analysis showed that use of the median income of the study participants instead of the minimum wage increased the indirect cost by 23%. However, the estimated average annual cost using the minimum wage rate of Bangladesh was US$853 (95% CI US$795.1-US$911.7), while it was US$864.7 (95% CI 806.5-922.9) using the median income of the study participants. The difference between these two estimates is insignificant as the CI overlapped each other. The results of a two-way sensitivity analyses showed that the average annual cost increased by 2.9% (US$865 vs US$890) when insulin use was increased by 25% and that decreased by 4% (US$865 vs US$830) when insulin use was decreased by 25%. A 25% increase in prevalence of complications lead to a 5.3% (US$865 vs UD$898) increment of average annual cost, while it decreased by 3.9% (US$865 vs US$819) with a 25% reduction in complications.

Determinants of cost-of-illness
The results of simple and multiple median regression analyses are presented in Table 4. In the simple median regression analysis, age group 61-80 years (US$221.91, p < 0.001) Multiple regression analysis showed that the average annual cost was higher for females (US$44.85, p = 0.036). People treated with insulin with a combination of OHA (US$152.87, p < 0.001) also had higher costs compared to those treated with OHA only. Patients with a duration of diabetes of more than 10 years (US$66.93, p = 0.025) incurred a higher cost. Likewise, patients with the presence of any complication (US$63.69 for one or two and US$440.93 for more than two, p < 0.001) had higher costs compared to those without any complication.

Discussion
Diabetes has become a major global economic burden in recent decades, but proper management of the factors related to it can be useful for reducing this burden. Diabetes is also an increasingly economic threat in Bangladesh, yet studies on an adequate estimation of COI for T2DM and its key drivers are limited; hence, the aim of this paper. This study involved a large representative sample that adequately investigated the economic burden of type 2 diabetes in Bangladesh from the patients' perspective. The key finding of this study was that the average annual cost for T2DM patient is US$865 with the medicine cost being the highest contributor followed by the hospitalisation cost. The average annual cost for patients with hospitalisation was 4.2 times higher compared to those without hospitalisation.
The average annual cost for each person with T2DM in Bangladesh appears to be considerably higher than that reported in previous studies conducted in Bangladesh (US$314) [19] and other South Asian countries such as India (US$525) [28] and Pakistan (US$197) [29]. One possible explanation of this difference may be because the studies conducted in Bangladesh and Pakistan addressed only outpatient department cost, which underestimated the overall cost. In contrast, some high-or upper middle-income Asian countries, for example, China (US$1501.7) [30] and Singapore (US$1575.6) [18], reported higher cost for diabetes management.
This study finding showed that cost increased with age, which is supported by previous studies [14,17,31]. Additionally, female gender was a factor more likely to incur higher cost. A study conducted in Bangladesh by Shamima et al. showed that females had better awareness about diabetes and were more regular in receiving follow-up check-ups [7], which may be related to higher cost. A study conducted in Hawaii by Bhattacharyya et al. [14] showed a similar result, while Krop et al. [31] in Maryland and Chaikledkaew et al. [32] in Thailand showed that the cost of care was higher for males.
An important finding of this study was that 13.5% of participants had income less than the estimated average annual cost. Overall, a person with T2DM spent 9% of his/her annual household income on management, which is a notable financial burden for a family. In South Asia, health insurance is practically non-existent, and almost all expenses are met through OOP, which creates a significant burden and sometimes leads to family impoverishment. This study showed that urban residents spent more than the rural residents (mean cost for urban: US$1024.4, vs rural: US$421.6). This may be because of people residing in urban areas have better education and a higher income, and thus can better afford to receive adequate treatment and access to specialised doctors.  Other service utilisation b 10.0 9.0 9, 36 54.0 1. 8 12,195.0 The present results showed that direct cost had the largest share (90.5%) of overall cost. Among all the cost components of overall direct cost, medicine cost was the major contributor (60.7%). The studies conducted by Khowaja et al. [29] in Pakistan and Shobhana et.al [28] in India reported similar features. However, compared to these studies, the present study showed a much higher proportion of medicine cost. This difference may be related to many factors. Firstly, 93% of the participants in this current study had high or medium adherence to medication, which incurs a higher medicine cost. Secondly, among them, 65.6% used either insulin alone or insulin with a combination of OHA, which is higher than that reported in another study in Bangladesh [33].   [14,[34][35][36][37] also showed the medicine cost as a major contributor to direct cost. Medicine was the highest source of direct cost (83.5%) for patients without hospitalisation. Furthermore, for patients with hospitalisation, medicine cost was 50.7% of the direct cost followed by a hospitalisation cost of 39.9%. However, a number of previous studies showed that the largest proportion of cost is attributable to hospitalisation followed by medicine cost. In the USA, hospitalisation cost accounted for 50% [38], while that was 55% in the Cost of Diabetes in Europe-Type II study [39].
In the present study, the cost of diabetes care substantially increased with the presence of comorbidities as well as complications related to T2DM. The cost was positively correlated with the increased number of comorbidities and complications leading to hospitalisation. This finding is supported by other studies in the developed [30,37,[40][41][42] as well as in developing countries [43,44]. In addition to complications and comorbidities, the duration of diabetes also accelerates cost; likewise, cost increased for patients who had poor glycaemic control compared to good control. Similar results were reported in previous COI studies [28,29,37,45].
A Median regression analysis showed that female gender, use of insulin, longer duration of diabetes, and presence of complications are the factors related to a higher annual average cost per person. A majority of these variables also appeared as contributing factors in previous studies [14,15,17,28,42]. Thus, optimisation of the management of diabetes-related complications is an imperative need for people with T2DM in Bangladesh, which has also reflected in the results of sensitivity analyses of the current study.
The present study showed that in 2017 the annual average cost per T2DM was US$864.7, which is 52% of per capita GDP of Bangladesh [10] and 9.8 times higher than the general health care cost [9]. The burden of diabetes is influenced by many socio-economic and health care system factors, which consequently affects the cost of care. Early screening is one of the factors that may help to diagnose T2DM patients at an initial stage, thus avoiding complications. However, in developing countries, people often seek medical help when they have already developed some complications. In addition, inadequate awareness about diabetes-related complications, lack of access to medical care resulting from income disparities, lack of social supports, and heterogeneous quality of care are other societal factors that influence the diabetes-related cost of care [46]. Thus, this study finding will be useful for policy makers in planning future health care needs and allocating scarce resources. Furthermore, it will play a significant role for both patients and provider in identifying and quantifying the costs attributed to T2DM in Bangladesh.
A strength of this study was that it addressed all possible cost components of both direct (including hospitalisation) and indirect cost from the patients' perspective, and cost was calculated based on primary data. However, other recent COI studies have not addressed all cost components [3,38], despite the cost being calculated from the societal perspective. Another strength was that a professional mix of patients residing in urban and rural locations was recruited randomly form six hospitals that provide primary to tertiary care.
Some limitations should also be noted. Firstly, in addition to direct and indirect costs, there are also intangible costs (e.g. pain, suffering, and loss of quality of life), which was not addressed in this study. Secondly, due to a lack of information, the calculation of indirect cost was done using the traditional human capital approach rather than the frictional cost approach [47]. Finally, as it was a descriptive cost-of-illness study, it did not provide information on the efficiency of resource use; thus, higher cost does not necessarily mean better services or value for money.

Conclusions
Diabetes is a major public health issue with a high economic burden in Bangladesh. The development and improvement of interventions toward better control of T2DM and the prevention of its complications are vital requirements for the country. Without these, in the near future, the private and public financing of diabetes treatment will be severely constrained, representing a health threat for the Bangladeshi population.