Out-of-pocket expenditure on medicines in Bangladesh: An analysis of the national household income and expenditure survey 2016-17.

BACKGROUND AND OBJECTIVES: High out-of-pocket expenditures (OOPE) increases the probability that households will become impoverished or will forgo needed care. The aim of this paper is to study household medicines expenditure and its associated determining factors to develop policies to protect households from financial hardship.
METHODS: The present cross-sectional and population-level study used the Bangladesh 2016-17 National Household Income and Expenditure Survey (HIES). The final sample size was 46,080 households. We analyzed the probability of OOPE for medicines, the share of total OOPE due to medicines out of total OOPE in health (reported as a ratio between zero and one), the OOPE amount for medicines reported (in United States Dollars), and the share of OOPE amount on medicines out of total household expenditure (reported as a ratio between zero and one). Predictors of analyzed outcomes were identified using three regression models.
RESULTS: Out of those households who spent on healthcare, the probability of having any OOPE on medicines was 87.9%. Of those who spent on medicines, the median monthly expenditure was US$3.03. The poorest households spent 9.97% of their total household expenditure as OOPE on medicines, nearly double that of the wealthiest households (5.86%). The characteristic which showed the most significant correlation to a high OOPE on medicines was the presence of chronic diseases, especially cancer. Twenty six percent of all surveyed households spend more than 10% of their OOPE on medicines.
CONCLUSIONS: Our study shows that financial protection should be targeted at the poorest quintiles and such protection should include enrollment of rural households. Further, outpatient medicines benefits should include those for non-communicable diseases (NCDs).

Introduction

One of the most important concerns in low- and middle-income countries (LMICs) is the increasing out-of-pocket health care expenditure (OOPE) made by households and individuals. OOPE is the amount of money paid by households to purchase health services and medicines when members of a household have a health care need. Healthcare costs are among the largest barriers to accessing health services and achieving universal health coverage (UHC), and among the most important factors associated with the reduction of the welfare of households. In particular, high OOPE increases the probability that households will become impoverished or will forgo needed care [1] and, as a result, households may decide to sell assets or take out loans to pay for this healthcare [2, 3].

Out of all possible healthcare financing mechanisms, OOPE is considered the most inequitable. It has been estimated that, due to health expenditure, 100 million households fall into extreme poverty every year -living on US$1.90 per day or less [4]. In this regard, in many countries medicines represent the largest proportion of OOPE on healthcare. A financial protection analysis in eight Southeast Asian countries showed that in seven out of eight countries medicines represent between 75–81% of OOPE [5]. Despite significant progress towards achieving UHC in many low- and middle-income countries, substantial challenges remain in terms of access to quality healthcare and lack of financial protection [6, 7].

Out of pocket expenditure is amendable by public policy. For instance, the implementation of health insurance should protect households from large OOPE, including medicines. Furthermore, policies regulating the payment of providers have an influence on their behavior, including ordering diagnostic tests, prescribing medicines and recommending surgery and other types of treatment. These considerations make the study of OOPE relevant for setting health policies and assessing their effect on economic development and poverty reduction. There is ample literature on the study of OOPE [8], the methodological foundation [9] and its current application to assess progress on UHC [1].

Evidence shows that in many countries the majority of health OOPE is for medicines. For instance, in India 90% of OOPE on health is on medicines, in Nepal it is 88%, and Indonesia it is 78%. Medicines OOPE are also important as a proportion of total household expenditure. According to the World Health Survey, up to 95% of the total expenditure of poorer households in LMICs is spent on medicines, and this is far higher than the 3.5% expended by poorer households in high income countries (HICs) [10]. Approximately half (41%-56%) of households in LMICs spent 100% of their health care expenses on medicines [10]. Medicines OOPE is a large driver of overall OOPE and countries have made a commitment to UHC. Thus, studies focusing on medicines OOPE have become increasingly relevant.

Bangladesh is a country with over 160 million inhabitants and rapid urbanization that has made great progress in health, education, and economic development over the past decades. Maternal mortality has fallen by 60% over the past two decades and child mortality by two-thirds [11]. Bangladesh is expected to ‘graduate’ from the World Trade Organization’s designation as a ‘least developed country’ as soon as 2024 [12]. However, Bangladesh faces significant challenges to improve the health and wellbeing of its population due to a lack of coherent social security or financial health protection. Bangladesh is incurring a demographic shift toward longer life expectancy and it is experiencing an epidemiological transition from predominantly infectious diseases toward chronic, non-communicable diseases that require sustained medication and life-long treatment. For example, the prevalence of diabetes in Bangladesh is relatively high at about 10% [13]. Most patients with diabetes require long-term medication, diagnostic and monitoring devices apart from other medical care. We have previously shown, ex ante, that household OOPE on insulin is likely to have an effect on the probability of individual households’ becoming impoverished, as well as having a much more expansive effect on the country’s welfare [14].

Bangladesh, however, spends only a relatively low percentage of GDP on health compared to several other countries in the region and relies heavily on OOPE as the main source of health financing. Just 2.27% of Bangladesh’s GDP is spent on health [15] out of which 74% is OOPE [5] and medicines are the largest component of OOPE in health. The 2005 National Household and Expenditure Survey identified the cost of medicine as greater than any other factor in determining OOPE on health [16]. The 2010 Bangladesh National Household and Expenditure Survey showed that medicines represented 61% of the OOPE on health [17] and a recent cross-sectional survey showed that prices of some essential medicines in Bangladesh are consistently expensive across both public and private sector facilities [18].

The contributions of this present study are twofold. First, there is a gap in our understanding of medicines OOPE determinants. Knowing these determinants would support the development of policies to protect households from financial hardship. Second, this study uses the most recent national-level survey data on healthcare utilization and OOPE for out-patient in Bangladesh, making the study highly relevant from public policy standpoint. The aim of this paper is to study household medicines expenditure and its associated determining factors.

Material and methods

Settings/design and analytical sample

The present cross-sectional and population-level study used the Bangladesh 2016–17 National Household Income and Expenditure Survey (HIES). Details of the survey design have been described elsewhere [19]. Briefly, HIES seeks to obtain detailed data on household income, expenditure and consumption, determine the poverty profile with urban and rural breakdown and district-level poverty, provide household level consumption data for compiling national accounts estimates, and provide relevant data for monitoring of the Poverty Reduction Strategy, five-year plan and the Sustainable Development Goals. This survey contains a wide range of socio-economic information at the household level that has strong influence on the decision-making process for the government.

The sample for the HIES is explicitly designed to produce estimates at the three levels (urban and rural, district-level, and household level) and is designed with an urban sample size large enough to understand Bangladesh’s urbanization patterns. A sample design was adopted for the HIES with 2,304 Primary Sampling Units (PSU) in eight administrative and geographical divisions (Barisal, Chittagong, Dhaka, Khulna, Mymensingh, Rajshahi, Rangpur and Sylhet) and 64 districts, selected from the last Housing and Population Census (2011) [20]. Within each PSU, 20 households were selected for interviews. The final sample size was 46,080 households [20] and was stratified at the district level, including a total of 132 sub-strata: 64 urban, 64 rural, and four main City Corporations.

For the present study, we excluded 3.1% of households surveyed with incomplete relevant information, the analytical sample included 43,659 households (representing all approximately 37.6 million households in Bangladesh). To test whether there is a difference between the included and the excluded households, we examined potential differences in covariates that could be associated with our outcome variables between our analytical sample and those excluded ones. We did not find any significant differences.

Variables

From our sample (see above), we analyzed five main outcome variables: 1) The probability of OOPE in medicines, 2) the share of medicines OOPE out of total OOPE in health (reported as a ratio between zero and one), 3) the OOPE amount in medicines reported (in United States dollars based on the annual average exchange rate obtained from the central bank of Bangladesh), 4), the share of OOPE on medicines out of total household expenditure (reported as a ratio between zero and one)- based on the method of Wagstaff et al. [2, 8], and 5) the share of OOPE on medicines out of a given household’s capacity to pay (reported also as a ratio between zero and one)- based on the method of Xu et al. [9, 21]. We calculated the amount of total OOPE in health by taking the sum of all expenditures reported by households in the last month and year before the survey, including out- and inpatient care and medicines.

We followed previous studies [17, 21, 22] and included as covariates such as age, sex (male = 1/female = 0), schooling (none, primary, secondary and tertiary), religion (Islam, Hinduism, Buddhism, other), marital status (married, never married, widowed/Divorced/Separated), and working during the last seven days. We also included h as covariates: number of ‘equivalent adults’which adjusts for the economy of scale in consumption. That is, a household with three members, including children, does not consume three times that of a one-person household. According to the OECD [23], the equivalence scale considers the age of the household members and establishes a standardization that allows comparison). We also considered as covariates the demographic dependence ratio (the number of dependents aged 0–14 and those over the age of 65, compared with the total population aged 15–64), an additive and unweighted disability index (measured by the presence in all household members of difficulty in seeing, hearing, walking, climbing, remembering or concentrating, washing all over or dressing and communicating) operationalized as a percentage, the presence (yes = 1/no = 0) of any member with any symptoms of illness/injury in the last 30 days, the use of health services (yes = 1/no = 0) by any of the members who reported a health need in the last 30 days, the presence (yes = 1/no = 0) of any member with a chronic disease (categorized as diabetes, cardiovascular disease, cancer, chronic diseases of infectious origin, disabilities, others), an asset and housing material-index as a measure of socioeconomic status constructed using factor analysis [24, 25] and expressed as a percentage, the participation in any safety nets or social programs (yes = 1/no = 0) and place of residence such as rural/urban and the administrative and geographical division.

Statistical analysis

We used survey weights to account for the complex survey design in all descriptive and multivariable analyses. We report population estimates for all results. All analyses were performed using the svy module of the statistical package Stata version 15.1 [26]. We first quantified the household characteristics described previously reporting mean, percentage and their 95% confidence intervals (CIs). The median of the expenditure was calculated for those households which had an expenditure greater than zero.

We developed four main outcome variables among surveyed households as a function of the quintile of monthly household expenditure per equivalent adult. We report median and interquartile range, and percentage and their 95% CI.

Predictors of analyzed outcome variables were estimating with three regression models: 1) a logistic regression model [27] for the probability of OOPE on medicines; 2) three fractional logistics multiple regression models [28, 29] for the share of OOPE on medicines out of: OOPE on health; total household expenditure; a given household’s capacity to pay (the latter was estimated using the STATA fracreg); and 3) a linear regression model [27, 28] for the OOPE on medicines per adult equivalent expressed as a napierian logarithm. For the first regression model, we reported adjusted odds-ratios (aORs) and CI95% and for the second and third models, we reported adjusted coefficients (aCoeff) and CI95%. Finally, based on the first regression analysis results, we estimated the adjusted share of OOPE on medicines out of total household expenditure by considering the following thresholds of the total household expenditure: 10, 15, 20, 30%). We also used the first regression model to estimate share of medicine OOPE out of a household’s capacity to pay by considering different thresholds (10, 15, 20, 30 and 40%) of the total household expenditure) [30] as well as according to the quintile of household expenditure per equivalent adult.

Following previous studies [22], our estimations had to consider the presence of a selection bias related to the decision to spend funds on health because there may be particular household characteristics that increase the probability of health expenditure. This bias applied to all households. Following Heckman (1979) [31], we used a logistic model to estimate the conditional probability that a given household would record any given health OOPE (as a function of the household characteristics mentioned above). We then calculated the Mills ratio [32] to capture the magnitude of the selection bias for each household analyzed. Subsequently, this parameter was incorporated as a regressor in the regression models described above.

Results

Table 1 describes the household characteristics. The mean age of household heads was 43.6 years and 87.7% of household heads were male. About 75% of all household heads have either no or only a primary level schooling. Nine out of ten household heads report Islam as their religion and 91.8% of the household heads reported being married. Nearly half (46.1%) of households had at least one member with at least one chronic disease (e.g. cardiovascular diseases, diabetes, cancer). Half of the households reported having a member with symptoms of illness/injury in the last 30 days and out of this half, 89.4% used health services. One in five households (21.2%) are benefiting from a social program.

Table 1

Main household characteristics.

HIES, Bangladesh, 2016/17.

Sample size (n) = 43,659 households	Mean or % and CI 95%
weighted sample (N) = 37,616,656 households
Household head
Age (yrs)	43.57 [43.27–43.86]
Male	87.67 [87.10–88.24]
Schooling
Nothing	30.59 [29.61–31.56]
Primary	43.28 [42.27–44.30]
Secondary	21.22 [20.27–22.16]
Tertiary	4.91 [4.35–5.48]
Religion
Islam	89.29 [87.72–90.87]
Hinduism	9.48 [8.22–10.74]
Buddhism	0.88 [0.32–1.44]
Other	0.35 [0.19–0.50]
Marital status
Marriage	91.78 [91.42–92.14]
Never marriage	2.10 [1.91–2.28]
Widowed/Divorced/Separated	6.12 [5.81–6.44]
Working	84.15 [83.34–84.96]
Household
Equivalent adults	2.73 [2.71–2.75]
Demographic dependence	79.30 [77.95–80.65]
Disability index	63.21 [58.31–68.12]
Any member with any symptoms of illness/injury in the last 30 days	53.32 [51.93–54.72]
Use of health services	89.32 [87.72–90.91]
Any member with a chronic disease	46.15 [44.77–47.53]
Often infectious origin	60.45 [59.18–61.72]
Disabilities	34.55 [33.29–35.81]
Diabetes	12.12 [11.22–13.02]
Cardiovascular disease	28.76 [27.61–29.91]
Cancer	0.53 [0.41–0.65]
Others chronic disease	18.85 [17.92–19.78]
Socioeconomic index	19.39 [18.77–20.01]
Beneficiary/member of any safety nets/social program	21.18 [20.07–22.29]
Place of residence
Rural	71.30 [69.03–73.57]
Division
Barisal	5.57 [5.19–5.95]
Chittagong	19.98 [17.86–22.11]
Dhaka	25.30 [22.77–27.83]
Khulna	10.90 [10.14–11.67]
Mymensingh	7.59 [6.18–9.00]
Rajshahi	12.55 [11.04–14.06]
Rangpur	10.95 [9.83–12.07]
Sylhet	7.15 [6.69–7.62]

Most surveyed households are in rural areas (71.3%), with only 25.3% in the urbanized Dhaka area. There are large disparities between the lowest quintile of monthly household expenditure (Quintile 1) and the wealthiest (Quintile 5), in particular regarding the demographic dependence (95.5 and 62.8 respectively) and socioeconomic index (11.9 and 32.7 respectively).

Among all households, the probability of any OOPE on health in the last month was 74.4%. Out of those households who spent anything on healthcare, the probability of having OOPE on medicines was 87.9% (Table 2). When we adjusted per equivalent adult, out of those households who spent anything on healthcare, their median total monthly health-related OOPE was US$3.1. With similar adjustment per equivalent adult, those households spending OOP on medicines had a median OOPE of US$3.0. Thus, 96.5% of the total monthly healthcare-related OOPE was on medicines.

Table 2

OOPE on health and medicines according to quintile of household expenditure.

HIES, Bangladesh, 2016.

Weighted sample (N) = 37,616,656 households	Overall	Quintile of monthly household expenditure per equivalent adult
		1st	2nd	3th	4th	5th
Total household expenditure per equivalent adult (US$), p50 and IQR	60.42 [41.59–92.24]	30.87 [25.92–34.73]	44.97 [41.59–48.47]	60.42 [56.31–65.20]	83.20 [76.18–92.24]	145.36 [120.99–192.29]
Probability of OOPE on health, %	74.41 [72.93–75.89]	70.24 [68.40–72.08]	75.85 [73.88–77.81]	74.13 [71.84–76.41]	73.42 [71.09–75.76]	78.42 [75.73–81.11]
OOPE on health per equivalent adult (US$), p50 and IQR	3.14 [1.12–8.77]	1.79 [0.66–4.52]	2.57 [0.94–6.18]	3.21 [1.22–8.26]	4.25 [1.51–11.81]	5.47 [1.74–16.49]
Share of OOPE on health out-off household expenditure, %	8.20 [7.90–8.50]	8.87 [8.39–9.34]	8.63 [8.19–9.08]	8.29 [7.80–8.78]	8.36 [7.76–8.96]	6.86 [6.32–7.41]
Share of inpatient expenditure out-off OOPE on health, %	4.36 [3.99–4.73]	2.78 [2.33–3.24]	3.70 [3.13–4.27]	4.26 [3.26–5.26]	5.15 [4.26–6.05]	5.76 [4.89–6.63]
Share of outpatient expenditure out-off OOPE on health, %	95.64 [95.27–96.01]	97.22 [96.76–97.67]	96.30 [95.73–96.87]	95.74 [94.74–96.74]	94.85 [93.95–95.74]	94.24 [93.37–95.11]
Probability of OOPE on medicines, %	87.92 [86.63–89.21]	86.96 [85.75–88.17]	88.36 [87.05–89.68]	88.49 [87.07–89.90]	89.24 [87.77–90.70]	86.57 [82.32–90.81]
OOPE on medicines per equivalent adult (US$), p50 and IQR	3.03 [1.32–7.32]	1.80 [0.85–4.09]	2.42 [1.12–5.31]	3.06 [1.41–6.84]	3.82 [1.65–9.31]	5.25 [2.12–13.06]
Share of OOPE on medicines out-off OOPE on health, %	71.49 [70.24–72.73]	74.96 [73.70–76.22]	73.85 [72.44–75.27]	71.83 [70.26–73.39]	70.52 [68.89–72.14]	66.67 [63.00–70.33]
Share of OOPE on medicines out-off household expenditure, %	8.06 [7.80–8.33]	9.97 [9.46–10.47]	8.59 [8.18–9.00]	8.13 [7.69–8.57]	7.98 [7.49–8.48]	5.86 [5.32–6.39]
Share of OOPE on medicines out-off household’s capacity to pay, %	14.16 [13.96–14.36]	19.85 [19.30–20.40]	16.87 [16.42–17.33]	14.42 [13.99–14.85]	11.74 [11.36–12.11]	7.60 [7.32–7.89]

The OOPE share out of total household expenditure on healthcare and medicines was 8.2% and 8.1%, respectively. In this regard, households in the first income quintile (poorest households) spent 8.9% of their OOPE on healthcare as a share of their total household expenditure whereas those in the 5th income quintile (wealthiest) spent only 6.7%. With regard to medicines, the proportion of total household expenditure relegated to medicine OOPE is the highest for the poorest households (9.9%)–nearly double that of the wealthiest households (5.9%). Expressed as the capacity of the household to pay, the disparity between households in the poorest and the wealthiest quintile is even larger: while poor households spent nearly a one-fifth (19.8%) of their disposable income on medicines, the wealthiest households spent 7.6% on medicines.

Moreover, the overwhelming proportion of the healthcare OOPE is for outpatient expenditures (95.6% outpatient versus 4.3% inpatient). See Table 2.

Several factors are associated with the increased probability of having any OOPE for medicines in a household, including for instance age of the household head, the household head never being married, being in a rural area, and reporting a chronic disease (except cancer) compared to those not having the disease (Table 3). By contrast, having a male household head, having a job, and being a beneficiary of a social program reduced the probability of OOPE on medicines. The actual amount of medicines expenditure (adjusted per number of equivalent adults in the household) increased with factors including the age of the household heads, the household head education, households with larger demographic dependence, living in rural areas, and with all chronic diseases (especially cancer). Being covered by a safety net/social program reduces the share of OOPE on medicines out of total health expenditure, reduces the amount of OOPE on medicines and the share of OOPE on medicines out of household expenditure.

Table 3

Factors associated with OOPE on medicines.

HIES, Bangladesh, 2016/17.

	Logistic regression model: OOPE on medicines > 0	Factional regression model: Share of OOPE on medicines out-off OOPE on health	OLS regression model: OOPE on medicines per adult equivalent (ln)	Factional regression model: Share of OOPE on medicines out-off total household expenditure or household’s capacity to pay
				out-off total household expenditure	out-off household’s capacity to pay
	Adjusted odds-ratios	Adjusted coefficient	Adjusted coefficient	Adjusted coefficient
Household head
Age (yrs)	1.036 [1.027―1.045]***	0.017 [0.014―0.020]***	0.010 [0.008―0.012]***	0.005 [0.003―0.007]***	0.004 [0.002―0.006]***
Male	0.543 [0.362―0.816]**	-0.380 [-0.508―-0.252]***	0.107 [0.026―0.187]**	0.114 [0.018―0.210]*	0.146 [0.046―0.245]**
Schooling
Nothing	Ref.	Ref.	Ref.	Ref.	Ref.
Primary	0.779 [0.633―0.958]*	-0.173 [-0.248―-0.097]***	0.116 [0.069―0.164]***	-0.058 [-0.111―-0.004]*	-0.093 [-0.147―-0.038]**
Secondary	0.668 [0.516―0.866]**	-0.338 [-0.434―-0.242]***	0.231 [0.161―0.301]***	-0.071 [-0.152―0.009]+	-0.167 [-0.248―-0.085]***
Tertiary	0.697 [0.472―1.029]+	-0.439 [-0.588―-0.291]***	0.304 [0.208―0.400]***	-0.126 [-0.256―0.004]+	-0.264 [-0.398―-0.129]***
Religion
Islam	Ref.	Ref.	Ref.	Ref.	Ref.
Hinduism	0.857 [0.624―1.178]	-0.043 [-0.136―0.049]	-0.098 [-0.167―-0.030]**	-0.061 [-0.142―0.019]	-0.013 [-0.084―0.058]
Buddhism	0.601 [0.342―1.054]+	0.108 [-0.408―0.623]	-0.287 [-0.646―0.071]	-0.292 [-0.646―0.063]	-0.203 [-0.568―0.162]
Other	0.623 [0.193―2.007]	-0.087 [-0.423―0.248]	-0.078 [-0.309―0.153]	-0.196 [-0.442―0.050]	-0.168 [-0.423―0.087]
Marital status
Marriage	Ref.	Ref.	Ref.	Ref.	Ref.
Never marriage	3.842 [2.109―6.997]***	0.731 [0.510―0.952]***	0.106 [-0.041―0.253]	0.196 [0.041―0.352]*	0.099 [-0.045―0.244]
Widowed/Divorced/Separated	1.027 [0.608―1.734]	0.008 [-0.146―0.161]	-0.252 [-0.344―-0.160]***	0.026 [-0.069―0.121]	0.064 [-0.036―0.163]
Working	0.741 [0.525―1.046]+	0.081 [-0.016―0.178]	-0.179 [-0.246―-0.112]***	-0.220 [-0.294―-0.146]***	-0.124 [-0.202―-0.047]**
Household
Equivalent adults	0.956 [0.822―1.111]	-0.106 [-0.142―-0.069]***		-0.129 [-0.158―-0.101]***	-0.132 [-0.160―-0.105]***
Demographic dependence	0.878 [0.762―1.013]+	0.035 [-0.012―0.082]	0.015 [-0.019―0.048]	0.047 [0.010―0.083]*	0.110 [0.075―0.146]***
Disability index	0.988 [0.932―1.048]	0.005 [-0.012―0.022]	0.020 [0.009―0.032]**	0.030 [0.020―0.041]***	0.031 [0.019―0.043]***
Any member with any symptoms of illness/injury in the last 30 days	0.978 [0.630―1.519]	-0.198 [-0.390―-0.007]*	-0.148 [-0.271―-0.026]**	-0.187 [-0.318―-0.055]**	-0.188 [-0.322―-0.055]**
Use of health services	10.322 [5.828―18.280]***	-0.521 [-0.711―-0.331]***	0.668 [0.539―0.797]***	0.419 [0.294―0.544]***	0.492 [0.365―0.620]***
Chronic disease
Often infectious origin	3.798 [2.705―5.334]***	0.173 [0.118―0.227]***	0.295 [0.252―0.338]***	0.185 [0.142―0.229]***	0.160 [0.115―0.205]***
Disabilities	2.595 [1.801―3.739]***	0.146 [0.081―0.211]***	0.344 [0.297―0.391]***	0.262 [0.213―0.310]***	0.265 [0.214―0.316]***
Diabetes	3.780 [1.482―9.639]**	0.168 [0.038―0.297]*	0.652 [0.586―0.717]***	0.333 [0.261―0.406]***	0.300 [0.225―0.376]***
Cardiovascular disease	6.022 [3.896―9.310]***	0.263 [0.192―0.334]***	0.503 [0.448―0.557]***	0.319 [0.264―0.373]***	0.322 [0.263―0.381]***
Cancer	4.481 [0.747―26.875]	-0.248 [-0.594―0.097]	1.497 [1.145―1.849]***	1.205 [0.813―1.598]***	1.031 [0.603―1.460]***
Others chronic disease	3.811 [2.643―5.494]***	0.070 [-0.007―0.148]+	0.663 [0.606―0.720]***	0.520 [0.462―0.579]***	0.431 [0.373―0.490]***
Socioeconomic index	0.984 [0.978―0.991]***	-0.010 [-0.012―-0.007]***	0.006 [0.004―0.008]***	-0.010 [-0.012―-0.008]***	-0.016 [-0.018―-0.014]***
Beneficiary/member of any safety nets/social program	0.732 [0.626―0.855]***	-0.107 [-0.160―-0.054]***	-0.018 [-0.061―0.025]	-0.038 [-0.084―0.008]	-0.041 [-0.085―0.004]+
Place of residence
Rural	0.990 [0.738―1.329]	0.089 [0.020―0.158]*	0.077 [-0.005―0.160]+	0.107 [0.047―0.167]***	0.146 [0.087―0.205]***
Division
Barisal	Ref.	Ref.	Ref.	Ref.	Ref.
Chittagong	1.688 [1.182―2.411]**	-0.073 [-0.158―0.013]+	0.214 [0.122―0.305]***	-0.015 [-0.095―0.065]	0.041 [-0.038―0.120]
Dhaka	0.860 [0.553―1.338]	0.025 [-0.072―0.121]	-0.170 [-0.279―-0.062]**	-0.328 [-0.411―-0.245]***	-0.284 [-0.364―-0.203]***
Khulna	0.382 [0.289―0.507]***	-0.372 [-0.450―-0.295]***	-0.514 [-0.610―-0.418]***	-0.346 [-0.420―-0.272]***	-0.200 [-0.269―-0.131]***
Mymensingh	0.878 [0.532―1.450]	0.222 [0.084―0.360]**	-0.427 [-0.549―-0.306]***	-0.366 [-0.484―-0.247]***	-0.090 [-0.201―0.021]
Rajshahi	0.298 [0.219―0.405]***	-0.373 [-0.457―-0.288]***	-0.518 [-0.614―-0.422]***	-0.295 [-0.371―-0.219]***	-0.125 [-0.198―-0.052]**
Rangpur	0.959 [0.684―1.344]	0.195 [0.104―0.287]***	-0.436 [-0.537―-0.335]***	0.002 [-0.078―0.082]	0.124 [0.051―0.198]**
Sylhet	1.783 [1.113―2.857]*	0.307 [0.197―0.417]***	-0.149 [-0.251―-0.047]**	-0.263 [-0.349―-0.177]***	-0.069 [-0.150―0.012]+
Intercept	6.882 [3.395―13.952]***	1.966 [1.753―2.180]***	0.845 [0.677―1.014]***	-2.272 [-2.441―-2.104]***	-1.759 [-1.928―-1.591]***
Weighted sample (N)	27,991,346	27,991,346	24,609,301	27,991,346	27,322,598

***P<0.001,

**P<0.01,

*P<0.05,

+P<0.10.

In order to capture the magnitude of the selection bias in the occurrence of household expenditure in health, five models were adjusted also by Mills ratio.

The characteristic which showed the most significant association with a high OOPE on medicines relative to total household expenditure was the use of health services and the presence of chronic diseases, especially cancer, followed by diabetes and cardiovascular diseases. The amount, as well as the share, of OOPE on cancer medicines out of the total household expenditure has by far the highest incremental risks (Fig 1).

Fig 1

Adjusted incremental risk of OOPE on medicines according to household presence chronic disease.

HIES, Bangladesh, 2016.

Twenty six percent of households spend more than 10% of their OOPE on medicines. By increasing the spending threshold to 15% of OOPE, the percentage of households incurring this level of spending dropped to 6.8% (Table 4). Nearly two percent (1.8%) of households spent 20%, and 0.1% spent 30% of their total expenditure on OOPE for medicines. These proportions decrease with increasing wealth. For instance, 27.3% of households in the 1st quintile spent 10% of their household expenditure on medicines compared to 23.6% in the 5th quintile.

Table 4

Adjusted share of OOPE on medicines out-off total household expenditure according to quintile of household expenditure.

HIES, Bangladesh, 2016/17.

Threshold	Overall	Quintile of monthly household expenditure per equivalent adult
		1st	2nd	3th	4th	5th
≥10%	26.178 [25.003―27.353]	27.279 [25.475―29.083]	26.246 [24.583―27.909]	26.442 [24.485―28.398]	27.525 [25.524―29.527]	23.614 [21.276―25.952]
≥15%	6.820 [6.362―7.278]	6.649 [5.809―7.488]	6.760 [5.991―7.529]	7.200 [6.241―8.160]	7.113 [6.255―7.972]	6.396 [5.429―7.363]
≥20%	1.824 [1.619―2.028]	1.746 [1.390―2.102]	1.725 [1.370―2.080]	1.911 [1.399―2.423]	1.910 [1.488―2.333]	1.825 [1.402―2.249]
≥30%	0.138 [0.095―0.181]	0.124 [0.038―0.210]	0.078 [0.007―0.149]	0.200 [0.068―0.333]	0.128 [0.046―0.210]	0.159 [0.058―0.260]

We use household capacity to pay to calculate the proportion of medicines expenditure out of total household expenditure. We find that 71% of households spent more than 10% of their entire disposable income on medicines (Table 5). Large disparities exist between the poorest and the wealthiest households. While 23.8% of the poorest households spent over 20% of their disposable household income on medicines, only 12.1% of the wealthiest households did so. The disparity between the lowest and wealthiest households increases with higher thresholds: while 0.5% of household spent over 40% of their disposible household income on medicines, only 0.2% of the wealthiest households did so.

Table 5

Adjusted share of OOPE on medicines out-off household’s capacity to pay according to quintile of household expenditure.

HIES, Bangladesh, 2016/17.

Threshold	Overall	Quintile of monthly household expenditure per equivalent adult
		1st	2nd	3th	4th	5th
≥10%	71.058 [69.262―72.854]	80.376 [78.767―81.984]	78.532 [76.635―80.429]	74.696 [72.274―77.118]	69.612 [66.928―72.296]	53.394 [49.288―57.500]
≥15%	42.213 [40.557―43.869]	52.660 [50.628―54.691]	48.027 [45.742―50.312]	43.939 [41.495―46.383]	40.161 [37.592―42.729]	27.519 [24.531―30.508]
≥20%	18.547 [17.568―19.525]	23.816 [22.170―25.463]	20.960 [19.493―22.427]	19.501 [17.748―21.253]	16.957 [15.380―18.534]	12.078 [10.518―13.637]
≥30%	2.760 [2.503―3.016]	3.458 [2.796―4.120]	3.114 [2.594―3.633]	2.806 [2.197―3.414]	2.613 [2.118―3.108]	1.885 [1.460―2.310]
≥40%	0.296 [0.230―0.363]	0.527 [0.329―0.725]	0.268 [0.126―0.412]	0.250 [0.099―0.401]	0.250 [0.120―0.380]	0.206 [0.092―0.320]

Discussion

Bangladesh has made remarkable progress in relation to its population health and economic development over the past 20 years [33]. Ensuring the continuation of this progress depends partly on strengthening existing social programs and developing new programs such as a financial health-related protection for all citizens [34]. To guide the development and implementation of these policies, it is critical to identify determinants of OOPE on healthcare, especially those that determine OOPE on medicines, as these represent the largest proportion of healthcare OOPE.

The findings of this present study fill an important knowledge gap in terms of OOPE on medicines in Bangladesh. We show that the probability of having any healthcare expenditure within the previous month is high (74.4%)–in other words healthcare expenditures are frequent and medicines themselves represent nearly three quarters of all health related OOPE (71.5%). Both findings show the significance of medicines expenditure and the importance of including medicines in the benefit package of any Bangladeshfinancial protection program. Financial healthcare protection is much more relevant for poorer households as it significantly affects their overall household expenditure. Enrollment of poor households in social insurance programs is often a critical challenge to reach those most in need. For example, Seguro Popular, the pro-poor insurance program that Mexico implemented between 2003 and 2018, included medicines as part of this benefit catalogue after recognizing its relevance especially on poor households [35]. Moreover, the fact that almost all OOPE by households was for outpatient care highlights the critical importance of including outpatient medicines benefits in any effective financial protection scheme. In several countries, pharmaceutical benefit packages are limited to inpatient care which has resulted in insufficient financial protection of households [36].

Our findings not only show that poorer households pay proportionally much more for medicines than wealthier households, but also that the disparity between them is very large. When looking at medicines OOPE as a fraction of total household expenditure on healthcare, poor households pay nearly double that of their wealthiest counterparts. Household expenditure that exceeds 30% is regarded as impoverishing [37]. One of our most important findings is that 0.1%—one in every 1,000- Bangladeshi households spent more than 30% of their total monthly household expenditure on medicines. This means that medicines expenditure in Bangladesh can result in an estimated 376,000 households incurring catastrophic expenditure every month, and rural households are more affected compared to urban ones. The large number of households affected each month has the potential to significantly reduce opportunities for economic development and welfare of the population that is already very vulnerable.

Furthermore, our results demonstrate that chronic diseases have an increasing impact on household OOPE expenditure. Inclusion of common NCDs in an insurance benefit package is therefore critical to lower the probability, and especially the amount, of OOPE in NCDs. Cancer is particularly associated with very costly treatment and there is growing literature about what is called the “financial toxicity” of cancer treatment [37], especially in countries where the entire costs are born by patients, as it is in Bangladesh. Cost drivers are primarily the cost of surgery, radiation and medicines to treat cancer along with other indirect costs such as transportation, food, childcare [36]. Moreover, with the arrival of many biological oncology medicines, the costs of treatment have dramatically increased [38].

Limitations

The results of this study have several limitations. First, this observational and cross-sectional analysis explores associations and does not prove causation. Second, we do not rule-out the existence of recall bias and lower accuracy in the self-report of analyzed variables. However, the HIES is the most detailed national representative survey available anywhere and we note that HIES uses a sophisticated method of collecting data to minimize recall bias. For instance, data pertaining to daily consumption is collected by the same enumerator every day visiting the household. A third limitation is that health expenditure is largely influenced by health status and the Bangladesh 2016–17 HIES provides only partial information on diagnoses or any other clinical information for household members. Fourth, we used the household as a unit of analysis. Although this is standard practice, it is noteworthy that it does not account for any complexity of diversity of families. Fifth, the data on health are self-reported which can introduce recall errors due to the fact that the respondent may not know or remember health related information. This could have affected our analysis of expenditure by type of disease. Sixth, this study does not analyze medicines prices or price elasticity as a factor affecting medicines OOPE because information on medicine prices is not collected as part of the survey. Linking outside data source with this analysis is very challenging as individuals in the survey do not report on specific products that they purchased. Finally, this study focuses on determinants of OOPE that are recorded within this survey. We did not link other databases to study determinants of expenditure such as distance to the nearest government/private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power.

Conclusions

The introduction of efficient financial protection for the population is critical to protect households from increasing health and medicine expenditure. Our study shows that financial protection should be targeted to the poorest quintiles, enrollment of rural households should be ensured, and outpatient medicines benefits should including those for NCDs. The results of Bangladesh’s 2016 HIES can serve as a baseline for measuring the progress achieved by the introduction of a new financial protection scheme in health with particular focus on medicines. Findings from this study would be also supportive to the healthcare financing strategy of the Bangladesh’s Government for monitoring the progression towards UHC.

Adjusted incremental risk of OOPE on medicines according to household presence chronic diseae.

(DOCX)

Click here for additional data file.

23 Jun 2021

PONE-D-21-14254

Out-of-pocket expenditure on medicines in Bangladesh: an analysis of the national household income and expenditure survey 2016-17

PLOS ONE

Dear Dr. Wirtz,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Aug 07 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Mohammad Bellal Hossain

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The paper examines the determinants of OOP expenses for medicines in Bangladesh using a set of cross-sectional data. The paper is well-written, but I have the following concerns on the current version of it:

1. The studies examining the determinants of OOP healthcare expenses in different contexts are voluminous. Contemporary researchers have now moved towards more experimental type research designs like randomized control trials to examine behavioral aspects of individual choices with regard to household expenses, and those studies still make significant contributions to the literature on consumer behavior. In this backdrop, this study uses a set of cross-sectional data from Bangladesh HIES 2016/2017 to examine the determinants and their association with OOP household expenses on medicines. I am therefore wondering whether the study makes any significant contribution to the literature. What is the novelty of the study? Which research gap the study is going to bridge? Very precisely, you need to elaborate what the contributions of this study are.

2. The study is lacking a sound theoretical foundation as well. You might want to postulate related hypotheses based on a theory or a set of theories as this study uses the deductive research approach. Therefore, my recommendation would be to develop a conceptual framework using related theories of consumer behavior before proposing the methodology.

3. Did you use the total sample or only the sample of households with positive expenditure for medicines to estimate fractional models and OLS regression? It is not clear from the result tables as they do not include the number of observations used for each model. First, I recommend authors to include vital information like number of observations, post-estimation test results, and measures of model fit under each model. Also, as I understood, household decision making process with regard to the demand for medicines has two stages: Whether to spend OOP for medicines or not, if yes, then how much to be spent. The current analytical process has not taken into account this two-stage nature of household decision making. For instance, Tobit model, Double hurdle model, and Heckman two-stage model may be better alternatives to check the robustness of current findings and to account for approximately the real nature of household decision making process for the demand for medicines.

4. Household expenses for healthcare should be examined on a broader backdrop of healthcare utilization. The magnitude of OOP expenses for medicines depends on whether a household utilizes in-patient care, out-patient care, or clinic services like dental clinics and maternal clinics (Kumara and Samaratunge, 2019). Otherwise, the analysis on OOP expenses would be incomplete. Can you statistically investigate the interplay between OOP expenses for medicines and healthcare utilization? If required, use the following study for more literature

Ajantha Sisira Kumara and Ramanie Samaratunge (2019), Relationship between healthcare utilization and household out-of-pocket healthcare expenditure: Evidence from an emerging economy with a free healthcare policy, Social Science & Medicine 235, pp. 1-12

5. The importance of supply-side factors in determining OOP expenses for medicines is completely ignored from the study, making it lopsided. Do you have data in HIES on healthcare supply-side factors? For instance, distance to the nearest government /private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power. My recommendation would be incorporate such variables and see whether they play a role in Bangladesh like in other contexts.

6. As the study uses a set of cross-sectional data, you need to be careful of addressing the issue of endogeneity. As the study has many omitted variables, it would definitely lead the way for the issue. For instance, you have omitted the variable of health insurance ownership which might affect both OOP healthcare expenses and health status of household members (via moral hazard), leading the way for endogeneity. Thus, checking the results for endogeneity would be advised, and endogeneity-corrected models like IV-regression models might need to be applied to have more accurate results.

Reviewer #2: I have some major concern on conceptualization and analyses of the paper.

Conceptualization: The paper analyzed OOP on medicine in Bangladesh which is partial. as authors mentioned medicine accounts 61% of OOP in the country and so it is a unique analyses. I differ in this ground. The paper would have focused on OOPE and a sub-section on medicine. If such analyses have already been carried out, authors need to think differently.

Methods: Authors used budget share approach and provided the varying incidence of CHE. Findings suggest, at 10% threshold, the CHE of fourth quintile is higher than first quintile. This is primarily due to limitations of budget share approach in estimating the CHE. I suggest the authors should use the capacity to pay approach as the expd and income survey provides required variable for estimation of CHE.

In table 4, authors must mention whether it is incidence of CHE ?

Adjusted per equivalent adult and out of those households

192 who spent on health, their median total monthly health-related OOPE was US$3.14.

I am not sure whether is is per capita or per household. I think it should be per capita

authors must mention how they derive adult equivalent scale ? What weight they assign to children? Need mentioning

Authors disease classification is not adequate. Moreover, they must mention as limitation of self reported diseases as expd survey typically collect self reported data

Abstract: Background and objective

There is only one sentence of aim

Keep a line on background

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Ajantha Sisira Kumara

Reviewer #2: Yes: Sanjay K Mohanty

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

13 Oct 2021

Responses to the reviewers' comments

Manuscript [PONE-D-21-14254] - [EMID:661117b46576190d]

We would like to thank the reviewer for the constructive comments. We have carefully addressed the each of the reviewer’s comments in this revision. We believe that after incorporating your thoughtful feedback, our manuscript has been greatly improved.

Reviewer #1:

The paper examines the determinants of OOP expenses for medicines in Bangladesh using a set of cross-sectional data. The paper is well-written, but I have the following concerns on the current version of it: 1. The studies examining the determinants of OOP healthcare expenses in different contexts are voluminous. Contemporary researchers have now moved towards more experimental type research designs like randomized control trials to examine behavioral aspects of individual choices with regard to household expenses, and those studies still make significant contributions to the literature on consumer behavior. In this backdrop, this study uses a set of cross-sectional data from Bangladesh HIES 2016/2017 to examine the determinants and their association with OOP household expenses on medicines. I am therefore wondering whether the study makes any significant contribution to the literature. What is the novelty of the study? Which research gap the study is going to bridge? Very precisely, you need to elaborate what the contributions of this study are.

RESPONSE: We agree with the reviewer that different designs are used to study out-of-pocket expenditure. Our original contribution lays in using a secondary dataset (the HIES household survey) that has not be explored to answer an important policy question: what factors contribute to high OOP on medicines in Bangladesh in 2016? As we mention in our introduction there are other studies which have looked at OOP on health in general. However, no recent study has analyzed the factors that drive high OOP on medicines in Bangladesh, a country with rapidly increasing life expectancy fueling the epidemiological transition from a largely infectious towards a growing non-communicable disease burden.

We have changed the background section to address the point raised by the reviewer: “Although there are some previous studies of the OOPE on medicines in Bangladesh the contributions of this study are twofold: there is a gap in our understanding of the determinants of OOPE on medicines. Knowing these determinants would support the development of policies to protect households from financial hardship. Additionally, this study uses the most recent national-level survey data on healthcare utilization and OOP expenditure for out-patients in Bangladesh making the study highly relevant from a public policy standpoint.”

2. The study is lacking a sound theoretical foundation as well. You might want to postulate related hypotheses based on a theory or a set of theories as this study uses the deductive research approach. Therefore, my recommendation would be to develop a conceptual framework using related theories of consumer behavior before proposing the methodology.

RESPONSE: Thank you for your suggestion to add more on the theoretical foundation in the background of our manuscript. We have added the following section:

“Out-of-pocket expenditure is amendable by public policy: for instance, the implementation of health insurance should protect households from large out-of-pocket expenditure on health, including medicines. Furthermore, policies regulating the payment of providers also have an influence on their behaviors, e.g., the ordering of diagnostic tests, the prescribing of medicines and recommendations of surgery and other types of treatment. This makes the study of out-of-pocket expenditures relevant for setting health policies and assessing their effect on economic development and poverty reduction in a country. There is ample literature on the study of OOPE [8], the methodological foundation [9] and its current application to assess progress on UHC [1].” It is noteworthy that this study is not analyzing consumer behavior.

3. Did you use the total sample or only the sample of households with positive expenditure for medicines to estimate fractional models and OLS regression? It is not clear from the result tables as they do not include the number of observations used for each model. First, I recommend authors to include vital information like number of observations, post-estimation test results, and measures of model fit under each model. Also, as I understood, household decision making process with regard to the demand for medicines has two stages: Whether to spend OOP for medicines or not, if yes, then how much to be spent. The current analytical process has not taken into account this two-stage nature of household decision making. For instance, Tobit model, Double hurdle model, and Heckman two-stage model may be better alternatives to check the robustness of current findings and to account for approximately the real nature of household decision making process for the demand for medicines.

RESPONSE: Thank you for these reflections. For each of the tables we have included the total number of observations and included an explanation whether the analysis included the total sample or only a subgroup such as those households with a positive expenditure on health:

We now have used the Heckman two-stage model for our analysis to consider the presence of the selection bias that the reviewer mentions:

“Following previous studies [19], our estimations had to consider the presence of a selection bias related to the decision to spend funds on health because there may be particular household characteristics that increase the probability of health expenditure. This bias is applied to all households. Following Heckman (1979) [28], we estimated a logistic model to calculate the conditional probability that a given household would record any given health OOPE (as a function of the household characteristics mentioned before). We then calculated the Mills ratio [29] to capture the magnitude of the selection bias for each household analyzed. Subsequently, this parameter was incorporated as a regressor in the four regression models described above.”

It should be noted that we do not study causal inference. We are studying the magnitude of out-of-pocket expenditure and the factors that are associated with the magnitude and frequency of medicines expenditure.

4. Household expenses for healthcare should be examined on a broader backdrop of healthcare utilization. The magnitude of OOP expenses for medicines depends on whether a household utilizes in-patient care, out-patient care, or clinic services like dental clinics and maternal clinics (Kumara and Samaratunge, 2019). Otherwise, the analysis on OOP expenses would be incomplete. Can you statistically investigate the interplay between OOP expenses for medicines and healthcare utilization? If required, use the following study for more literature Ajantha Sisira Kumara and Ramanie Samaratunge (2019), Relationship between healthcare utilization and household out-of-pocket healthcare expenditure: Evidence from an emerging economy with a free healthcare policy, Social Science & Medicine 235, pp. 1-12

RESPONSE: We would like to thank the reviewer for the recommendation of the publication. We have now added health care utilization as one of the factors that influences health expenditure. We have revised all tables to reflect this change as well as the methods and results section to accommodate the incorporation of health service use as a factor influencing health and medicine expenditures. As expected, health service use increases health and medicines expenditure and is an important driver not only of the probability of expenditure but also the amount of spending.

5. The importance of supply-side factors in determining OOP expenses for medicines is completely ignored from the study, making it lopsided. Do you have data in HIES on healthcare supply-side factors? For instance, distance to the nearest government /private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power. My recommendation would be incorporate such variables and see whether they play a role in Bangladesh like in other contexts.

RESPONSE: The purpose of this study is to analyze household and individual factors that are associated with medicines OOPE. The reviewer is correct that additional data on the supply side could be added to complete this analysis. However, the survey does not provide geospatial data at the unit of analysis which is the household. Adding supply side factors such as nearest public dispensary, etc. would be approximations based on the smallest geographical area. We have added in the discussion a section where we expand on the need for analysis of the supply-side factors.

“Finally, this study focuses on determinants of out-of-pocket expenditure that are recorded within this survey. We did not link other databases to study determinants of expenditure that are not included in the survey such as distance to the nearest government /private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power.”

6. As the study uses a set of cross-sectional data, you need to be careful of addressing the issue of endogeneity. As the study has many omitted variables, it would definitely lead the way for the issue. For instance, you have omitted the variable of health insurance ownership which might affect both OOP healthcare expenses and health status of household members (via moral hazard), leading the way for endogeneity. Thus, checking the results for endogeneity would be advised, and endogeneity-corrected models like IV-regression models might need to be applied to have more accurate results.

RESPONSE: The reviewer is correct in asserting that health insurance would be expected to be an important determinant of out-of-pocket expenditure. However, in Bangladesh it is estimated that less than 1% of the population have health insurance (Rahman, 2019). Furthermore, the survey instrument does not have a separate question on health insurance affiliation only whether the household is Beneficiary/member of any safety nets/social program. We have included this variable in our models: “Beneficiary/member of any safety nets/social program”. Our study results show that only one fifth (21.18%) of the population are covered by some form of safety nets or social program which include health insurance. The regression models show that being covered by a safety nets/social program reduces the share of OOP on medicines out of total health expenditure, reduces the amount of OOP on medicines and the share of OOPE on medicines out of household expenditure.

We have added this aspect in the results section: “Being covered by a safety nets/social program reduces the share of OOP on medicines out of total health expenditure, reduces the amount of OOP on medicines and the share of OOPE on medicines out of household expenditure.” However, it is important to recognize that the survey manual clarifies that the social programs do not include support of health service use or reimbursement of medical expenditure.

Reviewer #2:

1. I have some major concern on conceptualization and analyses of the paper. Conceptualization: The paper analyzed OOP on medicine in Bangladesh which is partial as authors mentioned medicine accounts 61% of OOP in the country and so it is a unique analyses. I differ in this ground. The paper would have focused on OOPE and a sub-section on medicine. If such analyses have already been carried out, authors need to think differently.

RESPONSE: The reviewer is correct that OOPE on medicines is only a part of the total OOP on health. We expanded the background section to explain the relevance of medicine OOP in the context of health expenditure and its relation to total household expenditure.

“Out of pocket expenditure is amendable by public policy: for instance, the implementation of health insurance should protect households from large out-of-pocket expenditure on health including medicines. Furthermore, policies regulating the payment of providers also have an influence on their behavior to order diagnostic tests, prescribe medicines and recommend surgery and other types of treatment. This makes the study of out-of-pocket expenditure relevant for setting health policies and assessing their effect on economic development and poverty reduction in a country. There is ample literature on the study of out-of-pocket expenditure (Wagstaff and van Doorlaer, 2003), the methodological foundation (Xu et al, 2007) and its current application to assess progress on UHC (Wagstaff et al, 2019).

More specifically, the study of medicines OOP within health OOP expenditure is very relevant as evidence shows that in many countries the majority of health OOP is for medicines: for instance, data from Southeast Asia show that in India 90% of out of pocket expenditure on health is on medicines, in Nepal it is 88%, and Indonesia it is 78%. Not only in relation to health OOP are medicines OOP expenditure relevant; they are also important as a proportion of the total household expenditure: according to the World Health Survey, up to 9·5% of the total expenditure of poorer households in LMICs is spent on medicines, far higher than the 3·5% expended by poorer households in high income countries (HICs) (Wagner et al, Health Policy). Approximately half (41%-56%) of households spent 100% of their health care expenses on medicines (Wagner et al, Health Policy).

Since medicines OOP is a larger driver of the overall health OOP and countries have made a commitment to Universal Health Coverage, studies focusing on medicines OOP become increasingly relevant.”

As we explain, focusing the manuscript on medicines OOP is important. At the same time, our manuscript reports on health OOP. While medicines OOP is the central theme of the manuscript, we believe it gives sufficient attention to health OOP as well.

2. Methods: Authors used budget share approach and provided the varying incidence of CHE. Findings suggest, at 10% threshold, the CHE of fourth quintile is higher than first quintile. This is primarily due to limitations of budget share approach in estimating the CHE. I suggest the authors should use the capacity to pay approach as the expd and income survey provides required variable for estimation of CHE. In table 4, authors must mention whether it is incidence of CHE?

RESPONSE: Thank you for this suggestion. We have now revised the analysis and incorporated the reviewer’s suggestion to include the capacity to pay approach. We have also added an analysis that uses the disposable household income as described by Xu et al. 2007. The methods section under the main outcome variables reads: “4) based in Wagstaff’s approach [2,8], the share of OOPE on medicines out of total household expenditure (reported as a ratio between zero and one), and 5) based in Xu’s approach [9,21], the share of OOPE on medicines out-off household’s capacity to pay (reported also as a ratio between zero and one).”

As expected the disparity between households in the poorest and the wealthiest quintile is even larger when using the capacity to pay approach regarding the proportion of medicine expenditure out of total household expenditure. The results section now reads: “Expressed as the capacity of the household to pay, the disparity between households in the poorest and the wealthiest quintile is even larger: while poor households spent nearly a one-fifth (19.85%) of their disposable income on medicines, the wealthiest households spent 7.6% on medicines.”

“Using the capacity to pay approach to calculate the proportion of medicines out of total household expenditure, we show that 71% of households spent more than 10% of their disposable income on medicines. Large disparities exist between the poorest and the wealthiest households: while 23.8% of the poorest households spent over 20% of their disposable household income on medicines, only 12.1% of the wealthiest households do so.”

We believe that following the reviewer’s suggestion and complementing our analysis using these two approaches have substantially strengthened the paper.

3. Adjusted per equivalent adult and out of those households 192 who spent on health, their median total monthly health-related OOPE was US$3.14. I am not sure whether is is per capita or per household. I think it should be per capita. Authors must mention how they derive adult equivalent scale? What weight they assign to children? Need mentioning

RESPONSE: Thank you for the question by the reviewer. We agree that it is important to be clear on the meaning of “adult equivalent” and how it was calculated:

“Adult equivalent’ adjusts for the economy of scale in consumption – a household with three members, including children, for example, does not consume three times that of a one-person household. The equivalence scale considers the age of the household members and establishes a standardization that allows comparison.”

Per ‘Adult equivalent’ adjustment is standard in welfare studies. It has been developed to account for the variation in the consumption of goods by household members on the basis of age and economies of scale of consumption. For more information we refer the reviewer to the following resource which we have cited in the paper: Haughton J, Khandker SR. Handbook on Poverty and Inequality. 1st ed. World Bank Publications; 2009.

4. Authors disease classification is not adequate. Moreover, they must mention as limitation of self-reported diseases as expd survey typically collect self-reported data

RESPONSE: We thank the reviewer for this note. We have added to the limitation the following sentence: “Fifth, the data on health are self-reported which can introduce errors due to the fact that the respondent may not know or remember health related information. This could have affected our analysis of expenditure by type of disease.”

5. Abstract: Background and objective. There is only one sentence of aim. Keep a line on background

RESPONSE: We have added a line “High OOPE increases the probability that households will become impoverished or will forgo needed care.”

REFERENCE

Rahman, S. Universal health coverage in Bangladesh: the challenges. The Financial Express 2021, Sep 2. https://thefinancialexpress.com.bd/views/universal-health-coverage-in-bangladesh-the-challenges-1549037007 [accessed Sep 1 2021]

According to the World Health Survey, up to 9·5% of the total expenditure of poorer households in LMICs is spent on medicines, far higher than the 3·5% expended by poorer households in high income countries (HICs). Approximately half (41%-56%) of households spent 100% of their health care expenses on medicines (Wagner et al, Health Policy).

17 Dec 2021

8 Feb 2022

We would like to thank the reviewer for the constructive comments. We have carefully addressed the each of the reviewer’s comments in this revision. We believe that after incorporating your thoughtful feedback, our manuscript has been greatly improved. Our responses to the reviewer’s comment can be found in bold.

Reviewer #2: While you have estimated the catastrophic health spending using capacity to pay approach, you have not estimated the estimates using threshold of 40%. In CTP approach a single threshold of 40% is used. If you will do so, you will find different estimates by quintile. Your revised table 4 is not correct. It is based on budget share approach Pl read the following reference to get complete idea of estimating CHE using CTP approach

https://equityhealthj.biomedcentral.com/articles/10.1186/s12939-021-01421-6

Addressing data and methodological limitations in estimating catastrophic health spending and impoverishment in India, 2004–18

Authors’ response: Thank you very much for your comment. In response to your suggestion and following the article you shared with us, we disaggregated Table 4 into two separate tables (see below), the first (Table 4) for the adjusted ratio of spending on medicines with respect to total household spending, and the second (Table 5) for the adjusted ratio of spending on medicines with respect to of households' ability to pay. In Table 5 we have the additional threshold of 40% of the households' ability to pay. We clarify this point in the methods section as follows:

[…] Finally, based on the first regression analysis results, we estimated the adjusted share of OOPE on medicines out of total household expenditure (considering the following thresholds of the total household expenditure: 10, 15, 20, 30%); and out of a household’s capacity to pay by considering different thresholds (10, 15, 20, 30 and 40% of the total household expenditure) [31] and according to the quintile of household expenditure per equivalent adult […]

We have added further descriptions of the results:

“The disparity between the lowest and wealthiest households increases with higher thresholds: while 0.5% of household spent over 40% of their disponible household income on medicines, only 0.2% of the wealthiest households do so.”

Reference: 31. Mohanty, S.K., Dwivedi, L.K. Addressing data and methodological limitations in estimating catastrophic health spending and impoverishment in India, 2004–18. Int J Equity Health 20, 85 (2021). https://doi.org/10.1186/s12939-021-01421-6

We have included the revised Table 4 and a new Table 5 in the manuscript.

Submitted filename: Response to Reviewers .docx

Click here for additional data file.

22 Apr 2022

20 May 2022

Reviewer #1: The paper has been written well, and it analyzes the patterns, trends, and determinants of OOP expenses for medicines in Bangladeshi households using nation-wide survey data from the HIES. The paper provides a lot of empirical evidence on the subject using appropriate analytical methods. For instance, the paper analyses trends and determinants of OOP household expenses on medicines across different income quintiles.

RESPONSE: Thank you.

Please check whether you could attend the followings which may further improve your paper:

• From the view point of public policy, it may be important to discuss about income elasticity of OOP expenses on medicine. Can you calculate the elasticity using double-log regression framework? Analyzing the data from elasticity ground may be more important than analyzing the same across different income quintiles. This type of an analysis may be useful to determine whether medicine is an essential or a luxury for Bangladeshi households.

RESPONSE: As the reviewer suggests we have incorporated this aspect in the discussion section.

“Sixth, this study does not analyze medicines prices or price elasticity as a factor affecting medicines out-of-pocket expenditure because information on medicine prices is not collected as part of the survey. Linking outside data source with the data set used in this study is challenging as individuals in the survey do not report on specific products that they purchased.”

Moreover, it is noteworthy to mention to the reviewer that medicine price information collected through market studies often do not reflect the prices patients pay for their medicines.

• The price-levels of medicines play a significant role in determining the extent of OOP expenses on medicines. Did take price-factor into account when analyzing? If not, the current analysis is lop-sided, and resultantly, the paper is missing a big picture coming from the price factor. You may use the data from price indices of medicine or any other source to capture the impact of prices on OOP expenses on medicines.

RESPONSE: Thank you for this comment. We have included an explanation in the discussion section about this point. See response to the comment above.

It is important to note that the National Household and Expenditure Survey asks the respondents to provide the information on the total costs of their medicines over the past 30 days. It does not ask the respondents about the price of each of the medicine purchased. We do not know how many medicines individuals consumed.

It is also important to take into consideration that the amounts analyzed are at current market prices and that there are no reliable price indices for medicines in the different geographical areas, so the price factor mentioned could not be effectively controlled. Furthermore, it is important to mention that correcting for inflation would not modify the results obtained since this survey is cross-sectional and not a comparative one of different points of time.

• The covariates that the paper considered represent only the demand-side of medicines. The paper misses the story coming from supply-side factors in determining OOP expenses on medicines. Availability of public/private avenues for treatments, proximity to such avenues, health insurance coverage, facilities available in public/private avenues, influence of doctors are some of supply-side factors. Can you strengthen the analysis by incorporating the information pertaining to those supply-side factors? Accordingly, you may strengthen the sections of discussions and conclusions by describing the situation and the role of such supply-side factors in determining OOP expenses on medicines.

RESPONSE: The reviewer is correct that this analysis focuses on the demand-side. It would be very interesting to link the demand side data with the supply side. However, this is outside the scope of this manuscript. In addition, it is noteworthy, that linking supply side factors such as availability of medicines in public and private sector and prescriber behavior in the catchment area of individual households are often absent or not updated. We have included this limitation in the discussion sector.

“Finally, this study focuses on determinants of out-of-pocket expenditure that are recorded within this survey. We did not link other databases to study determinants of expenditure such as distance to the nearest government/private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power.”

Reviewer #2: Thank you for implementing necessary changes. the paper has improved in content and presentation. Limitations of the study may be highlighted.

RESPONSE: Thank you.

________________________________________

Submitted filename: Responses_PLoS One Bangladesh OOP medicines_Final.docx

Click here for additional data file.

13 Jul 2022

7 Aug 2022

Dear Editor,

We would like to thank you for your response to our letter to the Editor on July 26 (see the Annex to this letter). You asked us to edit the language. In response to this request, we have significantly edited the language. You will see these changes in the track-change version that we uploaded.

Additional responses to the reviewer’s comments can be found in our letter to the Editor on July 15, 2022 (also in the Annex).

Annex: Recent correspondence with PLoS

From: plosone

Sent: Tuesday, July 26, 2022 5:49 PM

To: Wirtz, Veronika

Subject: RE: PLOS ONE Decision: Revision required [PONE-D-21-14254R3]

Dear Dr. Wirtz,

Thank you for your patience, here is the response of the AE regarding to your concern in reviewers comment.

******

Thanks for your email. I have gone through the comments and responses. I think the authors are valid in their argument. But they still need to edit the language. Please ask them to do so, and then I will provide my final decision.

*******

Feel free to message me if you have any further question or concern

Kind regards,

Liezl Callo

Straive Editorial Assistant

PLOS ONE | plosone@plos.org

Empowering researchers to transform science

Boston, July 15,2022

Dear Editor

We are responding to the most recent request for changes to this manuscript. Respectfully, we would like to ask that the Journal make a final determination as to whether or not our manuscript [PONE-D-21-14254R3] - [EMID:e4ab9e939629ff6c] is acceptable for publication without making any additional changes. One of the reviewers, Ajantha Sisira Kumara, keeps insisting (now for the third time) that we provide two additional analyses. In our view, these analyses are both out of the scope of our study and, in fact, are not feasible. We have explained our reasoning in our detailed responses over the past three re-submissions, which can be found below. Indeed, we believe these requests would amount to writing a separate paper.

It is noteworthy that all other reviewers over the course of the three revisions have been satisfied with our responses and how we incorporated their comments except Ajantha Sisira Kumara who keeps coming back with the same request over and over regardless of our explanation.

Since it is clear from Kumara’s comments that nothing is wrong with the current manuscript and that this request for additional analyses is complementary and not essential, we respectfully ask the Journal to please accept the manuscript at this stage.

Sincerely,

Veronika J. Wirtz, B.Pharm., M.Sc., Ph.D., FISPE

Professor - Department of Global Health, Boston University School of Public Health

Director, World Health Organization Collaborating Center in Pharmaceutical Policy

vwirtz@bu.edu; T: +1 617 358 3046

https://www.bu.edu/sph/profile/veronika-wirtz/

pronouns: she/her/hers @VeroWirtz

1st Revision October 2021

Ajantha Sisira Kumara: “The importance of supply-side factors in determining OOP expenses for medicines is completely ignored from the study, making it lopsided. Do you have data in HIES on healthcare supply-side factors? For instance, distance to the nearest government /private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power. My recommendation would be incorporate such variables and see whether they play a role in Bangladesh like in other contexts.”

RESPONSE: The purpose of this study is to analyze household and individual factors that are associated with medicines OOPE. The reviewer is correct that additional data on the supply side could be added to complete this analysis. However, the survey does not provide geospatial data at the unit of analysis which is the household. Adding supply side factors such as nearest public dispensary, etc. would be approximations based on the smallest geographical area. We have added in the discussion a section where we expand on the need for analysis of the supply-side factors.

“Finally, this study focuses on determinants of out-of-pocket expenditure that are recorded within this survey. We did not link other databases to study determinants of expenditure that are not included in the survey such as distance to the nearest government /private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power.”

2nd Revision in December 2022

All comments successfully responded to.

3rd Revision May 2022

Ajantha Sisira Kumara: “Please check whether you could attend the followings which may further improve your paper: • From the viewpoint of public policy, it may be important to discuss about income elasticity of OOP expenses on medicine. Can you calculate the elasticity using double-log regression framework? Analyzing the data from elasticity ground may be more important than analyzing the same across different income quintiles. This type of an analysis may be useful to determine whether medicine is an essential or a luxury for Bangladeshi households.”

RESPONSE: As the reviewer suggests we have incorporated this aspect in the discussion section.

“Sixth, this study does not analyze medicines prices or price elasticity as a factor affecting medicines out-of-pocket expenditure because information on medicine prices is not collected as part of the survey. Linking outside data source with the data set used in this study is challenging as individuals in the survey do not report on specific products that they purchased.”

Moreover, it is noteworthy to mention to the reviewer that medicine price information collected through market studies often do not reflect the prices patients pay for their medicines.

Ajantha Sisira Kumara: “The covariates that the paper considered represent only the demand-side of medicines. The paper misses the story coming from supply-side factors in determining OOP expenses on medicines. Availability of public/private avenues for treatments, proximity to such avenues, health insurance coverage, facilities available in public/private avenues, influence of doctors are some of supply-side factors. Can you strengthen the analysis by incorporating the information pertaining to those supply-side factors? Accordingly, you may strengthen the sections of discussions and conclusions by describing the situation and the role of such supply-side factors in determining OOP expenses on medicines.”

RESPONSE: The reviewer is correct that this analysis focuses on the demand-side. It would be very interesting to link the demand side data with the supply side. However, this is outside the scope of this manuscript. In addition, it is noteworthy, that linking supply side factors such as availability of medicines in public and private sector and prescriber behavior in the catchment area of individual households are often absent or not updated. We have included this limitation in the discussion sector.

“Finally, this study focuses on determinants of out-of-pocket expenditure that are recorded within this survey. We did not link other databases to study determinants of expenditure such as distance to the nearest government/private hospital, nearest public/private clinics, nearest public/private dispensary, and availability of doctors, specialists, dentists per 1,000 of population as they might have an explanatory power.”

4th Revision July 2022

Ajantha Sisira Kumara: “However, I still believe that you can calculate income elasticity of medical expenditure at the household-level. For this purpose, you no need to have the data pertaining to unit prices of medicine. I hope your dataset has household income or otherwise, it can be proxied by household expenditure.

Further, I still believe that you can incorporate health-related supply-side factors to calculate their impact on OOP expenditure on medicine. Those data are available at the country's macro-level and can be collected from annual reports of relevant ministries.

I firmly believe that the paper will be in a better shape after you address at least these two issues. Congratulations!”

Submitted filename: Responses to the editors and reviewers comments_August 7.pdf

Click here for additional data file.

2 Sep 2022

Out-of-pocket expenditure on medicines in Bangladesh: an analysis of the national household income and expenditure survey 2016-17

PONE-D-21-14254R4

Dear Dr. Wirtz,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Mohammad Bellal Hossain

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

6 Sep 2022

PONE-D-21-14254R4

Out-of-pocket expenditure on medicines in Bangladesh: an analysis of the national household income and expenditure survey 2016-17

Dear Dr. Wirtz:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Mohammad Bellal Hossain

Academic Editor

PLOS ONE