Appl Health Econ Health Policy DOI 10.1007/s40258-014-0100-2

ORIGINAL RESEARCH ARTICLE

Disease-Specific Impoverishment Impact of Out-of-Pocket Payments for Health Care: Evidence from Rural Bangladesh Syed Abdul Hamid • Syed M. Ahsan Afroza Begum

•

Ó Springer International Publishing Switzerland 2014

Abstract Background Analysing disease-specific impoverishment impact of out-of-pocket (OOP) payments for health care is crucial for priority setting in any informed policy

This paper was drafted while Syed M. Ahsan was working as the Team Leader, Syed Abdul Hamid as the Project Coordinator and Afroza Begum as a Research Associate at the Institute of Microfinance (InM), Dhaka, Bangladesh. S. A. Hamid (&) University of Dhaka, Dhaka, Bangladesh e-mail: [email protected]; [email protected] S. M. Ahsan Concordia University, Montreal, QC, Canada e-mail: [email protected] S. M. Ahsan South Asian University, New Delhi, India S. M. Ahsan CESifo, Munich, Germany A. Begum University of Chittagong, Chittagong, Bangladesh e-mail: [email protected]

discussion. Lack of evidence, particularly in the Bangladesh context, motivates our paper. Objective To examine disease-specific impoverishment impact of OOP payments for health care. Methods The paper estimates the poverty impact of OOP payments by comparing the difference between the average level of headcount poverty and poverty gap with and without health care payments. We used primary data drawn from 3,941 households, distributed over 120 villages of seven districts in Bangladesh during August–September 2009. Findings We find that OOP outlays annually push 3.4 % households into poverty. The corresponding figures for those who had non-communicable diseases (NCDs), chronic illness, hospitalization and catastrophic illness were 4.61, 4.65, 14.53 and 17.33 %, respectively. Note that NCDs are the principal reason behind the latter two situations (about 88 % and 85 % of cases, respectively). Looking into individual categories of NCDs we found that major contribution to headcount impoverishment arose out of illnesses such as cholecystectomy, mental disorder, kidney disease, cancer and appendectomy. The intensity of impoverishment is the largest among the hospitalized patients, and more individually among cancer patients. Conclusions The poverty impact of OOP outlays for health care, in general, is quite high. However, it is especially high for NCDs, particularly for chronic NCDs and those requiring immediate surgical procedures. Hence, these illnesses should be given more priority for policy framing. In addition to suggesting some ex-ante measures (e.g. raising awareness regarding the risk factors causing NCDs), the paper argues for reforms to enhance efficiency in the public health care facilities and increasing the quality of public health care.

S. A. Hamid et al.

Key Points for Decision Makers Out-of-pocket outlays annually push 3.4 % households into poverty, while non-communicable diseases (NCDs), particularly chronic ones and those requiring immediate surgery, inflict substantially higher burdens on impoverishment. In view of the non-viability of voluntary health insurance, the paper proposes reforms of both financing (e.g.,carefully designed system of affordable user fees) and of governance (strengthening of local government and its integration in the health care delivery system at the local level) of public health care facilities. The paper also argues for the gradual adoption of effective regulation of the price of essential drugs, and restrictions in the sale of over-the-counter drugs.

1 Introduction The poor population in low-income countries mainly finance health care from out-of-pocket (OOP) payments that severely affect their consumption during periods of major illness, or forces them to forego treatment, which raises the chance of long-term deterioration of health and earning capacity [1, 2]. Costs of health care are therefore claimed to be a major cause of poverty in low-income countries [3], and is also a cause of aggravating poverty [4– 6].1 Hence, it seems that OOP payment is a major threat to the success of national poverty reduction initiatives of developing nations. However, OOP outlays would plausibly vary across illness categories (e.g. between non-communicable diseases [NCDs], and communicable diseases [CDs], and accordingly one would expect the impoverishment impact of OOP payments to also depend on the type of illness. A growing body of literature in the developmental context have documented the general impoverishment impact of OOP payments [6–13]. Although analysing the disease-specific impoverishment impact of OOP payments is crucial for priority setting in any informed policy discussion (e.g. especial focus needs to be given to the most burdensome illnesses), there is hardly any evidence, particularly in the developmental context. Van Doorslaer et al. [6], who analysed the data of 11 Asian countries, including 1

Of course the choice of the coping strategy also has implications on poverty outcome [7].

Bangladesh, only measured the headcount impact of OOP outlays on poverty, not disease-specific poverty, using the Household Income and Expenditure Survey (HIES) data of the Bangladesh Bureau of Statistics (BBS). Indeed, the lack of evidence in this arena, particularly in the Bangladesh context, motivates the present paper, which aims at examining the disease-specific impoverishment impact of OOP payments, giving a particular focus on NCDs and chronic illnesses; hence, contribute to the literature by providing such evidence.

2 Methodology 2.1 Data This paper uses the data on OOP payments for health care obtained from the baseline survey of a longitudinal research project conducted in 2009, which successfully collected data from 3,941 (out of a target of 4,010) rural households (accounting for 19,424 individuals) from 120 villages spread over seven of 14 districts in rural Bangladesh where Grameen Kalyan (GK), a social business company affiliated with the Grameen Bank (GB), had been operating a prepaid card-based micro health insurance (MHI) scheme. The survey used a program-control design such that ten health care delivery centres were selected purposively taking into consideration a suitable mix of old and new centres and the geographic variation among these locations. One comparable ‘Union Council’, the smallest civil administrative unit in Bangladesh, adjacent to each GK programme centre was then selected purposively to serve as the ‘control’ area in question. The control areas lay wholly outside the radius of the GK operational boundary but shared similar characteristics in all other aspects. A sample of seven villages was randomly selected from each of the ten programme strata and five villages from each of the ten control strata from a listing of all the villages in both these strata. Thus, the survey covered 120 (70 programme and 50 control) villages.2 In the next stage, a census was conducted in all the selected villages and about 30,000 households were thus listed. In the programme villages, the listed households were divided into two groups: GK MHI card holders (CH) and non-card holders (NCH). In each programme stratum, 150 households were 2

In most cases the programme area included a part of the adjacent administrative unions in addition to the core union where the health care centre of the Grameen Kalyan was situated. On the other hand, as control, we selected one administrative union comparable to the core union of the programme area. Hence, the number of villages was higher in programme areas compared with control areas. As we sampled the same proportion of villages from both programme and control areas, some difference is inevitable.

Impoverishment Impact of Out-of-Pocket Payments

randomly selected from the NCH and 105 from the CH group, except one area where only 65 CHs were available. A total of 2,510 households (1,010 CHs and 1,500 NCHs) were thus selected from the programme areas. In each control stratum, 150 households were randomly selected from the listed households, yielding a total of 1,500 households for all control areas. Thus, the total target sample stood at 4,010 households (2,510 and 1,500 from the programme and control areas, respectively). A series of questions regarding OOP payments were posed to the respondents for each episode of illness within the household over the 12 months preceding the survey. Ideally, the recall period for healthcare utilization should be set in order to satisfy the dual objectives: minimizing the recall bias and maximizing the sample of target subjects [6]. In the literature regarding impoverishment impact of OOP payments, some have used a 12-month recall period for both inpatient and outpatient cases [8], while others have preferred a 1-month recall for outpatients and 12-month recall for hospitalization [10, 13]. In a study of 11 Asian countries, van Doorslaer et al. [6] used different recall periods (varying from 1 month to 12 months) for different countries; for Bangladesh they used 1 month for both cases. Similarly, Yardim et al. [12] used 1 month for both inpatient and outpatient care. Although the present survey has collected information both over 90 days as well as for 12 months, the analysis presented here is based on 12-month data to get the adequate number of observations. In addition, we collected information about morbidity, health care seeking pattern, demographic condition, occupation, education, income, expenditure, assets, borrowing, etc., from the household. Household heads were the main respondents. Note that the respondents were not directly asked about any particular diseases (e.g. cholecystectomy, appendectomy, cancer). The disease categories were determined by recoding the original responses which were collected in the respondent’s mother language and after cross-matching with the symptoms. The actual questionnaire was developed in Bengali and the disease categories were written in layman’s language. There were medical doctors in the research team who helped to categorize the illnesses after cross-matching the symptoms. 2.2 Measurement of Out-of-Pocket (OOP) Payments In measuring OOP payments, we have mainly considered the payments for direct medical input used by the sick. More precisely, ‘the OOP payments’ variable was constructed by adding the expenses that a household incurred for consultations, drugs, diagnostic tests, surgical operations, and bed charge for each episode of illness for the 12 months preceding the survey. These types of OOP payments may be termed as direct OOP payments. Total

OOP payments may also be constructed by adding the payments for transportation and other items (food, lodging, accommodation and unofficial fees) with direct OOP payments. OOP payments for CDs, NCDs, chronic illnesses, acute illnesses, hospitalization and catastrophic illnesses were similarly constructed by adding the relevant expenses incurred for each episode of illness in each such category.3 2.3 Impoverishment Impact of OOP Health Payments We have estimated the impoverishment impact of OOP payments for each category of illness by comparing the difference between the average level of head count poverty (H) or poverty gap (G, which is the intensity of poverty or poverty deepening) with and without OOP, as has been done in earlier studies [4, 6, 9].4 Prepayment headcount poverty (H prepayment ) was calculated by comparing per capita household expenditure (including OOP payments for health care) with a poverty line estimated by the authors.5 Similarly, the postpayment headcount poverty (H postpayment ) was measured by comparing per capita household expenditure (excluding direct OOP payments for health care) with the poverty line.6 Assume zi to be per day per capita expenditure (including OOP payments for health care), yi is per day per capita OOP payments, PL is the poverty line and n is the number of individuals. Prepayment and postpayment headcount poverty measures can be expressed, respectively, as:

3

We used both WHO fact sheets and Council for Medical Schemes guidelines to define chronic diseases. 4 Headcount poverty measures the percentage of individuals or households living below the poverty line, while poverty gap measures poverty deepening or intensity of poverty (the amount by which the poor households fall short of the poverty line). 5 We used both food and non-food expenditure as a proxy for household income. For measuring food expenditure, we considered expenditure on the food bundle consumed by the household for the week preceding the survey. We considered expenditure for non-food consumption against the following items: clothing, toiletries, cookware, blankets, furniture, lamp, torch light, candle, match, kerosene, electricity, transportation, fuel, maintenance and repair of household effects, taxes, donation and tolls, recreation, tobacco, tuition fees, stationeries, mobile and land telephone bills, festivals and traditional ceremonies, electronic equipments and health expenses (both direct and indirect). Note that we included health expenses (both direct and indirect) for prepayment poverty measurement, and excluded direct health expenses for the postpayment measurement. 6 The prepayment health care financing (insurance) mechanism does not usually cover expenses such as transportation cost, and cost of food, lodging, accommodation and unofficial fees. In order to link policy discussion with the insurance mechanism we did not include such expenses. Thus, we have meant ‘direct out-of-pocket payments’ as OOP payments in the remaining part of the paper.

S. A. Hamid et al.

H prepayment ¼ 1=n

n X

ai ;

such that

3 Results

ai ¼ 0

i¼1

if

zi  PL and ai ¼ 1; if

H postpayment ¼ 1=n

n X

bi ;

ð1Þ

zi \PL such that

bi ¼ 0

i¼1

if

ðzi  yi Þ  PL and bi ¼ 1 otherwise

ð2Þ

Similarly, the prepayment and postpayment poverty gap can be defined, respectively, as: n X Gprepayment ¼ 1=n ci ðPL  zi Þ; ð3Þ i¼1

Gpostpayment ¼ 1=n

n X

ci fPL  ðzi  yi Þg;

ð4Þ

i¼1

where, ci = 1 (i.e. the household is poor) if zi \ PL and ci = 0 (i.e. the household is non-poor) if zi C PL. The headcount poverty is higher in Eq. (2) compared with Eq. (1) if OOP payments are positive. Similarly, the poverty gap is higher in Eq. (4) compared with Eq. (3). Thus, the difference between Eq. (2) and Eq. (1) depicts the headcount impoverishment impact of OOP payments. Similarly, the difference between Eq. (4) and Eq. (3) illustrates the intensity of poverty on account of OOP payments. More precisely, the headcount and poverty gap impoverishment impact of OOP payments can be expressed, respectively, as: ðH postpayment  H prepayment Þ and ðGpostpayment  Gprepayment Þ. It is often helpful to use the normalized poverty gap (the size of poverty gap in relation to poverty line, ðGpostpayment  Gprepayment Þ=PL , for a comparative analysis. 2.4 Poverty Line We estimated poverty line expenditure by using the costs of basic need (CBN) approach where we used, following Ravallion and Sen [14], the cost of a normative food bundle (consisting of rice, wheat, pulses, milk, mustered oil, beef, fish, potato, and both leafy and non-leafy vegetables) which provides the minimal nutritional requirement of 2,122 kcal per day per capita. We computed the food poverty line, as done in the report of the Bangladesh HIES, 2005 [15], by adding the products of the price of each item and the respective quantities given in the bundle. This resulted in Bangladesh Taka (BDT)33.42 (US$0.48, using 2009 exchange rates) as the food poverty threshold. We next estimate the ‘upper non-food allowance’ by taking the median amount spent on non-food items by households whose per capita food expenditure is close to the food poverty line. The estimated poverty line expenditure per day per capita turned out to be BDT61 (i.e. equivalent to US$0.88), which is taken as PL in the estimations to be presented below.

3.1 Sample Characteristics A total of 3,941 households, out of a target of 4,010, were successfully interviewed in this study. The overall response rate was 98.28 % (98.66 % in programme areas and 97.60 % in control areas). Household heads were the respondents in most cases (83 %) and spouses in 15 % of cases. Most households (about 88 %) were male-headed. Average education level of the household head was 3.2 years, and the average age about 46 years. The average household size was 4.45. The mean per capita daily consumption (both food and non-food) was about BDT66 (or US$0.96). About 30 % of the household heads were engaged in agriculture, followed by day labour (about 16 %) and small business (about 14 %). 3.2 Pattern of Morbidity and Care-Seeking The survey enquired whether any member in the household had suffered from any acute or chronic condition during the 12 months preceding the interview. They were also asked whether they had received any treatment for their condition and the type of care they had received. The survey covered 19,424 individuals, of which about 33 % had experienced some form of self-reported morbidity over 12 months (Fig. 2 in the Appendix). About 88 % of households reported at least one episode of illness, and about 55 % (or 48 % of the sampled households) had more than one (where about 35 % had two episodes and about 20 % had three or more) in 1 year. About one-third of the ill suffered from ‘general cough and fever’. Other major symptoms were gastrointestinal disorder, abdominal pain, diarrhoea, typhoid, headache, blood pressure, skin diseases and dysentery. The incidence of CDs and NCDs was about equal in the sample, while about 80 % of the patients suffered from acute conditions, and the remaining 20 % faced chronic conditions. The overwhelming majority (about 98 %) of the ill sought some kind of care, although most (95 %) care seekers went for outpatient services (Fig. 2 in the Appendix). 3.3 OOP Payments In the present dataset it is seen that OOP outlays and OOP payments on account of drugs for all episodes of illnesses during the 12 months preceding the survey stood at US$60.87 and US$36.41, respectively, per affected household (Table 1).7 Thus, the cost of drugs appears to be the 7

OOP payments and OOP costs on account of drugs for all episodes of illnesses over 12 months, when averaged over all sampled households, decline to US$52.74 and US$31.49, respectively (not shown in the table).

Impoverishment Impact of Out-of-Pocket Payments Table 1 OOP payments per affected household by expenditure quintiles Quintile

Mean OOP costs (US$) over 12 months

Mean OOP costs (US$) on account of drugs over 12 months

OOP costs as a percentage of total household (both food and non-food) expenditure

First (the poorest)

38.51 (66.29)

26.16 (45.14)

4.49 (664)

7.44 (664)

67.93 (664)

Second

42.48 (80.93)

27.12 (44.91)

3.69 (693)

6.52 (693)

63.83 (693)

Third Fourth

61.19 (117.30) 65.33 (128.71)

37.83 (73.74) 40.59 (82.19)

4.50 (672) 3.98 (696)

8.22 (672) 7.80 (696)

61.82 (672) 62.13 (696)

Fifth (the richest)

96.07 (224.00)

49.97 (93.30)

4.06 (690)

10.00 (690)

52.01 (690)

Programme

56.99 (126.01)

34.72 (71.20)

3.78 (2,121)

7.61 (2,121)

Control

67.26 (153.46)

39.16 (71.41)

4.69 (1,294)

9.14 (1,294)

58.22 (1,294)

Card holders

57.99 (139.62)

35.64 (81.46)

3.49 (818)

7.55 (818)

61.46 (818)

Non-card holders

56.35 (116.72)

34.16 (63.96)

3.99 (1,303)

7.64 (1,303)

60.62 (1,303)

Total

60.87 (137.13)

36.41 (71.30)

4.11 (3,415)

8.18 (3,415)

59.81 (3,415)

OOP costs as a percentage of food expenditure

Payments for drugs as a percentage of OOP costs

60.94 (2,121)

Consumption expenditure has been scaled up to 12 months Figures in parentheses are standard errors in columns 2–3 and number of observations in columns 4–6 One US$ was equivalent to BDT69 while the survey was conducted (mid-2009). This rate is used throughout the paper OOP out-of-pocket, BDT Bangladesh Taka

major component (about 60 %) of OOP payments. In terms of annual expenses, OOP payments are about 4 % of the value of total household consumption (food and non-food) and about 8 % of the value of food consumption (Table 1). Although OOP payments show a definite positive pattern as one moves up the expenditure quintiles, there is no systematic variation when considered as a share of either total or food expenditures. However, the share of drug costs in OOP payments shows an unambiguous negative pattern across quintiles, and the difference (68 % vs. 52 % between the poorest and the richest quintiles, respectively) is significant (p \ 0.05). Health care for the very poor therefore appears to be largely synonymous with ‘accessing drugs’. Although there is a significant difference in both categories of OOP payments (at 5 % and 10 %, respectively) between the programme and control areas, the difference is negligible between GK CHs and NCHs. For the broad category of illnesses, it is seen that OOP payments per episode of CDs, NCDs and accidents and injuries (AI) was US$12.25, US$52.12 and US$61.54, respectively (Table 2). Quite plausibly, therefore, NCDs and AI involved significantly (p \ 0.01) higher OOP costs than CDs per episode. Presumably a good part of the AI expenses would be for whatever ‘emergency care’ that was available at the time. OOP costs per episode of chronic condition (US$76.99) was significantly (p \ 0.01) higher than for acute conditions (US$21.64), and similarly for an episode of inpatient vs. outpatient care (US$222.61 and US$22.87, respectively). Insofar as drug expenses (as a share of OOP costs) are concerned, the pattern appears most stable (over 59 %) for the disease nature (i.e. acute vs. chronic). However, when interpreted by the type of illness along the CDs/NCDs/AI

orientation, drugs expense ratio rises to 74 % for CDs, while staying at 56 % and 60 %, respectively, for NCDs and AI. Let us turn to the incidence of illness that leads to catastrophic health expenditure (CHE) where we have used total household expenditure as the capacity to pay, and considered 10 % as the threshold level. We included all health expenditure within the family (involving multiple members as appropriate) in measuring the ratio of household consumption for the year. About 10 % of the sampled households (or 12 % of the affected households), namely 404, incurred CHE at the 10 % threshold level over the 12 months preceding the survey.8 As may be anticipated from the above discussion, the poorest quintile again emerges as the group having suffered the most from such high-level expenses as a share of household consumption (11.80 % vs. 9.53 % for the poorest and the richest quintiles, respectively), although the difference is not statistically significant (tables are not shown). 3.4 Impact on Poverty 3.4.1 Headcount Impoverishment Impact of OOP Payments It is seen that overall prepayment headcount poverty is 56.34 % and postpayment (deduction of expenses for health care from total household expenditure) headcount

8

The actual number of households experiencing catastrophic healthcare expenditure at the 10 % level comes to 404, which is about 10 % of the sample figure (i.e. 3,937), but when expressed as a share of all households who actually sought medical treatment for illness (i.e. 3,419), the ratio rises to about 12 %.

S. A. Hamid et al. Table 2 OOP payments by per episode of illness Patient attributes

Mean OOP costs (US$) over 12 months

Mean expenses (US$) on drugs

Expenses on drugs as a percentage of OOP costs

Type of illness Communicable diseases

12.25 (28.10) [3,120]

9.03 (20.42) [3,120]

73.73

Non-communicable diseases

52.12 (127.48) [3,093]

29.39 (62.09) [3,093]

56.40

Accidental diseases and injuries

61.54 (113.25) [139]

36.68 (74.06) [139]

59.61

Acute diseases

21.64 (57.93) [5,078]

13.03 (29.80) [5,078]

60.21

Chronic diseases Type of care

76.99 (170.26) [1,274]

45.55 (84.43) [1,274]

59.17

Inpatient

222.61 (298.10) [314]

89.00 (118.67) [314]

39.98

Outpatient

22.87 (53.59) [6,038]

15.94 (37.87) [6,038]

69.71

Condition of illness

Figures in round parentheses are standard errors and those in square brackets are the number of observations OOP out-of-pocket

poverty is 59.74 % (Table 3). Thus, 3.4 % of households annually fall into poverty due to payments for health care. The impoverishment impact of payments for health care can be observed by plotting prepayment as well as postpayment consumption expenditure against the cumulative proportion of the households ranked by prepayment consumption expenditure (per day per capita) in a Pen’s parade graph [4]. Figure 1 shows the Pen’s parade graph for the pre-and postpayment consumption expenditure. The point at which the prepayment parade intersects the poverty line measures the prepayment headcount poverty, which is about 56.34 %. The ‘paint drops’ from the prepayment curve portray that payments for health care drag the consumption expenditure down the prepayment level. The lower boundary of the ‘paint drops’ plots the postpayment curve. The postpayment headcount poverty (which is 59.74 %) is depicted by the proportion below the poverty line. The difference of the two headcounts (3.4 percentage points) measured on the x-axis thus emerges as the headcount poverty impact of OOP payments. Turning to the broad categories of illnesses, we see that, after accounting for payment for health care, about 4.65 % of those who had chronic illnesses fall into poverty, while the corresponding figure for acute illnesses, NCDs, CDs and AI is 2.66, 4.61, 0.95 and 4.48 %, respectively (Table 3). Health care spending pushes 14.53 and 2.74 % of households into poverty among inpatient and outpatient care seekers, respectively. The headcount poverty burden is thus much higher for hospitalization, chronic illnesses and NCDs than their respective counterparts. The Pen’s parade graphs also attest to this (figures not shown). For the individual categories of NCDs, as well as chronic illnesses, we see that there are double-digit impacts of health expenses on headcount poverty among households having cholecystectomy, mental disorder, kidney disease, cancer, appendicectomy and hysterectomy (Table 4, column 4). The

impact of health care payment on headcount poverty is also large for households with paralysis, urinary tract infection (UTI), rheumatic fever, benign tumour and asthma patients. It is evident that cholecystectomy holds the highest rank in headcount poverty burden of health expenses, followed by mental disorder, kidney disease, cancer, appendectomy, hysterectomy, paralysis, UTI, rheumatic fever, benign tumour, asthma, ulcer, hypertension, heart diseases, sexually transmitted diseases (STDs) and diabetes (Table 4, column 4). It is evident from Appendix Table 5 that overall head count impoverishment impact is a bit lower in the programme compared with control areas. This impact is even lower among the microinsurance CHs compared with NCHs. While focusing on chronic illnesses and NCDs, a similar difference is evident between the CHs and NCHs. Although the GK MHI scheme provides only basic primary health care, the timely access may nevertheless help to lower the incidence of large (or even catastrophic) expenses as the illness is often easily contained. In contrast, this potential benefit is not available to NCHs who lack a similar access. It must be noted, however, that in the GK scheme the prepayment (although very modest) only allows discounts on charges at the point of service, thus also entailing considerable postpayment outlays. However, the differential impact (between CH and NCH) is not statistically significant in most cases (Table 5). 3.4.2 Average and Normalized Poverty Gaps The average prepayment and postpayment poverty gap (per day per capita) is 14.2 cents and 15.4 cents, respectively (Table 3, columns 7–8). In other words, per day per capita income before paying for health care is less than the poverty line income by 14.2 cents, while after paying for health care the gap stands at 15.4 cents. Thus, OOP payments raise the average poverty gap per day per capita by 1.2 cents, or by 8.6 %, and the normalized poverty gap by 1 % (Table 3,

43.56

56.44

52.28

58.75

52.97

60.45

33.45

56.75

Acute illness

Chronic illness

Communicable diseases

Non-communicable diseases

Accidental diseases and injuries

Inpatient cases

Outpatient cases

59.49

47.97

64.93

57.58

59.7

56.93

59.1

60.89

59.74

Postpayment headcount (%)

2.74*

14.53**

4.48

4.61**

0.95

4.65*

2.66

17.33***

3.4**

Poverty impact (percentage points)

94.94 (6,026)

5.06 (321)

2.15 (139)

49.02 (3,171)

48.83 (3,159)

20.02 (1,295)

79.98 (5,174)

11.83 (404)

33.30 (6,469)

Prevalence/ distribution of illnessb % (n)

0.143 [0.169] (3,281)

0.061 [0.114] (296)

0.164 [0.182] (134)

0.131 [0.164] (2,341)

0.150 [0.173] (2,097)

0.131 [0.167] (1,140)

0.141 [0.168] (2,966)

0.102 [0.157] (404)

0.142 [0.170] (3,937)

Prepayment gap (per day per capita in US$) Mean [SD] (n)

0.154 [0.174] (3,281)

0.100 [0.144] (296)

0.182 [0.185] (134)

0.145 [0.170] (2,341)

0.156 [0.175] (2,097)

0.148 [0.174] (1,140)

0.151 [0.172] (2,966)

0.161 [0.183] (404)

0.154 [0.175] (3,937)

Postpayment gap (per day per capita in US$) Mean [SD] (n)

Poverty intensity (poverty gap)

0.011***

0.040***

0.018

0.015***

0.006

0.017**

0.010**

0.059***

0.012***

Poverty impact (average poverty gap in US$)

1

5

2

2

1

2

1

7

1

Poverty impact (normalized poverty gap in %)

b

Percent change as a proportion of prepayment headcount We have provided prevalence for ‘all episode of illness’ (there were 6,469 individuals out of 19,422 who had some illnesses; hence, prevalence stands at 33.33 %) and ‘catastrophic events’. For the other cases we have provided the distribution of illnesses by different categories

a

***, ** and * indicates significance at the 1, 5 and 10 % level, respectively

4.83

43.41

7.41

8.7

1.62

8.89

4.71

39.78

6.03

Poverty impact (%)a

n* represents the total number of individuals seeking health care in the sample

56.34

Catastrophic events

Prepayment headcount (%)

Poverty headcount

All episodes of illness

Categories

Table 3 Impact of out-of-pocket payments on incidence and intensity of poverty

Impoverishment Impact of Out-of-Pocket Payments

10.0

Postpayment expenditure

9.0

Prepayment expenditure

8.0

poverty line expenditure

7.0

($1=BDT69)

prepayment exp.; postpayment exp.;

Fig. 1 Poverty Pen’s parade graph for all ailments. BDT Bangladesh Taka, exp. expenditure

Poverty line exp. (per day per capita) in US$

S. A. Hamid et al.

6.0 5.0 4.0 3.0 2.0 1.0 0.0 0

10

20

30

40

50

60

70

80

90

100

Cumulative Proportion of households ranked by prepayment household expenditures (in %)

columns 9–10). In the Pen’s parade graph, the extent of the poverty gap is measured by the area below the poverty line above each parade [4]. The graph illustrates that health care expenses increase the intensity of poverty (Fig. 1). Payments for health care raise the poverty gap (normalized poverty gap) for chronic illnesses by 1.7 cents (2 %), while it is 1 cent (1 %), 1.5 cents (2 %) and 0.6 cent (1 %) for acute illnesses, NCDs and CDs, respectively (Table 3, columns 9–10). Similarly, payments for health care raise the average poverty gap (normalized poverty gap) for inpatient and outpatient care by 4 cents (5 %) and 1.1 cents (1 %), respectively, whereas the corresponding figure for those who incur CHE is 5.9 cents (7 %). Furthermore, OOP payments are seen to raise the average poverty gap (normalized poverty gap) for cholecystectomy, mental disorder, kidney disease, cancer, appendectomy, benign tumour and hysterectomy patients by 2.5 cents (3 %), 3.4 cents (4 %), 4 cents (5 %), 6.8 cents (8 %), 3.4 cents (4 %), 1.5 cents (2 %), and 3.2 cents (4 %), respectively (Table 4, columns 9–10). The corresponding figures for paralysis, UTI, rheumatic fever and asthma patients are 3.4 cents (4 %), 2.6 cents (3 %), 1.9 cents (2 %) and 1.8 cents (2 %), respectively. In terms of normalized poverty gap, therefore, the burden of health expenses is the highest for cancer patients followed by those with kidney diseases, mental disorder, paralysis, appendectomy, hysterectomy, UTI, cholecystectomy, rheumatic fever, heart diseases, STDs, ulcer, asthma, benign tumour, diabetes, and hypertension, in that order.

4 Discussion and Conclusions This study measures the disease-specific impact of OOP payments for health care on poverty (both headcount and poverty gap) by comparing prepayment poverty (where

payments for health care are included) and postpayment poverty (where payments for health care are excluded). We find that spending for health care annually pushes 3.4 % of households into poverty in the central areas of rural Bangladesh covered by the survey data. The corresponding figures for those who suffered from NCDs, chronic illness, hospitalization and catastrophic illness were 4.61, 4.65, 14.53 and 17.33 %, respectively. Looking into individual categories of NCDs, we find that the headcount impoverishment impact of OOP payments is immense for cholecystectomy (22.22 %), mental disorder (18.75 %), kidney disease (15.22 %), cancer (12.5 %), appendectomy (12.5 %), and hysterectomy (9.84 %). The impact on the intensity of poverty is the largest among hospitalized patients. Individually, the intensity is highest among cancer patients (see Tables 3, 4). The impact would have been even higher if informal payments and quasi-formal payments were included in OOP outlays. Nevertheless, the overall impact cited above is somewhat lower than what van Doorslaer et al. [6] had reported earlier for Bangladesh (3.8 % for the international poverty line of $1.08 in purchasing power parity (PPP) terms). Figures of a similar order have also been found for both China (5 % [11]) and India (3.2 % [8]). We also found a moderate impact on poverty deepening, such that the average poverty gap is raised by 1.2 cents per capita daily or US$4.4 per capita annually and the normalized poverty gap by 1 %. Note that payment for drugs accounts for a major part of the impact because, as seen in Table 1, the former makes up for about 60 % of overall OOP payments. The impact (both poverty headcount and poverty deepening) is much higher for NCDs, chronic illnesses and hospitalization compared with CDs, acute illnesses, and outpatient care, respectively. AI has almost a similar burden as NCDs. This impact is exceptionally high for catastrophic expenses and hospitalization, where most of the episodes were NCDs (about 88 % and 85 %, respectively). Although

67.57 45.95

43.75

58.02

57.84

44.00

65.52

64.29 41.67

40.98

43.75

50.00

50.00

50.00

22.22

Hypertension

Ulcer (including peptic ulcer)

Shortness of breath, including asthma

Benign tumour

Rheumatic fever

UTI Paralysis

Hysterectomy

Appendectomy

Cancer

Kidney disease

Mental disorder

Cholecystectomy

44.44

68.75

65.22

62.5

56.25

50.82

71.43 50

72.41

50.00

63.73

61.07

45.83

70.27 48.65

31.25

Postpayment headcount (%)

22.22

18.75

15.22

12.5

12.5

9.84

7.14 8.33

6.89

6.00

5.89

3.05

2.08

2.7 2.7

1.56

Poverty impact (percentage points)b

The magnitudes are not significant at the conventional significant level

Percent change as a proportion of prepayment headcount

100.00

37.5

30.44

25.00

28.57

24.01

11.11 19.99

10.52

13.64

10.18

5.26

4.75

4.00 5.88

5.25

Poverty impact (%)a

0.42 (27)

0.25 (16)

0.74 (48)

0.12 (8)

0.25 (16)

0.94 (61)

0.22 (14) 0.56 (36)

0.46 (30)

0.77 (50)

2.38 (154)

2.03 (131)

2.29 (148)

0.96 (62) 1.75 (113)

1.04 (67)

Distribution of illnessc % (n*)

0.036 [0.078]

0.130 [0.159]

0.121 [0.154]

0.104 [0.162]

0.134 [0.176]

0.083 [0.118]

0.201 [0.199] 0.106 [0.182]

0.128 [0.142]

0.100 [0.142]

0.170 [0.187]

0.150 [0.176]

0.119 [0.169]

0.171 [0.168] 0.103 [0.154]

0.059 [0.106]

Prepayment gap (per day per capita in US$) Mean [SD]

0.061 [0.096]

0.164 [0.173]

0.161 [0.172]

0.173 [0.169]

0.168 [0.201]

0.114 [0.143]

0.227 [0.207] 0.140 [0.197]

0.147 [0.142]

0.116 [0.151]

0.188 [0.189]

0.168 [0.183]

0.125 [0.173]

0.189 [0.178] 0.121 [0.167]

0.070 [0.123]

Postpayment gap (per day per capita in US$) Mean [SD]

Poverty intensity (poverty gap)

0.025

0.034

0.040

0.068

0.034

0.032

0.026 0.034

0.019

0.015

0.018

0.017

0.007

0.018 0.018

0.011

Poverty impact (average poverty gap in US$)b

3

4

5

8

4

4

3 4

2

2

2

2

1

2 2

1

Poverty impact (normalized poverty gap in %)

In this column we have presented the distribution of total number of ill individuals (i.e. 6,469) among the ‘selected illness’ categories. For example, there were 67 individuals (out of 6,469) who had diabetes; hence, diabetes accounts for 1.04 % of the illnesses. As we did not present all categories of illness, the sum of this distribution does not stand at 100 %

c

b

a

STDs sexually transmitted diseases, UTI urinary tract infection

n* represents the total number of individuals seeking health care in the sample

29.69

STDs Heart disease

Prepayment headcount (%)

Poverty headcount

Diabetes

Categories

Table 4 Impact of out-of-pocket payments on incidence and intensity of poverty for major categories of illness

Impoverishment Impact of Out-of-Pocket Payments

S. A. Hamid et al.

there is an equal proportionate incidence of CDs and NCDs, the absolute impoverishment burden is about five times higher for NCDs than CDs. A closer examination reveals that the impact (especially headcount poverty) is excessively high for NCDs, such as cholecystectomy, mental disorder, kidney disease, cancer, appendectomy, hysterectomy, paralysis, UTI, rheumatic fever, benign tumour and asthma. The majority of these illnesses are chronic NCDs causing a major impoverishment burden. It is also noticeable that some non-chronic NCDs (e.g. cholecystectomy, appendectomy, hysterectomy), which usually require immediate surgical procedures, cause substantial impoverishment impact (see Table 4). Hence, there is a clear indication that NCDs, particularly of a chronic nature, and those requiring immediate surgical procedures, should be given more priority for policy framing. As we use the 12-month recall period, there is some potential recall bias. As far as OOP is concerned, the recall problem usually causes a downward bias. This is because respondents often cannot recall small expenses and, more generally, those related to minor illnesses. Thus, the overall impoverishment impact found in the study may prove to be an under-estimate. Moreover, as already cited, the data analysed in the paper does not represent Bangladesh as a whole or even all of rural Bangladesh. Future research may deal with the issue using a nationally representative sample and more suitable recall periods. Despite these limitations, there are indications that payment for NCDs is a visible threat to the poverty reduction initiatives in the country. Thus, Bangladesh stands to gain hugely if viable alternatives can be found to finance the provision of health care away from the OOP mode. Developing appropriate risk-pooling modalities such as low-cost voluntary MHI schemes are gaining popularity in many contexts similar to that in Bangladesh. We also find some impact of the limited primary outpatient care scheme run by GK. However, this is not a viable route for dealing with chronic NCDs and catastrophic illnesses. Moreover, there is little evidence of the replicability and scalability due, presumably, to both demand and supply side constraints. Introduction of social insurance is not quite feasible due to the large informal economy. Thus, hope lies in the general taxation-based public provision of health care. However, the scope of rendering the government facilities more efficient is plagued by both budgetary limitations as well as by daunting governance issues in the face of reported endemic corruption. Evidence shows that there is very low use of public health care [16]. As depicted in the literature (e.g. Health Economics Unit [17] and Andaleeb et al. [18]), loss of faith in public facilities (due to various supply side constraints, e.g. appropriate skill-mix, input-mix, absenteeism of doctors, politicisation in the posting of doctors, shortage of drugs) is one of the main reasons for low demand for public

health care. Thus, access to public health care can only be increased via enhancing its quality as well as its efficiency. This would require significant reforms, e.g. strengthening of local government bodies and integrating these in the management of subdistrict and rural hospitals and health centres, and allowing all levels of hospitals to impose some user fees (combined with a proper safety net for the poor and the vulnerable) and retention of these fees on their part for smoothening timely service delivery. The high share of drug costs in OOP reflects, in part, the unrestricted dispensation of all drugs, including antibiotics, over-the-counter [19, 20], as well as lack of control of drug prices [20]. Although the Bangladesh National Drug Policy professes to ensure the rational pricing of essential drugs, the regulatory authorities have little control over actual drug prices (witness the large hike, 30 % or more, in retail prices in early 2012). Moreover, there is overuse of drugs in Bangladesh [19, 20]. Evidence shows that at least half of all drugs are not prescribed, dispensed or sold under guidance [19, 20]. Selfmedication and purchase of all types of drugs without any prescription, and the continuous proliferation of unlicensed (and possibly illegal) and unregulated drug stores, are among the major reasons for the overuse of drugs [19, 20]. Thus, gradual adoption of measures of effective regulation of the price of essential drugs, and restrictions in selling over-the-counter drugs, may be long overdue. Acknowledgments The authors are grateful to the Department for International Development’s (UKAid) PROSPER (‘Promoting Financial Services for Poverty Reduction Programme’) project for providing funds for the longitudinal study ‘‘Microinsurance, Poverty & Vulnerability’’ housed at the Institute of Microfinance (InM). However, any opinions expressed and policy suggestions proposed in the document are the authors’ own and do not necessarily reflect the views of either InM or that of the funding agency. There are no conflicts of interest between the authors. The authors acknowledge Shubhasish Barua, Chowdhury Abdullah Al Asif, Rifat Haider, Suvadra Gupta, Raysul Naim and Shahidul Islam for valuable research and logistical assistance. The authors also acknowledge the anonymous reviewers and the Editor-in-Chief for their valuable comments and suggestions. The authors are also grateful to GK for its kind cooperation with this research. Author contributions Syed Abdul Hamid provided significant input (including conceptualization, sketching out overall structure, data analysis and write up) to prepare the primary draft and all subsequent drafts of the manuscript. He is the guarantor for the overall content. Syed M. Ahsan provided substantial input for improving the analysis in all aspects, beginning with conceptualization. Afroza Begum provided substantial input for data analysis, literature review and in drafting the manuscript.

Appendix See Table 5 and Fig. 2.

NCH

57.59

61.64

53.27

55.99

Control area

NCH

51.40

47.89

CH

57.58

52.78

Total

Programme area

56.98

61.31

55.71

Control area

52.99

45.38

NCH

49.62

51.81

CH

56.89

59.06

63.51

55.64

59.74

Total

Programme area

Control area

60.15 54.60

Postpayment headcount (%)

56.75 51.82

Prepayment headcount (%)

Poverty headcount

4.32

5.65*

3.51

4.80

3.99

5.59

4.23

5.08

3.42

3.77*

3.4* 2.78

Poverty impact (percentage points)

a

Percent change as a proportion of prepayment headcount

** and * indicates significance at the 1, 5 and 10 % level, respectively

CH card holder, NCH non-card holder

n* represents the total number of individuals seeking health care in the sample

Non-communicable diseases

Chronic illness

Programme area

All episodes of illness

Total CH

Area

Categories

8.11

10.08

7.33

9.09

7.53

10.04

9.32

9.80

6.15

6.31

5.99 5.36

Poverty impact (%)a

17.02 (1,233)

15.71 (1,166)

16.29 (772)

15.93 (1,938)

7.12 (516)

6.67 (495)

5.99 (284)

6.40 (779)

35.41 (2,566)

31.40 (2,331)

32.09 (3,903) 33.16 (1,572)

Prevalence of illness % (n*)

0.132 [0.167] (903)

0.142 [0.168] (868)

0.111 [0.150] (570)

0.130 [0.162] (1438)

0.133 [0.169] (451)

0.145 [0.172] (429)

0.105 [0.150] (260)

0.130 [0.165] (689)

0.139 [0.170] (1,463)

0.158 [0.178] (1,540)

0.143 [0.170] (2,474) 0.119 [0.154] (934)

Prepayment gap (per day per capita in US$) Mean [SD] (n)

0.148 [0.173] (903)

0.160 [0.174] (868)

0.120 [0.155] (570)

0.144 [0.168] (1438)

0.150 [0.175] (451)

0.166 [0.180] (429)

0.115 [0.157] (260)

0.146 [0.173] (689)

0.153 [0.176] (1,463)

0.172 [0.182] (1,540)

0.155 [0.174] (2,474) 0.127 [0.158] (934)

Postpayment gap (per day per capita in US$) Mean [SD] (n)

Poverty intensity (poverty gap)

Table 5 Impact of out-of-pocket payments on the incidence and intensity of poverty by study areas and major illnesses

0.015**

0.018**

0.008

0.014**

0.016

0.021

0.010

0.017

0.014**

0.013*

0.011** 0.008

Poverty impact (average poverty gap in US$)

2

2

1

2

2

2

1

2

2

1

1 1

Poverty impact (normalized poverty gap in %)

Impoverishment Impact of Out-of-Pocket Payments

S. A. Hamid et al.

Fig. 2 A schematic view of self-reported illnesses. CDs communicable diseases, NCD non-communicable diseases

References 1. Kochar A. Explaining household vulnerability to idiosyncratic income shocks. Am Econ Rev. 1995;85(2):159–64. 2. Gertler P, Gruber J. Insuring consumption against illness. Am Econ Rev. 2002;92(1):51–70. 3. Whitehead M, Dahlgren G, Evans T. Equity and health sector reforms: can low-income countries escape the medical poverty trap? Lancet. 2001;358(9284):833–6. 4. Wagstaff A, van Doorslaer E. Catastrophe and impoverishment in paying for health care: with applications to Vietnam 1993–1998. Health Econ. 2003;12:921–34. 5. Garg CC, Karan AK. Health and millennium development goal 1: reducing out-of-pocket expenditures to reduce income poverty—

evidence from India. EQUITAP project, Working Paper No. 15; 2005. 6. van Doorslaer E, O’Donnel O, Rannan-Eliya RP, Somanathan A, Adhikari SR, Garg CC, et al. Effect of payments for health care on poverty estimates in 11 countries in Asia: an analysis of household survey data. Lancet. 2006;368(9544):1357–64. 7. Flores G, Krishnakumar J, O’donnell O, van Doorslaer E. Coping with health-care costs: implications for the measurement of catastrophic expenditures and poverty. J Health Econ. 2008;17: 1393–412. 8. Limwattananon S, Tangcharoensathien V, Parkongasi P. Catastrophic and poverty impacts of health payments: results from national household surveys in Thailand. Bull World Health Organ. 2007;85(8):600–6.

Impoverishment Impact of Out-of-Pocket Payments 9. Garg CC, Karan AK. Reducing out-of-pocket expenditures to reduce poverty: a disaggregated analysis at rural–urban and state level in India. Health Policy Plan. 2009;24(2):116–28. 10. Somkotra T, Lagrada LS. Payments for health care and its effect on catastrophe and impoverishment: experience from the transition to Universal Coverage in Thailand. Soc Sci Med. 2008; 67(12):2027–35. 11. Sun X, Sleigh AC, Carmichael G, Jackson S. Health paymentinduced poverty under China’s New Cooperative Medical Scheme in rural Shandong. Health Policy Plan. 2010;25(5): 419–26. 12. Yardim MS, Cilingiroglu N, Yardim N. Catastrophic health expenditure and impoverishment in Turkey. Health Policy. 2010; 94(1):26–33. 13. Shi W, Chongsuvivatwong V, Geater A, Zhang J, Zhang H, Bromal D. Effect of household and village characteristics on financial catastrophe and impoverishment due to health care spending in Western and Central Rural China: a multilevel analysis. Health Res Policy Syst. 2011;9:16. 14. Ravallion M, Sen B. When methods matters: monitoring poverty in Bangladesh. Econ Dev Cult Change. 1996;44(4):761–92.

15. Bangladesh Bureau of Statistics (BBS). Report of the Household Income and Expenditure Survey 2005, Ministry of Planning, Government of Bangladesh. Dhaka: Bangladesh Bureau of Statistics; 2007. 16. Hamid SA, Ahsan SM, Begum A, Asif CAA. Inequity in formal health care use: evidence from rural Bangladesh. Institute of Microfinance, Working Paper No. 28; 2014. 17. Health Economics Unit (HEU). Expanding social protection for health towards universal health coverage: health care financing strategy 2012–2032. Ministry of Health and Family Welfare, Government of Bangladesh. Dhaka: Health Economics Unit; 2012. 18. Andaleeb SS, Shiddique N, Khandkar S. Patient satisfaction with health services in Bangladesh. Health Policy Plan. 2007;22:263–73. 19. Babu M. Factors contributing to the purchase of over the counter (OTC) drugs in Bangladesh: an empirical study. Int J Third World Med. 2007;6(2). 20. Akter SFU, Rashid MA, Mazumder SK, Jabbar SA, Sultana F, Rahman MH, Zahedee MNS. Essential drugs in Bangladesh and role of different stakeholders: a qualitative study. Int J Sci Environ Technol. 2012;1(5):506–18.