









































Pa
ge

 
1



Pa
ge

 
28

0

American Journal of  Economics and 
Business Innovation (AJEBI)

Hardship Financing in Healthcare among Rural Residents in the Gaibandha District: An 
Application of  Binary Logistic Regression

Most. Sabiha Sultana1, Md. Shawkat Ali2*

Volume 4 Issue 2, Year 2025
ISSN: 2831-5588 (Online), 2832-4862 (Print)

DOI: https://doi.org/10.54536/ajebi.v4i2.5387
https://journals.e-palli.com/home/index.php/ajebi

Article Information ABSTRACT

Received: May 15, 2025

Accepted: June 23, 2025

Published: August 07, 2025

High out-of-pocket expenditures lead to financial hardship in the absence of  health 
insurance coverage, which is prevalent in low- and middle-income countries, including 
Bangladesh. Households often incur “hardship financing” when they face additional losses 
by borrowing from a bank, microfinance, or moneylender (with interest) and selling assets, 
which was 31.4% of  516 sampled respondents in the rural areas of  the Gaibandha district 
in Bangladesh, selected as the study area. In contrast, financing in healthcare by current 
income or savings, and borrowing from relatives or friends (without interest), is termed 
“no hardship financing”, which was 68.6%. To investigate the factors affecting the risk of  
“hardship financing” as the outcome variable with “Yes or No” category, a binary logistic 
regression was applied, including six independent variables that made statistically significant 
contributions to the model, such as age, education, family size, the distance to the hospital, 
outpatient expenditures, and chronic illness. Subsidized healthcare services or health 
insurance schemes, and accessing to a wealthier social network were suggested, which may 
protect against the risk of  hardship financing. Besides, establishing more healthcare facilities 
in remote areas with assured quality has been emphasized. Hence, to achieve universal health 
coverage, policymakers should consider the significant factors associated with hardship 
financing.

Keywords
Bangladesh, Binary Logistic 
Regression, Gaibandha District, 
Hardship Financing, Out-of-Pocket 

1 School of  Social Sciences, Humanities and Languages, Bangladesh Open University, Bangladesh
2 Department of  Economics, Jahangirnagar University, Savar, Dhaka, Bangladesh
* Corresponding author’s e-mail: ssrony_bou@yahoo.com

INTRODUCTION
All member countries of  the United Nations (UN) are 
committed to achieving Universal Health Coverage 
(UHC) as part of  the Sustainable Development Goals 
(SDGs)-3, aiming to ensure financial risk protection for 
all by 2030 (Tadiwos, 2025; UNDP, 2019). Hence, UHC 
aims to provide access to quality healthcare without 
facing financial hardship (Kolesar et al., 2023). Financial 
hardship is particularly prevalent in low- and middle-
income countries due to limited health insurance coverage 
and high Out-of-Pocket (OOP) payments. In turn, high 
OOP payments increase economic vulnerability and lead 
to long-term poverty among households.
Limited access to health insurance and unexpected OOP 
payments can lead to asset depletion, debt, and reductions 
in essential consumption, which can prevent access 
to healthcare. This may eventually result in financial 
catastrophes, distress financing, and impoverishment 
(Islam et al., 2017; Leive & Xu, 2008; Russell, 2004; Chuma 
& Ezeoke, 2012). Health expenditures are considered 
catastrophic when they exceed a certain percentage of  
household income or expenditures (Tadiwos, 2025). 
Besides, health expenses that impoverish individuals 
or households are defined as those that exceed the 
internationally or nationally agreed-upon poverty line 
(Smit, 2009). There are an estimated 930 million (12.7% 
of  the global population) people worldwide who face 
catastrophic health expenditures (CHE) because they 
sacrifice at least 10% of  their household budgets to 
finance their healthcare (WHO, 2015).

Financing in healthcare includes sources such as current 
income or savings; borrowing from relatives or friends 
(without interest); borrowing from a bank, microfinance, 
or moneylender (with interest); and selling assets. 
Households often incur “hardship financing” when 
they face additional losses by borrowing from a bank, 
microfinance, or moneylender (with interest) and selling 
assets. On the other hand, financing in healthcare by 
current income or savings, and borrowing from relatives 
or friends (without interest), is termed as “no hardship 
financing” (Binnendijk et al., 2012). The second category 
has been regarded as less burdensome than the first 
in various studies (Asfaw et al., 2010; Kruk et al., 2009; 
Steinhardt et al., 2009). However, the consequences may 
differ between those of  richer households and poorer ones. 
Bangladesh has an extensive infrastructure in the 
healthcare delivery system. However, every year, an 
estimated 150 million people suffer severe financial 
hardship, and 100 million falls below the poverty line 
because of  high healthcare expenditures (Tahsina, 2018). 
In addition, public and private health services coexist in 
Bangladesh, with a dual healthcare system. In the public 
sector, outpatient, inpatient, and preventive care are 
provided largely. Whereas, curative care is mainly provided 
in outpatient and inpatient settings in the private sector 
(Islam et al., 2017). However, people often lack access to 
quality healthcare at an affordable cost, as they expect. 
In 2020, the OOP expenditures by households constituted 
a total of  68.5% (two-thirds) of  the Total Health 
Expenditures (THE) in Bangladesh, followed by 23.1% 



Pa
ge

 
28

1

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

from government healthcare financing. The remaining 
portion, including development partners, contributed 
5%, while NGOs contributed 2%. Additionally, private 
corporations, autonomous bodies, and voluntary 
health insurance, in combination, contributed 2% to 
healthcare financing (BNHA, 1997-2020). It is noted 
that approximately five million people in Bangladesh 
are impoverished by large OOP payments for healthcare 
every year (Van Doorslaer et al., 2006; Ahmed et al., 2013).
Generally, in most low-income rural settings, there is 
an irregular flow of  income, which instigates financial 
hardship. Consequently, different types of  suffering 
may occur, such as cutting a meal from a household’s 
regular food menu or a child’s schooling. However, in-
depth studies regarding factors associated with ‘hardship 
financing’ in healthcare among people in the rural areas 
of  northern Bangladesh are quite limited. The study will 
focus on how rural people of  the Gaibandha district 
finance their healthcare expenditures. Besides, the study 
will identify the factors associated with hardship financing 
in healthcare among respondents.

MATERIALS AND METHODS 
Research Approach 
The study was explorative, using both quantitative and 
qualitative data. 

Study Focus 
A cross-sectional study was carried out in the rural areas 
of  the Gaibandha district, which was selected as the study 
area. The survey was conducted from February 2019 
to March 2019 and focused on the rural people of  the 
Gaibandha district. 

Sampling 
A simple random sampling technique was used in 
two ways for the study. Firstly, four Upazilas were 
selected from seven Upazilas of  the Gaibandha district, 
named Fulchhari, Gaibandha Sadar, Gobindaganj, 
and Sundarganj (BBS 2011). Secondly, a total of  516 
respondents were selected finally by using the same 
technique for interviewing. 

Sources of  Data 
The primary data was collected from the respondents 
in the study area. Relevant books, journal articles, web 
pages, the Bangladesh budget, and the “Population and 
Housing Census, 2011, Zila report: Gaibandha” served as 
the sources of  secondary data. 

Data
To assess the socio-economic status, the survey 
questionnaire included socio-economic factors such as 
gender, age, family size, education, occupation, and annual 
income. Besides, the respondents were asked about their 
inpatient and outpatient healthcare expenditures in the 
year preceding the survey. The respondents were asked 
whether they had suffered from any type of  chronic 
illnesses. To identify chronic illness, respondents were 
asked a set of  questions related to disease symptoms, 
length of  illness, and regular medicine use. Hence, staying 
in a hospital for more than 24 hours has been considered 
inpatient care. On the other hand, outpatient care includes 
expenditures on drugs, diagnostic tests, and consultations 
with healthcare practitioners staying in a hospital for less 
than 24 hours. The inpatient care included medical direct 
expenses and hospital admissions of  the respondents. 
Besides, the respondents were asked about the distances 
to the hospital from their residences (travel distance in 
kilometers). To investigate the sources of  healthcare 
financing, the respondents were asked whether they had 
used current income, savings, borrowing from relatives, 
friends, neighbors, banks, moneylenders, or microfinance, 
and money received by selling assets.

Data Analysis and Presentation
Data has been analyzed using SPSS version 26. To 
investigate the factors affecting the risk of  hardship 
financing among rural residents in the Gaibandha 
district, “Binary Logistic Regression” was applied. Thus, 
“Hardship Financing” is the outcome variable, being 
a binary category with “Yes or No”. A 5% level of  
significance was used for the study to show the statistical 
significance. MS Excel was used for data presentation and 
tabulation.

Limitations of  the Study
Variations in knowledge and understanding regarding 
health conditions, memory recall issues, and overlapping 
symptoms among patients may lack of  accuracy in 
self-reporting illness episodes. The cross-sectional 
nature of  the study rendered it incapable of  capturing 
seasonal variations in household income or illness-related 
expenditure and coping strategies. In addition, it is not 
possible to draw the same conclusions elsewhere without 
reliable and sufficient data. 

RESULTS AND DISCUSSION
Socio-Economic Profile of  the Sampled Respondents

Table 1: The socio-economic status of  the sampled respondents
Variables Frequency (n) Percentage
Gender
Male 340 65.9
Female 176 34.1
Age 
Lowest up to 30 197 38.2



Pa
ge

 
28

2

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

31-45 198 38.4
45 above 121 23.2
Marital Status
Unmarried 3 0.6
Married 502 97.3
Widow/Divorce 11 2.1
Family size
Lowest up to 3 378 73.3
4 to 7 133 25.8
>7 5 1.0
Number of  children
0 to 3 427 82.8
4 above 89 17.2
Education
Primary 221 42.8
Secondary and Higher Secondary 80 15.5
Graduation / Masters 24 4.7
No education 191 37.0
Occupation
Service holder 117 22.7
Business 110 21.3
Farmer 94 18.2
Daily labor 195 37.8
Annual income
Lowest up to 100000 224 64
100001-200000 139 26.9
200000 above 47 9.1
Chronic illness
Yes 276 53.5
No 240 46.5
Distance to hospital (Km)
Lowest up to 3 129 25
4-20 278 53.9
20 above 109 21.1

Table 2: Healthcare expenditures from suffering any major disease among the study respondents 
Variables Frequency (n) Percentage
Households with inpatient care costs last year (BDT)
Lowest up to 1000 322 62.4
1001-10000 109 21.1
10001-25000 51 9.9
25000 above 34 6.6  
Households with outpatient care costs last year (BDT)
Lowest up to 1000 185 35.9
1001-5000 312 60.5
5000 above 19 3.7



Pa
ge

 
28

3

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

About 65.9 % of  the respondents were male and 34.1% 
of  the respondents were female. Most of  the households 
were male-headed who were responsible for key decisions 
at the household level. As a larger portion of  the male 
household heads are usually engaged in their occupation 
or performing daily duties during the daytime, all were 
not present at home while interviewing. The mean age 
of  the respondents was 38 years, and the age distribution 
ranged from 18 to 80 years. A large percentage of  
respondents, 38.2%, were age level from lowest up to 30 
years, which has been considered as younger individuals. 
A total of  38.4% were age level of  31-45 years, whereas 
23.2% were 45 years and above old and needed the 
care the most. Hence, there is a greater possibility of  
suffering from illness for older individuals than younger 
individuals because of  decreasing immunity as a general 
consideration.
Hence, daily labor was the main occupation in the study 
areas among rural residents, with a large percentage of  
37.8%. A total of  22.7% were service holders, although 
most of  them were employed in informal sectors, such 
as salesmen and waiters, and a very small portion was 
employed in formal sectors. About 21.3% were self-
employed in businesses, and 18.3% were self-employed 
in agriculture. On average, the annual income was 
BDT 112,852, with a minimum of  BDT 18,000 and a 
maximum of  BDT 650,000. Furthermore, 64% of  them 
reported their household income earning less than BDT 
100000, 26.9% reported BDT100001 to 200000, and 
9.1% reported above BDT 200000 annually.
 A larger portion of  respondents (37%) had no formal 
education, 42.8% had primary education, 15.5% had 
secondary or higher secondary education, and only 4.7% 
had a graduation or higher degree.  The average family 
size was five (5). The family size ranging from one to three 
(1-3) was 378 (73.3%), which accounted for the highest 
number of  respondents. Those with household sizes of  
four to seven (4-7) were 133 (25.8%), while those that 
had family sizes greater than seven (>7) were 5 (1.0%). 
It is noted that about 20 households had no children. A 
total of  82.8% of  children were below 3, 17.2% of  above 
4. About 53.5% had been suffering from chronic illness, 
whereas 46.5% had not.

Healthcare Expenditures
Table 2 shows that the majority of  respondents (62.4%) 
noticed their inpatient expenditures (Table 2) were up to 
BDT1000. On the other hand, 21.1% noticed the amount 

of  BDT1001-10000, 9.9% noticed of  BDT10001-25000, 
and 6.6% noticed of  BDT25000 above. For outpatient 
expenditures (Table 2), the majority of  respondents 
(60.5%) noticed their outpatient expenditures were 
BDT1001- 5000, followed by 35.9% noticing at the lowest 
up to BDT1000, and only 3.7% noticing BDT5000 above. 

Healthcare Financing
To meet the health expenditures (Table 3), the major 
portion of  the sampled respondents resorted to using 
their current income, savings, and borrowing from 
relatives/neighbors/friends, and hence they were 
considered to have no hardship financing. Besides, 
they often sold their assets or borrowed money with 
interest to finance their healthcare expenditures and were 
considered hardship financing (Binnendijk et al., 2012). 
Based on the criterion, hardship financing was found 
among 31.4% of  the sampled respondents during the 
year preceding the survey, and the remaining 68.6% had 
no hardship financing. 

Binary Logistic Regression (BLR) Analysis 
In regression analysis, an association between a response 
variable and one or more explanatory variables is 
determined. However, there is often a situation where the 
outcome variable is discrete, with two or more possible 
values (Hosmer & Lemeshow, 2000). Hence, a BLR is 
used to model the relationship between a dichotomous 
dependent variable, which is binary rather than continuous 
and multiple continuous or categorical independent 
variables (Hair et al., 2010; Hyeoun-Ae Park, 2013).
In the study, a BLR analysis was applied between hardship 
financing as the dependent variable with “Yes” or “No” 
category and the independent variables with gender, 
age, marital status, education, occupation, family size, 
number of  children, income, distance to the hospital, 
chronic illness, inpatient expenditures, and outpatient 
expenditures. However, a multivariate analysis was used 
under the BLR analysis to identify the factors influencing 
hardship financing while paying healthcare expenditures. 
The significance value (p-value) of  the Wald test has 
been used in the BLR analysis to determine whether the 
predictor variables meaningfully contribute to the model 
(Vakhitova et al., 2018). 

Assumptions of  BLR Analysis
BLR analysis requires certain assumptions to be satisfied 
to give a valid result. They are as follows:

Table 3: Healthcare financing of  the study respondents
Variables Frequency (n) Percentage
Covering costs by using present income 175(33.8) 341(66.2)
Covering costs by using savings 61(11.8) 455(88.2)
Covering costs by borrowing from relatives/neighbors/friends 357(69.1) 159(30.9)
Covering costs by selling assets 146(28.2) 370(71.6)
Covering costs by borrowing from banks/moneylenders/microfinance 80(15.5) 436(84.3)



Pa
ge

 
28

4

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

The Categories for a Dependent Variable Must be 
Mutually Exclusive And Exhaustive
It means that every data point should fall into only one 
category (mutually exclusive), and the categories provided 
should cover all possible outcomes (exhaustive).
The assumption was checked by defining the categories 
were defined that no single data point logically fell into 
more than one category. Besides, the assumption was 
checked with the numerical data, where the ranges did 
not overlap.

Binary Dependent Variable
There must be a dichotomous dependent variable, 
which was checked as dichotomized into “Yes” or “No” 
categories (Hosmer et al., 2013).  

Independence of  Observations
The observations should be independent of  each other. 
A simple random sampling has been applied to collect 
the data, ensuring the independence of  each observation. 
Besides, clustering was avoided, which might introduce 
dependencies. 

No Multicollinearity among Predictors
A key assumption in logistic regression is that the 
explanatory variables should not be highly correlated 
with each other. By using tolerance and variance inflation 
factor (VIF), multicollinearity can be detected easily. 
The VIF is defined as the reciprocal of  tolerance as 
follows:

VIF= 1/ Tolerance
The tolerance value (Table 4) close to 1 indicated 
that there was little multicollinearity, which can be 
accepted. Whereas a value close to zero suggests that 
multicollinearity may be a threat (Senaviratna and Cooray, 
2019). However, the calculated value of  VIF was <4, which 
shows no multicollinearity in the dataset (Jahan, 2022). 

Large Sample Size
Predictor variables should consist of  at least 10 events per 
variable as a general rule of  thumb (Peduzzi et al., 1996). 
Hence, there were 12 (twelve) predictor variables, and the 
events should be 120 (one hundred and twenty). However, 
516 (five hundred and sixteen) respondents were selected 
for the study, which shows a large sample size. 

Variables of  the Study
The dependent variable was hardship financing, 
categorized as “Yes” or “No”.
The independent variables used in the study were gender, age, 
marital status, education, occupation, family size, number of  
children, income, the distance to the hospital, chronic illness, 
inpatient expenditures, and outpatient expenditures

Table 4: Nature of  the dependent variable with the 
category codes
Categories of  hardship financing Codes
Yes 1
No 2

Table 5: Nature of  the independent variables with the category codes
Designation of  Variable Description of  Independent 

Variables
Codes

X1 (X1(1) =Male, X1(2) =Female) Gender of  respondent 1=Male, 2=Female
X2(X2(1) =Lowest up to 30 years,
X2(2) =31-45 years, X2(3) =45 years above)

Age of  respondent 1=Lowest up to 30 years, 2=31-45 
years, 3=45 years above

X3(X3(1) =Unmarried, X3(2) =Married, X3(3) 
=Married) 

Marital Status of  the 
Respondent

1=Unmarried, 2=Married, 3= 
Widow/Divorce

X4 (X4(1) =Lowest up to 3, X4(2) =4 -7, X4(3) 
=7 above) 

Family size of  the respondent 1= Lowest up to 3, 2=4- 7, 3= 7 above

X5(X5(1) =0 to 3, X5(2) =4 above) Number of  children 1= 0 to 3, 2=4 above
X6(X6(1) =Primary, X6(2) =Secondary / Higher 
Secondary, X6(3) =Graduation / Masters, X6(4) 
= No education)

Education level 1= Primary, 2=Secondary / 
Higher Secondary, 3= Graduation 
/ Masters,4= No education

X7(X7(1) =Service holder, X7(2) =Business, 
X7(3) =Farmer, X7(4) =Daily Labor)

Occupation 1= Service holder, 2=Business, 3= 
Farmer, 4= Daily Labor

X8(X8(1) =Lowest up to 100000, X8(2) 
=100001-200000, X8(3) =200000 above) 

Annual income (BDT) 1= Lowest up to 100000, 2=100001-
200000, 3=200000 above 

X9(X9(1) =Lowest up to 3, X9(2) =4-20, X9(3) 
=20 above)

The distance of  the hospital 
from the Patient’s residence (km) 

1= Lowest up to 3, 2=4-20, 3=20 
above 

X10(X10(1) =Yes, X10(2) =No) Chronic illness 1= Yes, 2=No
X11(X11(1) =Lowest up to 1000, X11(2) =1001-
10000, X11(3) =10001-25000, X11(4) = 25000 
above)

Inpatient expenditures (BDT) 1= Lowest up to 1000, 2= 1001-
10000, 3= 10001-25000, 4= 25000 
above

X12(X12(1) =Lowest up to 1000, X12(2) =1001-
5000, X12(3) =5000 above) 

Outpatient expenditures 
(BDT) 

1= Lowest up to 1000,
2= 1001-5000, 3= 5000 above



Pa
ge

 
28

5

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

Checking Model Adequacy
In Table 6 to 10, the full model, containing all predictors, 
was statistically significant, as indicated by a chi-square 
value of  123.061 with df  = 24, p-value = .000, meaning 
p < .001, which suggests that the model could distinguish 
between respondents who reported incurring hardship 
financing and those who did not. The model explained 

Table 6: Omnibus Tests of  Model Coefficients
Chi-square df Sig.

Step 123.061 24 .000
Block 123.061 24 .000
Model 123.061 24 .000

Table 7: Hosmer and Lemeshow Test
Step Chi-square df Sig.

1 6.845 8 .553

Table 8: Model Summary
Step -2 Log likelihood Cox & Snell R Square Nagelkerke R Square

1 519.078a .212 .298
a. Estimation terminated at iteration number 5 because parameter estimates changed by less than .001.

Table 9: Classification Table

Observed
Predicted

Hardship Financing Percentage Correct
Yes No

Hardship Financing Yes 77 85 47.5
No 40 314 88.7

Overall Percentage 75.8

between 21.2% (Cox and Snell R-square) and 29.8% 
(Nagelkerke R-square) of  the variance incurring hardship 
financing and correctly classified 75.8% of  cases as a 
whole. In the Hosmer-Lemeshow test, a high p-value 
indicates the model fits the data well (P> 0.05). In this 
case, it suggests that the model’s predictions are consistent 
with observed outcomes, implying a good fit. 

Table 10: Collinearity Statistics

G
en

de
r

A
ge

M
ar

ita
l

St
at

us

E
du

ca
tio

n

O
cc

up
at

io
n 

Fa
m

ily
 s

iz
e

N
um

be
r o

f
ch

ild
re

n

In
co

m
e

D
is

ta
nc

e 
to

 
th

e 
ho

sp
ita

l

C
hr

on
ic

 
ill

ne
ss

In
pa

tie
nt

ex
pe

nd
itu

re

O
ut

pa
tie

nt
ex

pe
nd

itu
re

VIF 1.116 1.308 1.081 1.215 1.388 1.302 1.293 1.282 1.113 1.115 1.469 1.491
Tolerance .896 .765 .925 .823 .721 .768 .773 .780 .898 .897 .681 .671

Parameter Estimates of  Binary Logistic Regression
Binary logistic regression was used to assess the factors 
associated with incurring hardship financing for healthcare 
services. As shown in Table 11 (parameter estimates), 
six independent variables made statistically significant 

contributions to the model, such as age, education, family 
size, the distance to the hospital, outpatient expenditures, and 
chronic illness. On the other hand, the influence of  gender, 
marital status, occupations, the number of  children, income, 
and inpatient expenditures was found to be insignificant.

Table 11: Parameter Estimates of  Binary Logistic Regression
Variables (Category) β (S.E) Wald df Sig. Expβ 95% CI

Lower Bound Upper Bound
Intercept=1 -2.293(1.089) 4.431 1 .035*** - - -
X1

Gender (Ref: Female)



Pa
ge

 
28

6

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

Male .471(.258) 3.340 1 .068 1.602 .966 2.654
X2

Age (Ref:45 above years)
Lowest up to 30 years .853 (.326) 6.862 1 .009*** 2.348 1.240 4.446
31-45 years .711(.295) 5.816 1 .016*** 2.037 1.142 3.631
X3

Marital Status (Ref: Widow/Divorce)
Unmarried -1.213 (1.566) .601 1 .438 .297 .014 6.394
Married -.499 (.806) .383 1 .536 .607 .125 2.949
X4

Family size (Ref: 7 above)
Lowest up to 3 -1.949 (.669) 8.497 1 .004*** .142 .038 .528
4 to 7 -1.452 (.600) 5.856 1 .016*** .234 .072 .759
X5

Number of  children (Ref: 4 above)
0 to 3 -.312 (.344) .820 1 .365 .732 .373 1.438
X6

Education (Ref: No education)
Primary .198 (.281) .494 1 .482 1.218 .702 2.113
Secondary / Higher 
Secondary

-.106 (.389) .074 1 .785 .899 .419 1.928

Graduation / Masters 1.848 (.713) 6.720 1 .010*** 6.346 1.570 25.656
X7

Occupation (Ref: Daily Labor)
Service holder .028 (.345) .007 1 .936 1.028 .523 2.022
Business -.336 (.317) 1.120 1 .290 .715 .384 1.331
Farmer .555 (.397) 1.952 1 .162 1.742 .800 3.795
X8

Annual income (Ref: 200000 above BDT)
Lowest up to 100000 -.389 (.491) .628 1 .428 .678 .259 1.773
100001-200000 -.760 (.484) 2.470 1 .116 .468 .181 1.207
X9

Distance to the hospital (Ref: 20 km above)
 Lowest up to 3 -1.251 (.435) 8.268 1 .004*** .286 .122 .672
4-20 -1.149 (.408) 7.943 1 .005*** .317 .143 .705
X10

Chronic illness (Ref: No)
Yes -.828(.244) 11.538 1 .001*** .437 .271 .705
X11

Inpatient expenditures (Ref: 25000 above BDT)
Lowest up to 1000 .620 (.491) 1.596 1 .206 1.859 .710 4.865
1001-10000 .709 (.509) 1.943 1 .163 2.032 .750 5.506
10001-25000 -.240 (.525) .208 1 .648 .787 .281 2.201
X12

Outpatient expenditures (Ref: 5000 above BDT)
Lowest up to 1000 1.851 (.641) 8.343 1 .004*** 6.366 1.813 22.351
1001-5000 .846 (.585) 2.090 1 .148 2.330 .740 7.336

Notes: ***P significant at 0.05 for the Wald test



Pa
ge

 
28

7

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

The Logistic Model 
A multiple linear regression model can be written as 
follows where the left-hand side expresses the log-odds or 
logit: 

the younger respondents were more likely to experience 
hardship financing than the older respondents.

Education 
The positive coefficient (β=0.198) with the primary 
education level, the negative coefficient (β=-0.106) with 
the secondary and higher secondary education level, and 
the positive coefficient (β=1.848) with the graduation/
master’s level represented the change in the log-odds of  
incurring hardship financing compared to those with 
no education. Hence, the respondents with primary 
education and graduation/master’s level had a much 
higher likelihood of  experiencing hardship financing than 
those without education. In contrast, the respondents 
with secondary and higher secondary education had less 
likelihood of  experiencing hardship financing than the 
respondents with no education.
The OR [exp(β)=1.218] indicated that the respondents 
who had primary education were 1.218 times (21.8%) 
more likely to report hardship financing than those who 
had no education; the respondents who had secondary 
and higher secondary education were 0.899 times (10.1%) 
less likely to report hardship financing than those who 
had no education; the respondents who had graduation/
master’s degree education were 6.346 times (534.6%) 
more likely to report hardship financing than those who 
had no education, holding all other variables constant. 
Thus, education level was a good predictor of  incurring 
hardship financing by the recorded value of  the odds ratio, 
and was statistically significant with hardship financing.

Family Size
The estimated change in the logit/log-odds for every 
1 member increases with the family sizes, lowest up to 
3 and 4-7 members at -1.949 and -1.452, respectively. 
Thus, the negative coefficients (β=-1.949 and β=-1.452) 
represented that the respondents of  the family size lowest 
up to 3 or 4-7 members were less likely to incur hardship 
financing than those of  7 or above.
The OR [exp (β)=0.142] indicated that the respondents 
of  the family size lowest up to 3 were 0.142 times (85.8% 
lower) less likely to incur hardship financing compared 
to the reference category (the respondents with family 
sizes of  7 and above. Added to this, the OR [exp (β) 
=0.234] indicated that respondents of  family sizes of  4-7 
members were 0.234 times (76.6% lower) less likely to 
incur hardship financing than those of  7 and above.

Distance to the Hospital
The estimated change in the logit/log-odds of  hardship 
financing for every 1 km increases with the distance of  
the hospital from the residence. The coefficients (β=-
1.251, when the respondents lived within the distances of  
the hospital of  lowest up to 3 km and β=-1.149,  when 
the respondents lived within the distances of  the hospital 
between 4-20 km) represented the estimated change in 
the log-odds of  incurring hardship financing compared to 
the respondents who lived within the distances of  above 

The equation can be re-written also as follows:
 

Where π is the event probability, α is the Y-intercept,   
B1.......BK are parameters of  the slope, and X1....XK  
are the explanatory variables. α and β’s is estimated by 
the maximum likelihood estimator (MLE) approach 
(Abdulqader, 2017; Peng et al., 2002).

The Model of  Binary Logistic Regression
The following model was run:

Where α (alpha) is the intercept (2.293), X1, ......., X12 are 
the explanatory variables, β1,........., β12 are the regression 
coefficients, and Ɛ  the error term. 

Factors Influencing Hardship Financing by Using 
BLR Analysis
In the BLR analysis, regression coefficients and odds 
ratios (OR) were interpreted with factors affecting 
hardship financing. They were as follows:

Socioeconomic and Demographic Parameters with 
Hardship Financing 
The β’s are the regression coefficients (Table 7), indicating 
the direction of  the relationship between the independent 
variables and the logit of  hardship financing. Hence, a 
direct or positive relationship exists between them when 
β’s is positive. On the other hand, a negative relationship 
exists between the two when β’s is smaller than zero 
(negative). The analysis is as follows:

Age Level
The estimated value of  β implies that the estimated 
change in the logit/log-odds of  hardship financing is for 
every 1-year increase in age level. The positive coefficients 
(β=0.853 and β=0.711) indicated that the respondents in 
the age level lowest up to 30 and between 31-45 years 
were more likely to incur hardship financing compared to 
the respondents aged 45 and above. Besides, the age level 
was statistically significant with hardship financing.
The OR [exp (β)=2.348] indicated that the respondents 
of  the age level lowest up to 30 years were 2.348 times 
more likely to experience hardship financing compared to 
the reference category (the respondents of  the age 45 and 
above). Added to this, the OR [exp (β)=2.037] indicated 
that the respondents in the age level between 31-45 years 
were 2.037 times more likely to experience hardship 
financing than the respondents aged 45 and above. Hence, 



Pa
ge

 
28

8

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

20 km. As the coefficients were negative, it expressed 
that the respondents who lived closer to the hospital (less 
than 20 km) were less likely to incur hardship financing 
compared to those who lived across away (above 20 km). 
The OR [exp (β)=0.286] indicated that the respondents 
who lived within the distances of  the hospital at the 
lowest up to 3 km, were 71.4% less likely to incur 
hardship financing than the respondents who lived within 
the distances of  above 20 km. Added to this, the OR 
[exp (β)=0.286] indicated that the respondents who lived 
within the distances of  the hospital between 4-20 km 
were  (or 68.3%) less likely to report hardship financing 
than the respondents who lived within distances of  above 
20 km, holding all other variables constant.  However, 
the distance was statistically significant with hardship 
financing.

Healthcare Expenditures with Hardship Financing
Annual Outpatient Expenditures
The positive coefficients (β=1.851, when the respondents’ 
annual outpatient expenditures were up to 1000 BDT 
and β=.846, when the respondents’ annual outpatient 
expenditures were between the 1001-5000 BDT) 
represented the change in the log-odds of  incurring 
hardship financing compared to the annual outpatient 
expenditures above 5000 BDT among respondents. 
The OR [exp (β)=6.366] indicated that the respondents 
with annual outpatient expenditures up to 1000 BDT 
were 6.366 times more likely to incur hardship financing 
than those with annual outpatient expenditures of  above 
5000 BDT. The OR [exp (β)=2.330] indicated that the 
respondents with annual outpatient expenditures between 
1001-5000 BDT were 2.330 times more likely to report 
incurring hardship financing than those with outpatient 
expenditures above 5000 BDT annually, holding all 
other variables constant. Thus, a strong and statistically 
significant relationship existed between annual outpatient 
expenditures and hardship financing. Nevertheless, the 
analysis confirmed that annual outpatient expenditures 
were a strong predictor of  incurring hardship financing 
by the recorded value of  the odds ratio. 

Morbidity Parameters with Hardship Financing
Chronic Illness
The negative coefficient (β=-0.828) represented the 
change in the log-odds (logit) of  incurring hardship 
financing among respondents who have suffered from 
chronic illness compared to those having no chronic 
illness. However, it expressed that having a chronic illness 
among respondents reduced the likelihood of  incurring 
hardship financing compared to the reference category 
(the respondents having no chronic illness).
The OR [exp (β) =.437] indicated that the respondents 
having chronic illness had 56.3% lower odds (1 - 0.437 = 
0.563 or 56.3% decrease) of  incurring hardship financing 
compared to the reference category (the respondents 
who have no chronic illness). However, it expressed that 
the respondents with chronic illness were less likely to 

require hardship financing compared to the reference 
category. Nevertheless, the chronic illness was statistically 
significant with hardship financing.

Discussion
The younger respondents may face more financial 
instability, which leads to seeking hardship financing. 
On the other hand, the older respondents may be more 
financially established, which reduces their need for 
hardship financing.
The respondents with primary education have limited 
opportunities for a stable job. This allowed them to face 
financial instability, and increased the likelihood of  their 
incurring hardship in financing them. The respondents 
with secondary/higher secondary education have better 
job opportunities with better stable income. This allowed 
them to plan and manage healthcare financing in a better 
way, which reduced incurring hardship of  financing. 
The respondents with an education level of  graduation/
master’s degree may have higher living standards. It 
enhanced their living expenditures even in healthcare, 
leading to financial stress. Moreover, the graduates don’t 
always get their job they expect. However, a mismatch 
between education level and the demands of  the job 
market exists. Thus, on one side, the underemployment 
situation among graduates and on the other side, increased 
expenditures even in healthcare led to a financially 
unstable situation.
The individuals with large family sizes may face higher 
financial pressure due to more dependents. Hence, the 
situation led them to rely more on hardship financing. 
In contrast, smaller families may have fewer financial 
obligations or be better able to manage their finances. 
The respondents who lived closer to the hospital were 
financially more relieved by spending less money on 
seeking healthcare than those who lived farther distances 
from the hospital. Thus, they were facing less financial 
hardship.  On the other hand, the respondents who were 
living far from the hospital were delaying in seeking 
treatment, which led to more suffering from illnesses and 
facing financial hardship. 
The respondents with lower outpatient expenditures 
were more likely to incur hardship financing than 
the respondents with high outpatient expenditures. 
Although the result may seem quaint at initially, it 
reflected immeasurable socioeconomic dynamics and 
healthcare access in Bangladesh. Households incurring 
lower outpatient expenditures often belong to poorer 
socioeconomic conditions. Thus, the lower outpatient 
expenditures don’t necessarily indicate lesser healthcare 
needs but rather financial constraints which limits 
their ability to seek or afford care. Consequently, even 
minimal outpatient expenditures can strain their limited 
resources, leading to hardship financing mechanisms 
such as borrowing or selling assets. On the contrary, high 
outpatient expenditures may reflect greater healthcare 
needs and financial ability, which allowing them to access in 
healthcare services without financial hardship. Moreover, 



Pa
ge

 
28

9

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

a time gap between the inflow of  income and outflow 
of  health expenditures can lead to hardship financing 
(Binnendijk, 2014). Added to this, the incalculable timing 
of  hospitalization and the urgent need for large amounts 
of  funds increase the risk of  hardship financing.
Having a chronic illness, the respondents were facing 
less hardship financing compared to those who had no 
chronic illness, which was an exception. Many indirect 
costs could be generated, such as loss of  income of  
the chronically ill patient, and transportation costs, 
which could independently lead to the risk of  hardship 
financing. However, the study explored that the 
respondents with chronic illnesses might plan better for 
healthcare expenditures compared to those with sudden 
illnesses (without chronic illnesses) and those incurring 
no hardship financing.

CONCLUSION
Improvement in healthcare accessibility can be ensured 
by financial assistance. Education cannot be a safeguard 
for financial distress; thus, educational outcomes should 
be aligned with labor market demand. However, the 
government should take initiatives to promote a better 
job market to reduce the unemployment situation. 
Besides, education policies should be developed by 
emphasizing skills relevant to the present job market 
to reduce underemployment situations. Thus, this will 
reduce the financial distress in healthcare financing. 
Moreover, the policymakers may extend financial support 
to people suffering from non-chronic diseases for those 
facing financial hardship. Those initiatives would enhance 
financial risk protection and would mitigate vulnerability 
due to the devastating economic effects of  health shocks. 
The frequency of  outpatient healthcare utilization is 
substantially higher than that of  inpatient healthcare 
utilization, which is a crucial consideration. Hence, health 
reforms are needed to tackle a less stable (declining) 
financial situation due to high outpatient expenditure, 
which is a significant predictor of  financial hardship. 
Thus, subsidized healthcare services or health insurance 
schemes can be introduced to reduce the economic 
burden among respondents. Additionally, accessing a 
wealthier social network may increase the likelihood 
of  households being able to obtain interest-free loans, 
which may protect against the risk of  hardship financing. 
Besides, establishing more healthcare facilities in remote 
areas with assured quality should be emphasized. 
Hence, in order to achieve universal health coverage 
and to protect households against financial hardship, 
policymakers should consider these determinants.

REFERENCES
Asfaw, A., Lamanna, F., & Klasen, S. (2010). Gender gap 

in parents’ financing strategy for hospitalization of  
their children: Evidence from India. Health Economics, 
19(3), 265–279. https://doi.org/10.1002/hec.1468

Ahmed, S., Hoque, M. E., Sarker, A. R., Sultana, M., Islam, 
Z., Gazi, R., & Khan, J. A. M. (2016). Willingness-

to-pay for community-based health insurance among 
informal workers in urban Bangladesh. PLOS ONE, 
11(2), e0148211. https://doi.org/10.1371/journal.
pone.0148211 

Binnendijk, E., Koren, R., & Dror, D. M. (2012). Hardship 
financing of  healthcare among rural poor in Orissa, 
India. BMC Health Services Research, 12, 23.

Chuma, J., Gilson, L., & Molyneux, C. (2007). Treatment-
seeking behaviour, cost burdens and coping strategies 
among rural and urban households in Coastal Kenya: 
An equity analysis. Tropical Medicine & International 
Health, 12(5), 673–686. https://doi.org/10.1111/
j.1365-3156.2007.01825.x

Ezeoke, O. P., Onwujekwe, O. E., & Uzochukwu, B. S. 
(2012). Towards universal coverage: Examining costs 
of  illness, payment, and coping strategies to different 
population groups in Southeast Nigeria. American 
Journal of  Tropical Medicine and Hygiene, 86(1), 52–57. 
https://doi.org/10.4269/ajtmh.2012.10-0660

Hosmer, D. W., & Lemeshow, S. (2000). Applied logistic 
regression (2nd ed.). John Wiley & Sons.

Park, H.-A. (2013). An introduction to logistic regression: 
From basic concepts to interpretation with particular 
attention to nursing domain. Journal of  Korean Academy 
of  Nursing, 43(2), 154–164. https://doi.org/10.4040/
jkan.2013.43.2.154

Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. 
(2010). Multivariate data analysis (7th ed.). Pearson 
Prentice Hall.

Islam, M. R., Rahman, M. S., Islam, Z., Zhang, B. C., 
Sultana, P., & Rahman, M. M. (2017). Inequalities in 
financial risk protection in Bangladesh: An assessment 
of  universal health coverage. International Journal for 
Equity in Health, 16, 59. https://doi.org/10.1186/
s12939-017-0556-4

Jahan, T. (2022). The importance of  attribution of  
primary health care services for community-based 
health facility according to the health service users in 
Kapasia, Gazipur. Journal of  Scientific and Technological 
Research, 4(1).

Kolesar, R.J., Erreygers, G., & Leng, S. (2023). Hardship 
financing, productivity loss, and the economic cost of  
illness and injury in Combodia. International Journal for 
Equity in Health, 22, 208

Kruk, M. E., Goldmann, E., & Galea, S. (2009). 
Borrowing and selling to pay for health care in low- 
and middle-income countries. Health Affairs (Millwood), 
28(4), 1056–1066. https://doi.org/10.1377/
hlthaff.28.4.1056

Leive, A., & Xu, K. (2008). Coping with out-of-pocket 
health payments: Empirical evidence from 15 African 
countries. Bulletin of  the World Health Organization, 86, 
849–856.

McIntyre, D., Thiede, M., Dahlgren, G., & Whitehead, 
M. (2006). What are the economic consequences for 
households of  illness and of  paying for health care 
in low- and middle-income country contexts? Social 
Science & Medicine, 62(4), 858–865. 



Pa
ge

 
29

0

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 4(2) 280-290, 2025

Peng, C. Y. J., Lee, K. L., & Ingersoll, G. M. (2002). 
An introduction to logistic regression analysis and 
reporting. The Journal of  Educational Research, 96(1), 
3–15.

Peduzzi, P., Concato, J., Kemper, E., Holford, T. R., & 
Feinstein, A. R. (1996). A simulation study of  the 
number of  events per variable in logistic regression 
analysis. Journal of  Clinical Epidemiology, 49(12), 1373–
1379.

Russell, S. (2004). The economic burden of  illness for 
households in developing countries: A review of  
studies focusing on malaria, tuberculosis, and human 
immunodeficiency virus/acquired immunodeficiency 
syndrome. American Journal of  Tropical Medicine and 
Hygiene, 71(2), 147–155.

Smith, P. C., Mossialos, E., Papanicolas, I., & Leatherman, 
S. (Eds.). (2009). Performance measurement for health 
system improvement: Experiences, challenges and prospects. 
Cambridge University Press.

Steinhardt, L. C., Waters, H., Rao, K. D., Naeem, A. J., 
Hansen, P., & Peters, D. H. (2009). The effect of  
wealth status on care seeking and health expenditures 
in Afghanistan. Health Policy and Planning, 24(1), 1–17.

Senaviratna, N. A. M. R., & Cooray, T. M. J. A. (2019). 
Diagnosing multicollinearity of  logistic regression 
model. Asian Journal of  Probability and Statistics, 5(2), 
1–9.

Tahsina, T., Ali, N. B., Siddique, M. A. B., Ahmed, S., 
Rahman, M., Islam, S., Rahman, M. M., Amena, B., 
Hoque, D. M. E., Huda, T. M., & Arifeen, S. E. (2018). 
Determinants of  hardship financing in coping with 
out-of-pocket payment for care seeking of  under-five 
children in selected rural areas of  Bangladesh. PLOS 
ONE, 13(5), e0196237. https://doi.org/10.1371/
journal.pone.0196237

Tadiwos, Y. B., Kassahun, M. M., & Mebratie, A. D. (2025). 
Catastrophic and impoverishing out-of-pocket health 
expenditure in Ethiopia: Evidence from the Ethiopia 
Socioeconomic Survey. Health Economics Review, 15, 15. 
https://doi.org/10.1186/s13561-025-00602-1 

United Nations Development Programme. (2019). 
Universal health coverage: Sustainable development – Issue 
brief. https://www.undp.org/publications/universal-
health-coverage-sustainable-development-issue-brief

Vakhitova, Z. I., & Alston-Knox, C. L. (2018). Non-
significant p-values? Strategies to understand and 
better determine the importance of  effects and 
interactions in logistic regression. PLOS ONE, 
13(11), e0205076. https://doi.org/10.1371/journal.
pone.0205076

Van Doorslaer, E., O’Donnell, O., Rannan-Eliya, R. P., 
Somanathan, A., Adhikari, S. R., & Garg, C. C. (2006). 
Effect of  payments for health care on poverty estimates 
in 11 countries in Asia: An analysis of  household 
survey data. The Lancet, 368(9544), 1357–1364.


