Out-of-pocket and catastrophic healthcare expenditure in Bangladesh: a nationwide comparative study of households with and without disabled members

Abstract

Background

Despite Bangladesh’s heavy reliance on out-of-pocket (OOP) healthcare financing and the absence of disability-specific financial protection provisions, no nationally representative study has compared OOP expenditure and catastrophic health expenditure (CHE) between households with and without disabled members in Bangladesh. The study aims to estimate OOP healthcare expenditure, incidence of CHE, and to identify socioeconomic factors associated with CHE incidence among households with and without disabled members in Bangladesh.

Methods

This cross-sectional study used secondary data from the Bangladesh Household Income and Expenditure Survey (HIES) 2022. The Washington Group short set of questions was administered to identify self-reported disability. CHE was defined using the capacity-to-pay approach, applying a 40% threshold of non-food expenditure. Survey-weighted multivariate binary logistic regression models were estimated to identify factors associated with CHE incidence.

Results

Of the 13,931 surveyed households, 2,732 (19.13%) had at least one disabled member. Among households incurring healthcare costs, those with disabled members had significantly higher mean monthly OOP expenditure (BDT 4,972 vs. BDT 3,288; p < 0.001) and allocated a greater share of non-food expenditure to healthcare (27.76% and 18.99%, respectively). CHE incidence was substantially higher in those households (23.46%; CI: 21.56–25.48), with an overall incidence of 15.53%. Multivariate regression analysis identified chronic illness (AOR: 5.20; CI: 2.99–9.04; p < 0.001), inpatient service use (AOR: 3.27; CI: 2.47–4.33; p < 0.001), and private facility use (AOR: 2.71; CI: 1.72–4.26; p < 0.001) as the strongest predictors of CHE in households with disabled members. Asset quintile was not significantly associated with CHE in those households, although it was significant in other households. No significant difference in coping strategy distribution was identified between groups (p = 0.687). Descriptively, households with disabled members relied more on regular income and borrowing, and less on savings, though these patterns are exploratory.

Conclusions

Households with disabled members face substantially higher OOP expenditure and CHE despite comparable non-food expenditure, indicating a potential gap in Bangladesh’s financial protection policy. Findings suggest that closing this gap may require policy responses including strengthening public provision of specialised disability care, integrated care pathways for disability and chronic disease management, and routine disability-disaggregated data monitoring to track progress.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12939-026-02944-6.

Background

Persons with disabilities face systemic barriers across education, employment, healthcare, and social participation worldwide, and Bangladesh, a low and middle-income country (LMIC), is no exception to that [1]. Discrimination, stigma, and a lack of awareness further increase these challenges in the country. These barriers translate directly into household-level financial vulnerability, such as limited economic opportunities, lower wages, and additional healthcare expenses related to disability care and rehabilitation, assistive devices and accessibility modifications. Moreover, due to the vulnerable health conditions, disabled individuals need comprehensive and consistent healthcare services because of their disability-related health problems in addition to general health conditions, which create higher medical expenditures than those of persons without disability [2, 3]. All of these together simultaneously increase expenditure, creating a compounding burden that is particularly acute when households must pay for healthcare out-of-pocket [4]. A systematic review of 150 LMIC studies found that 81% reported a positive association between disability and poverty markers [5]. In Bangladesh, the cumulative income loss attributable to disability has been estimated at approximately US$ 1.18 billion per year, equivalent to 1.74% of GDP, reflecting severely constrained household capacity to absorb unplanned healthcare costs [6]. Moreover, persons with disabilities die up to 20 years earlier than those without disabilities, have twice the risk of developing health conditions, and face many health inequities [2]. It is precisely this intersection of reduced income and elevated healthcare expenditure that places households with disabled members at heightened risk of catastrophic health expenditure (CHE).

In Bangladesh, an estimated 5.71% of the total population lives with disabilities, and a significant number of disabled people live in lower-income households [7, 8]. This socio-economic group is already found to be most exposed to CHE risk in the existing financial protection literature [9–11]. This burden is projected to intensify as the population ages. According to the Bangladesh Bureau of Statistics (BBS), 9.25% (15.3 million) or one in 10 of Bangladesh’s population is aged 60 years or older, and according to the World Bank, by 2050, this population will reach one in five [12, 13]. Since functional difficulty and disability are directly associated with ageing [2], this demographic shift will substantially increase the number of households facing disability-related healthcare costs and constrained income security, amplifying the population exposed to the specific OOP and CHE risk. Bangladesh relies heavily on out-of-pocket (OOP) financing for healthcare, with the OOP share of current health spending rising from 64% in 2000 to 74% in 2020, one of the highest rates globally, making CHE a persistent and structurally embedded challenge [14]. Financial protection from healthcare costs is a core component of Universal Health Coverage under SDG target 3.8, and OOP expenditure and CHE incidence are its principal monitoring indicators [15]. Despite this policy priority, Bangladesh’s Health Care Financing Strategy (2012–2032) contains no disability-specific financial protection provisions, leaving a population with demonstrably higher healthcare costs without targeted safeguards. Bangladesh, therefore, offers an ideal research setting: a well-established CHE evidence base for direct comparison, rising OOP costs, a rapidly growing disability-associated demographic, and newly available nationally representative data. These make it not merely a convenient LMIC case; however, the most analytically appropriate setting in South Asia [9–11, 14, 16, 17].

A body of research from Bangladesh has examined CHE incidence and its socioeconomic and geographic distribution. Khan et al. estimated the CHE incidence and its poverty linkage nationally [10]; Ahmed et al. quantified CHE across income quintiles and administrative divisions using HIES 2016 [11]; Sheikh et al. examined disease-specific and time-trend dimensions of financial protection [18]; Rahman et al. decomposed rural-urban CHE inequality [9]; and Alam et al. applied multilevel modelling to map district-level CHE variation [19]. Collectively, these studies have advanced the understanding of financial protection across economic and geographic strata. However, none has stratified analysis by disability status or examined whether the magnitude, determinants, or coping responses associated with CHE differ systematically between households with and without disabled members. This is a critical omission that remains underrepresented in the evidence base, particularly in LMICs - a gap explicitly highlighted in review literature [20]. Using HIES 2022 data, this study addresses this gap and, to our knowledge, is the first to compare CHE by disability status between affected and non-affected households.

The pathway through which disability leads to CHE is conceptualised through an integrated framework. Disability simultaneously constrains household capacity to pay through constrained income and employment; however, it elevates OOP expenditure through higher healthcare utilisation, chronic comorbidities, and disability-specific requirements [20, 21]. As OOP expenditure rises against a constrained non-food expenditure denominator, the probability of exceeding the 40% catastrophic threshold increases systematically [21, 22]. When CHE occurs, households resort to distress financing strategies that deplete future financial resilience, perpetuating the disability-poverty trap and heightening vulnerability to subsequent health shocks [5].

Using data from HIES 2022, the most recent nationally representative household survey in Bangladesh, this cross-sectional, descriptive and comparative study has three specific objectives: (1) to compare mean OOP healthcare expenditure and the incidence of CHE between households with and without disabled members; (2) to identify socioeconomic factors, that are independently associated with CHE incidence in the overall population and separately in households with and without disabled members; and (3) to compare the distribution of healthcare coping strategies between disability and non-disability households. The findings are intended to generate household-level comparative evidence on financial protection by disability status, which will enrich international discussions surrounding the realisation of Universal Health Coverage and the pursuit of health equity. Consequently, this could catalyse prospective policy adaptations and interventions geared towards alleviating the economic strain of healthcare costs on persons with disabilities in Bangladesh and internationally.

Methods

Data source, study design and settings

This cross-sectional, population-based study utilised secondary data from the latest seventeenth round of the Household Income and Expenditure Survey (HIES) 2022, conducted by the Bangladesh Bureau of Statistics [7]. A cross-sectional design was appropriate for this study as it enables the estimation of OOP healthcare expenditure and CHE, and the identification of associated socioeconomic factors at a single point in time using a nationally representative sample. Consistent with this design, causal relationships cannot be established; findings are interpreted as associations.

HIES is a nationally representative survey that serves as the standalone data source for estimating household income, expenditure, consumption, inequality, poverty, health status, and healthcare-related costs of people in Bangladesh. The HIES 2022 employed a two-stage stratified random sampling technique. In the initial stage, 720 primary sampling units (PSUs) were selected using a probability proportional to size sampling method from all 64 nationwide districts. Households were randomly chosen from each PSU in the second stage. Consequently, 14,400 households (720 PSUs multiplied by 20 households per PSU) were selected for inclusion in the survey, and 13,931 households provided disability-associated data and were included for analysis. Details of the sampling technique and household selection for the survey can be found in the HIES 2022 final report [7].

The survey was carried out through face-to-face interviews in 2021. The primary respondent was the household head or another adult member (aged 18 or older) with comprehensive knowledge of all household members. Before each interview, oral consent was obtained from the respondents, and all participants’ identities were anonymised before data publication. The HIES implemented rigorous supervision and quality control measures to minimise biases and enhance data accuracy. The designated survey organisation was responsible for identifying and resolving data inconsistencies, verifying information, and managing missing data through established protocols. The survey adhered to internationally standardised household expenditure survey methodology, consistent with World Bank Living Standards Measurement Study (LSMS) guidelines, ensuring comparability with equivalent surveys in the region. Before analysis, all variables were checked for completeness, and a complete-case analysis was applied. 18 observations with missing data were excluded (0.13% of the total sample of 13,931 households) for the asset quintile variable, indicating negligible missingness and minimal risk of selection bias.

Identification of households with disabled member(s)

Disability status was identified by analysing the HIES 2022 survey data collected by the BBS. As part of the survey, the BBS administered the Washington Group short set of questions to all household members regardless of age, except for children under 2 to 3 years, due to the lack of discernible cognitive symptoms. This exclusion reflects the original HIES data collection protocol, and no additional exclusions were applied. Through face-to-face household interviews, BBS enumerators asked respondents whether any family member experienced difficulty in one or more of six functional domains: vision, hearing, mobility (walking and climbing), cognitive function (remembering and concentrating), self-care, and speech or communication. Functional difficulty data were collected even when assistive devices, such as glasses or hearing aids, were in use at the time of the interview. A household member was classified as having a disability if the respondent reported any level of difficulty, i.e., some difficulty, a lot of difficulty, or cannot do at all in at least one of the six functional domains. As the classification was used in the HIES reports, this threshold was not modified for the current analysis to track progress and enable comparison with other studies using the same HIES dataset, although this might have led to an overestimation of disability. The implications of this classification decision for the study’s key estimates are discussed further in the Limitations section. Households where no such difficulties were classified as households without any members with disabilities. This binary classification was used for primary comparisons of OOP expenditure and CHE incidence between disabled and non-disabled households. However, it masks heterogeneity within the disability group, combining households with varying severity and number of affected members despite likely differences in cost burdens.

Dependent variable: average out-of-pocket (OOP) expenditure for health and incidence of CHE

The OOP health expenditure was estimated by combining all the outpatient costs, inpatient costs, and costs for prostheses and orthoses incurred by households over a specified period. This approach is consistent with the WHO financial protection framework, which defines OOP payments as all direct payments made by households to health service providers at the time of service use [23].

Outpatient expenses were documented over the preceding 30 days, while inpatient costs were generated over the prior 12 months. Outpatient expenses encompassed consultation fees, medication, diagnostic procedures, transportation costs, and expenses for immunisation, contraception, routine medicine and medical check-ups borne by the households. Inpatient costs comprised expenditures related to surgery, consultations, accommodation/bed charges, medication, diagnostics, transportation, maternal healthcare, and any other formal or informal payments made within the preceding 12-month period. The cost for prostheses and orthoses included the cost of spectacles, hearing aids, surgical belts, glass eye, crutch, wheelchair, massage material, dentures, medical bed, neck brace, medicated shoes and others. All cost components were standardised to monthly equivalents before aggregation. Outpatient costs were recorded over 30 days and required no conversion, while 12-month inpatient costs were divided by 12 to derive monthly equivalents. Mean expenditure with standard deviation (SD), median, and interquartile range (IQR) are reported, given the right-skewed distribution of OOP data. This conversion approach is consistent with standard practice in household health expenditure surveys using mixed recall periods, including prior Bangladesh CHE studies and the World Bank Living Standards Measurement Study (LSMS) methodology [9–11]. For reader convenience and precision, OOP expenditure was calculated and reported using two complementary measures, each serving a distinct analytical purpose. The first measure, OOP_1, divides total household healthcare cost by the total number of surveyed households (Eq. 1), including those with no spending during the reference period, to show the overall financial burden across the population. The second measure, OOP_2, divides total healthcare cost only among households that incurred any costs (Eq. 2), reflecting the actual burden on those who made payment for healthcare services and better capturing affordability at the point of care

1

where Ci is the total monthly healthcare cost for household i, and N is the total number of surveyed households.

2

where M is the number of households with positive healthcare spending (M ≤ N). Notably, third-party payments such as those from health insurance were not considered in any healthcare expenditure category as HIES 2022 records direct household payments at the point of care without capturing subsequent insurance reimbursements. As a supplementary measure, values in BDT were converted to USD in the results section using an exchange rate of 84.8 BDT per USD, reflecting the average rate during the three‑month data collection period.

The incidence of CHE was estimated using the capacity-to-pay (CTP) approach, considering the fraction of a household’s healthcare cost exceeding 40% of the household’s non-food expenditure [21, 22]. The CTP approach is considered as it uses non-food expenditure as the denominator, excluding essential subsistence spending to provide a more accurate measure of household financial burden than total expenditure-based approaches. Additionally, the 40% CTP threshold is consistent with prior nationally representative CHE studies in Bangladesh, ensuring direct comparability with the existing evidence base [9–11, 18, 19, 24]. Sensitivity analysis using 25% of non-food expenditure was conducted to assess robustness; results are shown in Supplementary Table S1. We included both monetary and in-kind payments for all goods and services, along with the estimated monetary value of homemade product consumption, as part of household consumption expenditure. For each household, the proportion of non-food expenditure allocated to healthcare was calculated as:

3

denotes total monthly non-food expenditure for household i, used as a proxy for household capacity to pay. Finally, we calculated the proportion of out-of-pocket (OOP) payments in total non-food expenditure and thereafter the incidence of catastrophic health expenditure (CHE) using the following equations;

4

Independent variables

Independent variables for this study were selected based on the existing literature on this topic and the availability of variables in the dataset. These variables include a range of socio-demographic factors and are consistent with prior Bangladesh CHE studies [9–11, 18, 19]. The area of residence was categorised into rural and urban based on the BBS administrative classification. Gender of the household head was coded as male and female, and their education level into five distinct categories according to the education system in Bangladesh (no formal education, primary level education, secondary level education, higher secondary and undergraduate education and graduate and higher education). The presence of younger members in the household was considered if the household had any member less than 16 years old, while the presence of older adults was considered having a member over 65 years old. Household size was categorised as 1–2 members, 3–4 members, and 5 or more members. Administrative division was included as an eight-category variable corresponding to Bangladesh’s eight divisions. The household head’s employment status labelled “currently earning” was defined as a binary variable (yes/no) based on the HIES 2022 question asking whether the household head was engaged in any income-generating activity at the time of the survey, regardless of employment type or sector. The presence of chronic illness was defined as a binary variable (yes/no) based on the self-report of whether any household member suffered from a chronic illness in the last 12 months. Inpatient service utilisation was defined as at least one household member having been admitted overnight to a health facility during the preceding 12 months. Private facility utilisation was defined as at least one household member having used a private healthcare provider during the reference period.

Asset quintile was estimated and used as the proxy of households’ living standards and economic capacity, following the approach of Filmer and Pritchett (2001), widely applied in LMIC household survey research [25]. PCA was performed on 56 binary and categorical indicator variables drawn from the HIES 2022 household questionnaire, covering three domains: dwelling characteristics (construction material of walls, roof, floor etc.), access to basic amenities (source of drinking water, type of sanitation facility, access to electricity, cooking fuel type etc.), and ownership of durable goods and assets (television, refrigerator, mobile phone, motorcycle, bicycle, land ownership, livestock etc.). All categorical variables were converted to binary dummy indicators before PCA, and an index score was created. The index was then classified into five equal groups: lowest, 2nd, 3rd, 4th, and highest.

Regression analysis

Three distinct multiple regression models were utilised to identify the associated factors of the CHE incidence for three household compositions: overall households, households with disabled members and households without disabled members. In each model, the dependent variable is the incidence of CHE. All the independent variables were included in the regression models regardless of their individual statistical significance. A total of 18 observations (0.13% of the total sample) were excluded from the regression analysis due to missing data on asset quintiles. Consequently, the final sample comprised 13,913 observations overall, including 11,183 households without disabled members and 2,730 households with disabled members. The multiple logistic regression models were specified as:

In the models, Y_i represents the incidence of CHE coded as 0 (not facing CHE) or 1 (facing CHE). In all models, β₀ is the intercept and β₁ through β_k are the adjacent coefficients to the corresponding independent variables X_1i through X_ki. Adjusted Odds Ratios (AOR) with 95% Confidence Interval (CI) were calculated to assess the association between dependent and independent variables in the multiple logistic regression models. Model significance was assessed using the Wald chi-square test. Model fit was evaluated using pseudo-R². Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC, c-statistic). Multicollinearity among independent variables was assessed using the variance inflation factor (VIF), with mean VIF values reported. All regression models incorporated the HIES 2022 complex survey design, including PSU clustering, stratification, and household sampling weights. Data were analysed using Stata, College Station, Texas, USA.

Results

Characteristics of the sample

Table 1 presents the demographic and socio-economic characteristics of the study participants, with weighted values. Of the 13,931 households included in the analysis, 11,199 (80.87%) had no member with a disability, while 2,732 (19.13%) had at least one member with a disability. Among the latter, 1,191 households (8.55%) had one disabled member, 555 households (3.98%) had two disabled members, and 920 households (6.60%) had more than two disabled members. The sample indicates that 68.04% of households are located in rural areas, and 87.75% are male-headed. The majority of household heads are aged between 31 and 45 years (39.87%), with a considerable proportion lacking formal education (30.85%). Most household heads are economically active (82.8%). A large portion of the sample comes from the Dhaka division (29.34%), and the predominant household size is three to four members (48.61%). Additionally, 71.27% of households have at least one member suffering from a chronic illness in the last 12 months, while more than half (52.87%) sought healthcare services for illness, and 11.98% required inpatient care. Notably, only 14.76% of households utilised private healthcare services. Households are evenly distributed across five asset quintile groups based on living conditions, durable goods, and asset ownership. The distribution patterns for households without disabled members were largely consistent with the overall population. However, households with disabled members exhibited some variations. These households had a higher proportion of older household heads, with 26.11% aged over 60 years old, and a greater percentage of household heads without formal education (35.59%). Additionally, the proportion of economically active household heads was lower (74.96%) compared to both the overall sample and households without disabled members. Furthermore, a greater proportion of households with disabled members included any elderly member (43.83%), and a higher incidence of chronic illness was observed among household members (84.72%).

Table 1.

Demographic information of the study participant households (N = 13,931 households)

Variables	Overall (N = 13,931)	Households without disabled members (n = 11,199)	Households with disabled members (n = 2,732)
	N (%)	N (%)	N (%)
Area of residence
Rural	9479 (68.04)	7496 (66.93)	1987 (72.72)
Urban	4452 (31.96)	3703 (33.07)	745 (27.28)
Gender of the head of the household
Male	12,225 (87.75)	9825 (87.73)	2400 (87.85)
Female	1706 (12.25)	1374 (12.27)	332 (12.15)
Age of household head
Below 30 years	1748 (12.55)	1499 (13.38)	246 (9.02)
31 to 45 years	5555 (39.87)	4676 (41.76)	873 (31.92)
46 to 60 years	4254 (30.54)	3356 (29.96)	900 (32.95)
More than 60 years	2374 (17.04)	1668 (14.9)	713 (26.11)
Education level of the head of the household
No formal education	4298 (30.85)	3330 (29.73)	972 (35.59)
Primary level	3535 (25.37)	2877 (25.69)	657 (24.05)
Secondary level	4113 (29.53)	3365 (30.05)	747 (27.33)
Higher secondary/undergrad	807 (5.79)	655 (5.85)	152 (5.55)
Graduate and more	1178 (8.45)	972 (8.68)	204 (7.48)
Household head currently earning
No	2396 (17.2)	1719 (15.35)	684 (25.04)
Yes	11,535 (82.8)	9480 (84.65)	2048 (74.96)
Having a child in the household
No	3765 (27.03)	2948 (26.33)	819 (29.98)
Yes	10,166 (72.97)	8251 (73.67)	1913 (70.02)
Having an older person in the household
No	10,752 (77.18)	9200 (82.15)	1535 (56.17)
Yes	3179 (22.82)	1999 (17.85)	1197 (43.83)
Division
Barisal	731 (5.25)	564 (5.04)	167 (6.13)
Chittagong	2568 (18.44)	1872 (16.71)	703 (25.72)
Dhaka	4087 (29.34)	3530 (31.52)	550 (20.12)
Khulna	1497 (10.75)	1273 (11.37)	222 (8.14)
Mymensingh	992 (7.12)	816 (7.29)	175 (6.4)
Rajshahi	1839 (13.2)	1435 (12.82)	404 (14.8)
Rangpur	1505 (10.81)	1196 (10.68)	310 (11.34)
Sylhet	712 (5.11)	513 (4.58)	201 (7.36)
Household size
1 to 2	1737 (12.47)	1410 (12.59)	327 (11.95)
3 to 4	6772 (48.61)	5735 (51.21)	1028 (37.64)
5 and more	5422 (38.92)	4054 (36.2)	1377 (50.41)
Any member of the household suffering from chronic illness
No member	4003 (28.73)	3574 (31.92)	417 (15.28)
Have member	9928 (71.27)	7625 (68.08)	2315 (84.72)
At least one household member sought care for a chronic illness
No	6566 (47.13)	5327 (47.56)	1238 (45.32)
Yes	7365 (52.87)	5872 (52.44)	1494 (54.68)
At least one household member utilised an inpatient service
No	12,262 (88.02)	9992 (89.23)	2265 (82.92)
Yes	1669 (11.98)	1207 (10.77)	467 (17.08)
At least one household member utilised a private facility
No	11,875 (85.24)	9590 (85.64)	2283 (83.58)
Yes	2056 (14.76)	1609 (14.36)	449 (16.42)
Asset quintiles 1
Lowest	3007 (21.61)	2409 (21.54)	599 (21.93)
2nd	2880 (20.7)	2355 (21.06)	524 (19.19)
3rd	2892 (20.78)	2322 (20.76)	570 (20.88)
4th	2614 (18.79)	2057 (18.4)	557 (20.43)
Highest	2520 (18.11)	2040 (18.24)	480 (17.57)
Households with disabled members status
No disabled member in the household	11,265 (80.87)
Have a member in the household	1191 (8.55)
Have two members in the household	555 (3.98)
Have more than two members in the household	920 (6.6)

n = count; % = weighted column percentage; ¹ Sample sizes are 13,913 for the overall population, 11,184 for households without disabled members, and 2,731 for households with disabled members. The total sample size is lower due to 18 reported missing cases for the asset quintile variable in the dataset

OOP expenditure: households with disabled members vs. without disabled members

Table 2 outlines the average monthly healthcare and non-food expenses incurred by participant households. The average monthly non-food expenditure by households was estimated at BDT 17,427 (US$ 205.5) overall, BDT 17,312 (US$ 204.2) for households without any disabled member, and BDT 17,911 (US$ 211.2) for households with any member with disabilities.

Table 2.

OOP expenditure comparison between households with and without disabled members in the last 30 days (in BDT)

Cost items (Household level)	Overall household			Households without any disabled members (A)			Households with any disabled member (B)			Differences (B-A)	P- Value*
	N	Mean (SD)	Median (IQR)	N	Mean (SD)	Median (IQR)	N	Mean (SD)	Median (IQR)
Average monthly non-food expenditure	13,931	17427.01 (22397.96)	11582.08 (12608.75)	11,199	17312.43 (21069.63)	11537.17 (12427.58)	2,732	17911.12 (27305.3)	11678.67 (13396.33)	598.69	0.604
Among the total households (estimated using Eq. 1 for OOP1)
Total outpatient cost in BDT	13,931	2926.357 (6262.417)	1191 (2790)	11,199	2668.151 (6091.837)	1050 (2620)	2,732	4017.276 (6831.262)	1800 (3870)	1349.13	0.000
Total inpatient cost in BDT	13,931	421.0552 (2505.326)	0 (0)	11,199	345.3273 (2136.288)	0 (0)	2,732	741.0059 (3659.436)	0 (0)	395.68	0.000
Total prosthesis and Orthoses in BDT	13,931	7.173641 (41.5273)	0 (0)	11,199	6.3536 (40.1657)	0 (0)	2,732	10.63832 (46.69466)	0 (0)	4.28	0.002
Total healthcare cost in BDT	13,931	3354.585 (7117.382)	1295 (3145.833)	11,199	3019.832 (6763.245)	1150 (2835)	2,732	4768.92 (8305.144)	2040 (4486.667)	1749.09	0.000
Proportion of OOP of non-food expenditure	19.25			17.44			26.63
Among the households that made the payment (estimated using Eq. 2 for OOP2)
Total outpatient cost in BDT	12,894	3186.07 (6470.79)	1380 (2975)	10,275	2928.43 (6322.07)	1250 (2645)	2,619	4230.48 (6945.59)	1990 (3920)	1302.05	0.000
Total inpatient cost in BDT	1,748	3527.52 (6453.42)	1666.67 (2691.67)	1,293	3221.75 (5773.42)	1641.67 (2525)	455	4338.2 (7931.22)	1833.33 (3079.17)	1116.45	0.003
Total prosthesis and Orthoses in BDT	1,320	71.37 (112.18)	45.83 (50)	974	70.31 (115.62)	44.17 (50)	346	74.2 (102.55)	45.83 (36.67)	3.89	0.784
Total healthcare cost in BDT	12,997	3621.3 (7329.33)	1500 (3335)	10,361	3287.53 (6994.04)	1350 (3000)	2,636	4971.73 (8420.27)	2181.67 (4530)	1684.20	0.000
Proportion of OOP of non-food expenditure	20.78			18.99			27.76

BDT: Bangladeshi Taka; * Design-adjusted Wald tests; Exact p-values are reported to three decimal places; values below 0.001 are reported as p < 0.001

The mean OOP expenditure on healthcare over 30 days was 3,355 BDT (US$ 39.6) for the entire household (using Eq. 1). OOP is found higher for households with disabled members (BDT 4,769 or US$ 56.2) compared to households without any disabled members (BDT 3,020 or US$ 35.6), and this difference is statistically significant (p < 0.001). When examining OOP expenditure exclusively among those who incurred healthcare expenses, the average expenditure was 3,621 BDT (US$ 42.7) for the overall households (using Eq. 2). Households with disabled members also demonstrated significantly higher average OOP expenditure compared to those without disabilities (BDT 4,972 or US$ 58.6 versus BDT 3,288 or US$ 38.8, respectively).

Among all surveyed households, 19.25% of non-food expenditure was allocated to healthcare. This proportion escalates to 26.63% for households with any disabled member, compared to 17.44% for households without disabled members. Among households that made healthcare payments, the overall proportion was 20.78%, while the households with disabled members allocated 27.76% of non-food expenditure to healthcare compared to 18.99% for non-disabled households, a difference of 8.77% points.

Incidence of catastrophic health expenditure

Table 3 presents the incidence of catastrophic health expenditure (CHE) across different household compositions and compares overall percentages and confidence intervals (CI) between households with and without members with disabilities. Several estimates for households with disabled members present relatively wide confidence intervals, reflecting the smaller subsample size (n = 2,732) relative to households without disabled members (n = 11,199). Subgroup estimates with notably wide CIs should therefore be interpreted with caution as they may have limited precision. Findings indicate that the overall incidence of CHE stands at 15.53% (95% CI: 14.81–16.29) using the threshold of 40% of CTP. Households with disabled members experience a higher incidence of CHE (23.46%; 95% CI: 21.56–25.48) compared to those without disabilities (13.66%; 95% CI: 12.89–14.46). Among the overall households, CHE incidence is notably higher among rural residents (17.02%; CI: 16.1, 17.98) than urban ones (12.37%; CI: 11.26, 13.57), and male-headed households have a higher CHE incidence (15.76%; CI:14.98, 16.57) than female-headed ones (13.9%; CI: 11.99, 16.06). Household heads aged above 60 years had a higher CHE incidence (22.61%; CI: 20.63, 24.72). Economic factors further highlight that non-earning household heads face a greater financial burden (20.19%; CI: 18.25, 22.27) than earning ones (14.57%; CI: 13.8, 15.37). Smaller households and those with older persons have higher CHE (18.32%; CI: 16.16, 20.71) and (21.36%; CI: 19.68, 23.13), respectively. Households face CHE particularly when chronic illnesses are present (19.64%; CI: 18.7, 20.61). Healthcare utilisation trends reinforce the financial strain on households with disabilities, as they are more likely to seek care for chronic illnesses (21.48%; CI: 20.35, 22.66) vs. (8.86%; CI: 8.05, 9.74), use inpatient services (38.35%, CI: 35.54, 41.23; vs. 12.43%, CI: 11.72, 13.17), and private healthcare facilities (43%, CI: 38.25, 47.88); vs. 14.21%, CI: 13.5, 14.96). Moreover, CHE incidence is highest among lower-income households, with a decreasing trend in wealthier households. Sensitivity analyses using 25% of non-food expenditure thresholds confirmed the robustness of the primary findings: households with disabled members consistently exhibited higher CHE incidence than those without (39.77%, CI: 37.53, 42.05 vs. 24.72%, CI: 23.75, 25.72). Full results are presented in Supplementary Table S1.

Table 3.

Incidence of catastrophic health expenditure (using 40% of non-food expenditure threshold)

Variables	Overall	Households without any disabled members	Households with any disabled members
	Percentage (95% CI)	Percentage (95% CI)	Percentage (95% CI)
Area of residence
Rural	17.02 (16.1, 17.98)	15.08 (14.1, 16.11)	24.57 (22.26, 27.03)
Urban	12.37 (11.26, 13.57)	10.78 (9.63, 12.05)	20.52 (17.4, 24.03)
Gender of the head of the household
Male	15.76 (14.98, 16.57)	14.01 (13.19, 14.88)	23.15 (21.13, 25.29)
Female	13.9 (11.99, 16.06)	11.13 (9.23, 13.35)	25.75 (20.35, 32.01)
Age of household head
Below 30 years	14.43 (12.48, 16.62)	12.89 (10.9, 15.18)	24.05 (18.04, 31.3)
31 to 45 years	13.21 (12.14, 14.37)	11.3 (10.21, 12.48)	23.79 (20.42, 27.52)
46 to 60 years	15.07 (13.82, 16.42)	13.89 (12.54, 15.37)	19.61 (16.66, 22.93)
More than 60 years	22.61 (20.63, 24.72)	20.48 (18.23, 22.93)	27.72 (23.89, 31.91)
Education level of the head of the household
No formal education	17.9 (16.56, 19.33)	16.47 (15, 18.05)	22.96 (19.93, 26.29)
Primary level	15.89 (14.46, 17.43)	14.32 (12.81, 15.98)	22.98 (19.31, 27.12)
Secondary level	14.71 (13.4, 16.12)	12.78 (11.43, 14.27)	23.66 (20.04, 27.72)
Higher secondary/undergrad	14.49 (11.76, 17.73)	11.07 (8.47, 14.35)	29.73 (21.47, 39.57)
Graduate and more	9.41 (7.52, 11.72)	6.85 (5.15, 9.05)	22 (15.48, 30.28)
Household head currently earning
No	20.19 (18.25, 22.27)	16.76 (14.67, 19.09)	29.05 (24.99, 33.48)
Yes	14.57 (13.8, 15.37)	13.09 (12.28, 13.95)	21.59 (19.49, 23.86)
Having a child in the household
No	17.57 (16.13, 19.12)	15.66 (14.1, 17.35)	24.69 (21.24, 28.48)
Yes	14.78 (13.95, 15.65)	12.94 (12.08, 13.86)	22.94 (20.69, 25.35)
Having an older person in the household
No	13.81 (13.02, 14.64)	12.47 (11.66, 13.33)	22.11 (19.63, 24.8)
Yes	21.36 (19.68, 23.13)	19.12 (17.12, 21.3)	25.2 (22.32, 28.31)
Division
Barisal	23.56 (21.28, 26)	22.08 (19.58, 24.81)	28.69 (23.64, 34.33)
Chittagong	20.25 (18.26, 22.38)	17.77 (15.61, 20.15)	27.05 (22.83, 31.74)
Dhaka	12.34 (10.88, 13.97)	11.11 (9.62, 12.79)	20.53 (15.81, 26.22)
Khulna	16.03 (14.14, 18.13)	14.56 (12.6, 16.78)	24.69 (19.17, 31.18)
Mymensingh	15.42 (13.48, 17.59)	13.82 (11.79, 16.13)	23.14 (17.88, 29.38)
Rajshahi	11.7 (10.11, 13.51)	9.81 (8.17, 11.73)	18.65 (14.6, 23.51)
Rangpur	14.22 (12.36, 16.29)	12.1 (10.19, 14.31)	22.63 (17.83, 28.29)
Sylhet	20.4 (18.2, 22.78)	18.85 (16.39, 21.6)	24.45 (20.06, 29.45)
Household size
1 to 2	18.32 (16.16, 20.71)	16.2 (13.94, 18.75)	27.77 (22.14, 34.21)
3 to 4	13.95 (12.96, 15)	12.4 (11.38, 13.5)	22.83 (19.84, 26.11)
5 and more	16.62 (15.45, 17.86)	14.55 (13.28, 15.91)	22.92 (20.29, 25.78)
Any member of the household suffering from chronic illness
No member	5.36 (4.54, 6.3)	5.34 (4.48, 6.35)	5.48 (3.4, 8.72)
Have member	19.64 (18.7, 20.61)	17.56 (16.53, 18.63)	26.7 (24.54, 28.98)
At least one household member sought care for a chronic illness
No	8.86 (8.05, 9.74)	7.47 (6.64, 8.38)	15.04 (12.77, 17.64)
Yes	21.48 (20.35, 22.66)	19.27 (18.06, 20.54)	30.44 (27.64, 33.38)
At least one household member utilised an inpatient service
No	12.43 (11.72, 13.17)	10.95 (10.21, 11.73)	19.17 (17.25, 21.24)
Yes	38.35 (35.54, 41.23)	36.11 (32.89, 39.47)	44.32 (38.81, 49.97)
At least one household member utilised a private facility
No	14.21 (13.5, 14.96)	12.47 (11.72, 13.25)	21.78 (19.89, 23.79)
Yes	43 (38.25, 47.88)	39.63 (34.3, 45.22)	56.14 (45.9, 65.89)
Asset quintiles
Lowest	17.71 (16.08, 19.46)	16.39 (14.62, 18.33)	23.17 (19.35, 27.48)
2nd	16.24 (14.65, 17.97)	14.31 (12.65, 16.14)	25.21 (20.93, 30.03)
3rd	16.12 (14.56, 17.81)	14.1 (12.47, 15.89)	24.61 (20.5, 29.26)
4th	16.08 (14.44, 17.88)	14.45 (12.7, 16.4)	22.29 (18.33, 26.83)
Highest	10.95 (9.52, 12.56)	8.44 (7.06, 10.06)	21.91 (17.58, 26.96)
Members with disability
No member with disability	13.66 (12.89, 14.46)
Have a member	20.73 (18.1, 23.63)
Have two members	21.81 (17.96, 26.23)
Have more than two members	27.99 (24.54, 31.72)
Overall CHE (using 40% of the non-food expenditure threshold)	15.53 (14.81, 16.29)	13.66 (12.89, 14.46)	23.46 (21.56, 25.48)
Overall CHE (using 25% of the non-food expenditure threshold) a	27.6 (26.7, 28.52)	24.72 (23.75, 25.72)	39.77 (37.53, 42.05)

^a Full results are presented in Supplementary Table S1

Comparison of the CHE incidence between households having any member with and without disability is shown in Table 3.

Coping mechanisms

Table 4 presents the distribution of primary coping mechanisms adopted by households by disability member status. The analysis focuses on the primary coping strategy utilised by each household, reflecting their most accessible source of financial support. Results are reported as percentages with CI values. Among all surveyed households, regular income was the most common primary coping strategy (38.9%; 95% CI: 33.2–44.9), followed by household savings (32.1%; CI: 26.8–37.9), borrowing or loans (12.9%; CI: 9.4–17.4), donation and other assistance (10.0%; CI: 7.0–14.2), and selling household assets (6.1%; CI: 3.9–9.5). The distribution was broadly similar across both household groups: regular income and savings were the two most common strategies in households without disabled members (36.5%; CI: 30.0–43.5 and 33.5%; CI: 27.2–40.4, respectively) and also for households with disabled members (45.5%; CI: 34.6–56.9 and 28.2%; CI: 19.2–39.3, respectively). CI values of households with disabled members were substantially wider than those for other households, reflecting the smaller subsample of that group and reduced statistical precision. No statistically significant between-group differences were observed for any individual coping strategy (p = 0.687).

Table 4.

Differences in coping mechanisms for healthcare costs between households with and without disabled members (column percentage)

Coping strategies	Overall households, Percentage (95% CI)	Households with no disabled members, Percentage (95% CI)	Households with disabled members, Percentage (95% CI)	P-value
Regular income	38.9 (33.2–44.9)	36.5 (30.0–43.5)	45.5 (34.6–56.9)	0.687
Household savings	32.1 (26.8–37.9)	33.5 (27.2–40.4)	28.2 (19.2–39.3)
Selling household assets	6.1 (3.9–9.5)	6.8 (4.1–11.0)	4.4 (1.5–11.9)
Borrowing or loan	12.9 (9.4–17.4)	12.8 (8.8–18.4)	13.1 (7.5–22.0)
Donations and other	10.0 (7.0–14.2)	10.4 (7.0–15.3)	8.8 (3.8–19.1)

Estimates are weighted using survey design. Percentages represent column proportions. Pearson χ² test assesses differences in the distribution of coping strategies between households with and without disabled members

Associated factors for CHE: multiple regression analysis

In Table 5, findings from multivariate regression analysis identify the factors associated with the incidence of CHE in Bangladesh. All three models demonstrated acceptable to good discriminative ability, with AUC (c-statistics) values of 0.8102 (CI: 0.801, 0.820), 0.806 (CI: 0.796, 0.815), and 0.794 (CI: 0.784, 0.804) for Models 1, 2, and 3, respectively, indicating that the models adequately distinguish between households that experienced CHE and those that did not. In the overall population model (model 1), rural residence was significantly associated with higher odds of experiencing CHE (AOR: 1.24, CI: 1.08–1.43, p = 0.003), while female-headed households were associated with lower odds of CHE (AOR: 0.64, CI: 0.50–0.80, p < 0.001), as were households with an earning head (AOR: 0.74, CI: 0.61–0.89, p = 0.002). The presence of a child in the household was associated with reduced odds of CHE (AOR: 0.81, CI: 0.68–0.96, p = 0.018). Smaller household sizes (1–2 members) were associated with higher odds of CHE (AOR: 1.65, CI: 1.28–2.14, p < 0.001). Households in Dhaka, Mymensingh, Rajshahi, and Rangpur were associated with significantly lower odds of CHE compared to Sylhet. The presence of chronic illness in the household (AOR: 3.26, CI: 2.68–3.96, p < 0.001), seeking care for chronic conditions (AOR: 2.58, CI: 2.25–2.97, p < 0.001), inpatient services utilisation (AOR: 3.98, CI: 3.40–4.66, p < 0.001), and private healthcare facility use (AOR: 2.99, CI: 2.35–3.81, p < 0.001) were associated with significantly higher odds of CHE. Households belonging to the lowest asset quintile were associated with the highest odds of CHE (AOR: 2.02, CI: 1.57–2.59, p < 0.001), with a decreasing trend in the strength of association across quintiles. Additionally, having more than two disabled members in the household was associated with significantly higher odds of experiencing CHE (AOR: 1.75, CI: 1.4–2.19, p < 0.001) compared to households with no disabled members.

Table 5.

Factors associated with the incidence of CHE

Variables	Overall (Model 1) (n = 13,913)		Households without disabled members (Model 2) (n = 11,183)		Households with disabled members (Model 3) (n = 2,730)
	AOR (95%CI)	p-value	AOR (95%CI)	p-value	AOR (95%CI)	p-value
Area of residence
Rural	1.24 (1.08, 1.43)	0.003	1.21 (1.03, 1.43)	0.022	1.32 (1.01, 1.72)	0.045
Urban (reference)	1.0		1.0		1.0
Gender of the head of the household
Male (reference)	1.0		1.0		1.0
Female	0.64 (0.5, 0.8)	< 0.001	0.53 (0.4, 0.71)	< 0.001	0.95 (0.64, 1.42)	0.800
Age of household head
Below 30 years (reference)	1.0		1.0		1.0
31 to 45 years	0.92 (0.74, 1.14)	0.442	0.87 (0.68, 1.11)	0.270	1.1 (0.71, 1.7)	0.675
46 to 60 years	0.95 (0.76, 1.19)	0.646	0.96 (0.74, 1.25)	0.776	0.87 (0.55, 1.36)	0.532
More than 60 years	1.12 (0.86, 1.45)	0.392	1.1 (0.81, 1.49)	0.553	1.16 (0.72, 1.88)	0.540
Education level of the head of the household
No formal education	1.35 (0.98, 1.86)	0.062	1.8 (1.22, 2.66)	0.003	0.74 (0.42, 1.3)	0.299
Primary level	1.21 (0.88, 1.67)	0.241	1.56 (1.05, 2.31)	0.029	0.73 (0.42, 1.28)	0.276
Secondary level	1.22 (0.89, 1.67)	0.208	1.61 (1.09, 2.37)	0.017	0.71 (0.41, 1.22)	0.218
Higher secondary/undergrad	1.18 (0.81, 1.72)	0.395	1.2 (0.75, 1.92)	0.458	1.34 (0.68, 2.65)	0.396
Graduate and more (reference)	1.0		1.0		1.0
Household head currently earning
No (reference)	1.0		1.0		1.0
Yes	0.74 (0.61, 0.89)	0.002	0.69 (0.54, 0.87)	0.002	0.78 (0.58, 1.05)	0.103
Having a child in the household
No (reference)	1.0		1.0		1.0
Yes	0.81 (0.68, 0.96)	0.018	0.79 (0.64, 0.97)	0.025	0.91 (0.65, 1.27)	0.575
Having an older person in the household
No (reference)	1.0		1.0		1.0
Yes	1.15 (0.96, 1.37)	0.128	1.24 (0.99, 1.56)	0.056	0.99 (0.76, 1.31)	0.967
Division
Barisal	1.22 (0.99, 1.52)	0.065	1.33 (1.03, 1.71)	0.028	1 (0.67, 1.49)	0.988
Chittagong	1.07 (0.86, 1.34)	0.531	1.17 (0.89, 1.52)	0.255	0.89 (0.6, 1.32)	0.559
Dhaka	0.66 (0.53, 0.82)	< 0.001	0.64 (0.49, 0.83)	0.001	0.72 (0.46, 1.12)	0.147
Khulna	0.83 (0.66, 1.05)	0.116	0.82 (0.63, 1.07)	0.152	0.82 (0.52, 1.3)	0.404
Mymensingh	0.6 (0.48, 0.76)	< 0.001	0.56 (0.42, 0.74)	< 0.001	0.76 (0.48, 1.19)	0.226
Rajshahi	0.44 (0.35, 0.57)	< 0.001	0.43 (0.32, 0.58)	< 0.001	0.48 (0.31, 0.75)	0.001
Rangpur	0.59 (0.46, 0.75)	< 0.001	0.55 (0.41, 0.74)	< 0.001	0.65 (0.42, 1.01)	0.055
Sylhet (reference)	1.0		1.0		1.0
Household size
1 to 2	1.65 (1.28, 2.14)	< 0.001	1.72 (1.27, 2.33)	0.001	1.58 (0.98, 2.56)	0.061
3 to 4	1.24 (1.07, 1.44)	0.005	1.25 (1.05, 1.49)	0.013	1.27 (0.95, 1.7)	0.107
5 and more (reference)	1.0		1.0		1.0
Any member of the household suffering from chronic illness
No member (reference)	1.0		1.0		1.0
Have member	3.26 (2.68, 3.96)	< 0.001	3.03 (2.45, 3.74)	< 0.001	5.2 (2.99, 9.04)	< 0.001
At least one household member sought care for a chronic illness
No (reference)	1.0		1.0		1.0
Yes	2.58 (2.25, 2.97)	< 0.001	2.74 (2.33, 3.23)	< 0.001	2.27 (1.74, 2.97)	< 0.001
At least one household member utilised an inpatient service
No (reference)	1.0		1.0		1.0
Yes	3.98 (3.4, 4.66)	< 0.001	4.3 (3.57, 5.19)	< 0.001	3.27 (2.47, 4.33)	< 0.001
At least one household member utilised a private facility
No (reference)	1.0		1.0		1.0
Yes	2.99 (2.35, 3.81)	< 0.001	3.07 (2.31, 4.08)	< 0.001	2.71 (1.72, 4.26)	< 0.001
Asset quintiles
Lowest	2.02 (1.57, 2.59)	< 0.001	2.25 (1.66, 3.04)	< 0.001	1.39 (0.88, 2.18)	0.160
2nd	1.79 (1.4, 2.29)	< 0.001	1.86 (1.38, 2.52)	< 0.001	1.48 (0.96, 2.28)	0.073
3rd	1.63 (1.29, 2.07)	< 0.001	1.7 (1.27, 2.27)	< 0.001	1.33 (0.87, 2.03)	0.183
4th	1.39 (1.09, 1.76)	0.007	1.55 (1.16, 2.08)	0.003	0.99 (0.65, 1.5)	0.957
Highest (reference)	1.0		1.0		1.0
Members with disability
No member (reference)	1.0
Have a member	1.23 (1, 1.5)	0.047
Have two members	1.22 (0.9, 1.64)	0.201
Have more than two members	1.75 (1.4, 2.19)	< 0.001
Number of observations	13,913		11,183		2,730
Wald chi2	1104.12		855.94		225.77
Prob > chi2	< 0.001		< 0.001		< 0.001
Pseudo R2	0.1613		0.1643		0.1302
AUC (C-statistics) with 95% CI	0.8102 (CI: 0.801, 0.820)		0.806 (CI: 0.796, 0.815)		0.794 (CI: 0.784, 0.804)
Mean VIF	2.97		3.22		3.19

Exact p-values are reported to three decimal places; values below 0.001 are reported as p < 0.001; AUC (C-statistics): Area under the Receiver Operating Characteristic (ROC) curve; VIF: Variance Inflation Factor

In model 2 (households without disabled members), rural residence, non-earning household head status, and small household size remained significantly associated with the highest odds of CHE. Household heads with no formal education were associated with the highest odds of CHE in this subgroup (AOR: 1.80, 95% CI: 1.22–2.66, p = 0.003). The presence of chronic illness (AOR: 3.03, CI: 2.45–3.74, p < 0.001), healthcare-seeking for chronic conditions (AOR: 2.74, CI: 2.33–3.23, p < 0.001), inpatient service utilisation (AOR: 4.3, CI: 3.57–5.19, p < 0.001), and private facility use (AOR: 3.07, CI: 2.31–4.08, p < 0.001) were each associated with significantly higher odds of CHE in this subgroup.

In model 3 (households with at least one disabled member), rural residence was significantly associated with higher odds of CHE (AOR: 1.32, CI: 1.01–1.72, p = 0.043). However, education level and household size were not significantly associated with CHE in this subgroup. The presence of chronic illness showed the strongest association with CHE in this model (AOR: 5.2, CI: 2.99–9.04, p < 0.001). Similarly, inpatient service utilisation (AOR: 3.27, CI: 2.47–4.33, p < 0.001) and private healthcare facility use (AOR: 2.71, CI: 1.72–4.26, p < 0.001) were significantly associated with CHE. Asset quintile was not significantly associated with CHE in this subgroup across any quintile level, in contrast to its consistent and significant association in Models 1 and 2. Details of the findings from the multivariate regression analysis for different household compositions are presented in Table 5.

Discussion

This study presents compelling evidence that households with disabled members in Bangladesh face a higher burden of OOP healthcare expenditure, CHE, and economic vulnerability and identifies the factors associated with the incidence of CHE. This reinforces the argument that disability and financial vulnerability are deeply interconnected, forming a cyclical relationship that is difficult to escape without systemic intervention [5].

Analysis revealed that the non-food expenditures of households with and without disabled members were relatively similar (BDT 17,911 or US$ 211.2 vs. BDT 17,312 or US$ 204.2), indicating similar baseline consumption capacity. Despite this parity, average monthly OOP healthcare expenditure was significantly higher among households with disabled members (BDT 4,769; SD = 8,305) compared to households without any disabled members (BDT 3,020; SD = 6,763) (p < 0.001). This gap widened further when focusing solely on households that incurred healthcare spending (BDT 4,972; SD = 8420.27 vs. BDT 3,288; SD = 6,994). These findings align with global literature that documents the extra cost of disability, including unavoidable expenses for assistive devices, specialised care, and increased frequency of healthcare utilisation [20, 26, 27]. However, as expenditure was measured at the household level, direct attribution of these cost differences to disability-related care needs cannot be confirmed. Higher SD values in expenditure data are likely due to their characteristic right-skewedness, driven by substantial variation in spending among some households. Expressed as a proportion of non-food expenditure, households with disabled members allocated 27.76% to healthcare compared to 18.99% for non-disabled households, positioning the average healthcare-seeking disabled household 12.24% points below the 40% catastrophic threshold, compared to 21.01% points for non-disabled households. This proximity to the threshold directly explains the substantially higher CHE incidence observed in the disability group. Previous studies in Bangladesh and Pakistan similarly observed elevated OOP expenditures among individuals with disabilities [28, 29]. The World Health Organization (WHO) also emphasises these excess costs as structural barriers to health equity [30]. The findings of higher OOP among households with disabled members point to a potential policy blind spot. While Bangladesh’s Health Care Financing Strategy (2012–2032) outlines a goal to reduce OOP payments from 64% to 32%, it lacks specific provisions to address the distinct vulnerabilities of households with disabilities [31]. Evidence from our research of systematically higher OOP expenditure despite comparable non-food expenditure suggests that universal OOP reduction targets may be insufficient to protect disability households without complementary disability-specific measures. Possible directions informed by these findings include revision of the financing strategy to incorporate disability-adjusted provisions, risk-pooling mechanisms, and coverage of essential disability-related services. These approaches align with international recommendations advocating for prepayment and risk-pooling to prevent financial catastrophe [32].

Using the capacity-to-pay method, our analysis reveals a markedly higher incidence of CHE in households with disabled members (23.46%) compared to those without (13.66%), with an overall rate of 15.53%. The overall incidence of CHE shows a slight increase compared to the 2017 estimate of 10.9% in a similar context [11]. Overall, the CHE rate is higher than in neighbouring countries such as Nepal, and Afghanistan at 10.3%, and 13.53%, respectively, except Bhutan, which exhibited a very low incidence at 0.51%; although these comparisons should be interpreted cautiously given methodological differences in threshold and denominator across studies, as detailed in the Methods [33–35]. Our findings affirm that households with disabled members incur significantly higher healthcare expenses and are more likely to experience CHE compared to households without disabled members. The higher CHE incidence among disability households compared to non-disabled households has been documented in South Korea and other LMIC contexts; however, those studies relied on national health insurance claims data or single-disease samples rather than population-based household expenditure surveys that the current study adopts [36, 37]. The vulnerable health conditions of disabled members, as well as their mobility constraints, specialised healthcare, and the high costs of supportive devices, may compel households to spend more than usual, leading them to excessive healthcare spending and pushing them toward catastrophic situations [20, 38]. The highest CHE incidence was found for those using private healthcare facilities among households with any disabled members (56.14%; CI: 45.9, 65.89) and other households also, reflecting the substantially higher direct costs of private healthcare relative to public facilities in Bangladesh [39]. This pattern is consistent with the hypothesis that inadequate public provision of specialised disability care, including rehabilitation, assistive technology, and chronic disease management, may be driving households toward higher-cost private facilities. Strengthening public provision of specialised and need-based disability care services is an evidence-informed strategy that may help to reduce catastrophic health spending.

The socioeconomic factors associated with CHE in households with members with disabilities were distinct and more intense. It is worth noting that the between-group differences in covariate associations reported across multivariate regression models 2 and 3 are based on visual comparison of adjusted odds ratios rather than formal statistical testing of disability-covariate interaction terms. Also, wide CI for certain variables suggest variability and should be interpreted with caution. Chronic illness was strongly associated with CHE across all household groups, with the highest effect observed among households with disabled members (AOR: 5.2; 95% CI: 2.99–9.04). These findings highlight the compounding effect of disability and chronic disease in driving financial risk, consistent with Yeo and Sawyer’s documentation of the compounding burden of chronic illness and disability [40]. Moreover, fragmented care pathways, where disability and chronic disease management require separate referrals, generate cumulative OOP costs, potentially contributing to the higher expenditure observed in disability households. Integrated care delivery within a single pathway may represent a plausible structural response to this pattern; such reforms have been associated with reduced costs (0.87; CI 0.80–0.94) and improved outcomes (1.15; 95% CI 1.11–1.18) in a global systematic review [41]. However, this approach requires prospective evaluation to determine its effectiveness and feasibility in the Bangladesh context.

Female-headed households in households with disabled members showed higher crude CHE prevalence than male-headed households (25.75% vs. 23.15%). However, after accounting for these confounding factors, female headship appears to be associated with a reduced risk of CHE (OR: 0.95; CI: 0.64, 1.42), likely driven by socioeconomic structural disadvantages, such as lower income and differential access to healthcare, rather than household headship itself. The presence of more than two disabled household members was associated with 1.75-fold higher odds of CHE (AOR: 1.75, 95% CI: 1.40–2.19), indicating that CHE risk increases with the number of disabled members in a household, though the cross-sectional design of this study precludes causal inference.

Asset quintiles were found to be significantly associated with CHE among the general population and households without disabled members, consistent with findings from other global studies conducted with overall households [9, 10, 17, 19, 35, 37, 42, 43]. However, among households with any disabled member, the association is not statistically significant. One plausible hypothesis, consistent with evidence from Korea and other LMIC contexts, is that expenditure on assistive devices, specialist care, and chronic illness management may constitute substantial and unavoidable financial burdens, such that wealth status alone does not provide meaningful protection against catastrophic health spending among households with disabled members [36, 37]. However, this remains speculative without disaggregated cost analysis; future studies with itemised cost data are needed to confirm this.

The coping mechanism analysis found no statistically significant difference in the overall distribution of primary financing strategies between households with and without disabled members (p = 0.687), with both groups relying predominantly on regular income and household savings. This finding should be interpreted in light of the limited statistical power available within the disability subsample, where CIs were substantially wider, indicating that modest differences in coping strategy proportions would likely be undetectable at this sample size. Given the non-significant result, the observed numerical differences in borrowing and savings proportions between groups are treated as descriptive trends only and are not interpreted as evidence of differential coping behaviour. Future studies with larger disability-specific samples, longitudinal designs, and detailed coping sequence data would be better powered to formally test whether disability status is associated with differential financial coping behaviour in Bangladesh.

Limitations

While this study contributes valuable insights to the existing literature, several limitations should be considered when interpreting the findings. The use of cross-sectional secondary data precludes causal inference and restricts control over data collection processes and variable definitions, although HIES 2022 is a nationally validated survey with established quality control procedures. Although nationally representative, findings may mask sub-national disparities; further region-specific analyses are warranted. Self-reported expenditure data are subject to recall bias, and variations in CHE thresholds and estimation methods across studies necessitate caution when making direct comparisons with other literature.

Regarding disability measurement, the Washington Group Short Set of Questions was validated for adults. It may not reliably capture functional difficulties in pre-school-aged children below five to seven years, meaning disability prevalence and CHE incidence among households with young children may be underestimated. More substantively, inclusion of “some difficulty” into disability identification departs from the Washington Group’s recommended threshold and likely overestimates the number of households with a disabled member. As individuals at this severity level typically have lower disability-related healthcare needs than those with severe functional impairment, this may attenuate OOP expenditure and CHE differences between household groups. This classification was retained for consistency with the HIES 2022 report and to enable comparability with prior Bangladesh studies using the same dataset, though its influence on key estimates warrants careful consideration.

A related but distinct limitation concerns the heterogeneity masked by the binary classification of disability. Beyond severity, aggregating households across six functionally distinct domains (vision, hearing, mobility, cognition, self-care, and communication) into a single variable obscures substantially different cost structures and healthcare utilisation patterns associated with each disability type. Consequently, the reported OOP and CHE estimates represent averages across a heterogeneous population, and future research disaggregating by disability type would provide a more precise characterisation of financial risk across the disability spectrum. Endogeneity concerns may additionally affect the regression findings. Reverse causality cannot be excluded; financial catastrophe may worsen functional status over time while disability simultaneously generates higher healthcare costs. Finally, the absence of routine disability-disaggregated monitoring in Bangladesh’s national health information systems limits the ability to track intervention progress and identify emerging protection gaps over time. Integrating disability indicators into national surveys and health information systems would support more comprehensive monitoring of financial protection in future research.

Conclusion

This study provides the first nationally representative comparison of OOP healthcare expenditure and CHE between households with and without disabled members in Bangladesh, using HIES 2022 data and a validated disability identification instrument consistent with international standards. Households with disabled members face significantly higher OOP expenditure (BDT 4,769 vs. BDT 3,020) and CHE incidence (23.46% vs. 13.66%) despite comparable non-food expenditure to other household groups. Chronic illness, inpatient service use, and private facility use are the dominant drivers of CHE for all the household groups, specifically higher for households with disabled members. Asset quintile, a significant predictor in the overall households, was not significantly associated with CHE in households with disabled members, indicating that financial catastrophe in this group is driven by care-specific costs rather than wealth position alone. No statistically significant difference in coping strategies was detected between groups, though households with disabled members showed higher borrowing and lower reliance on savings.

Collectively, these findings indicate the presence of a potential gap in Bangladesh’s financial protection policy, whereby its primary instruments of OOP reduction targets and income-based safety nets may not adequately reflect the distinct financial vulnerability of households with disabled members. While the cross-sectional design of this study precludes causal inference, the evidence points toward the need for policy attention to this population, with possible directions including disability-sensitive revisions to Bangladesh’s Health Care Financing Strategy (2012–2032), cross-quintile healthcare subsidies, expanded social health insurance, and comprehensive coverage of disability-related services. Integrating routine disability-disaggregated indicators into national health information systems would further support monitoring of financial protection progress and identification of emerging protection gaps. The nationally representative scale of this analysis provides an empirical foundation for these policy considerations, with relevance to comparable LMICs where disability-inclusive health financing remains absent from national agendas.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

RA: Conceptualisation, Methodology, Formal analysis, Writing original draft; AM: Conceptualisation, Methodology, Supervision, Validation, Writing- Review, and editing; JL: Conceptualisation, Investigation, Writing- Review, and editing; AB: Conceptualisation, Supervision, Investigation, Validation. All authors read and approved the final manuscript.

Funding

This work was supported by the UK Engineering and Physical Sciences Research Council (EPSRC) grant EP/S02249X/1 for the Centre for Doctoral Training in Prosthetics and Orthotics.

Data availability

The datasets used and/or analysed during the current study are available from the Bangladesh Bureau of Statistics (BBS) on reasonable request to be sent to the Director General of BBS. Website: https://bbs.gov.bd/.

Declarations

Ethics approval and consent to participate

This study uses secondary data (Bangladesh Household Income and Expenditure Survey 2022) conducted by the Bangladesh Bureau of Statistics (BBS). Since this study utilised publicly available secondary data, no further ethical approval was needed.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.