A Multi-Criteria Decision Analysis Framework to Explore Determinants of Catastrophic Healthcare Expenses

Abstract

Despite significant advances in the medical sciences, out-of-pocket (OOP) healthcare costs have remained a concern, especially for lower-middle-class and poor people. The current study aims to investigate the critical factors that notably contribute to catastrophic healthcare expenses (CHCEs). To this end, the ongoing research is conducted through two phases. The first phase aims to identify the key determinants of CHCEs through expert and household evaluations. A multi-criteria decision analysis (MCDA) framework using the FullEX method is developed to analyze expert and household opinions. In the second phase, the experts investigate the hierarchical relationships among key determinants. Interpretive structural modeling (ISM) and MICMAC analysis are carried out to examine the structural relationships among the determinants. The findings of the FullEX analysis reveal that experts and households are in consensus. It is found that low-income level, number of dependable members, frequent birth rate, high child mortality, and lack of job security and risk pooling mechanisms notably contribute to the higher CHCEs. The ISM analysis indicates the strong driving power of income, education, and job security, leading to disparities in rural economic conditions, reflecting the need for holistic development. The MICMAC analysis confirms the hierarchical relationships among the key determinants of CHCEs. The findings necessitate formulating an inclusive strategy to reduce financial distress and improve the healthcare outlook for rural households, leading to sustainable development.

1. Introduction

Catastrophic healthcare expenditure (CHCE) refers to a financial crisis in which a household’s out-of-pocket health costs exceed a specified percentage of its income, typically 10% or 25%. This may result in significant financial hardship, including borrowing, liquidating assets, or reducing expenditures on necessities. It typically causes financial problems that make it hard to meet basic needs. It might even push the affected families into poverty. Families with CHCE typically have to give up fundamental needs, which makes their quality of life worse overall. In China, older adults with high CHCE experience a significant decline in quality of life [1,2]. CHCE compels families to give up essential items, such as food, education, and housing, which are important to their quality of life. Low-income families with chronic illnesses may spend their savings to cover extra medical expenses. This situation makes their financial condition worse [2,3,4]. Demographic differences also influence people’s ability to earn a living, with the elderly, people with disabilities, and those with multiple chronic conditions most affected. People living in rural areas sometimes have higher rates of CHCE because they cannot easily access healthcare services and supplies [5].

The number of CHCE events varies by location and socioeconomic status [3]. Families with lower incomes face higher risks because they lack adequate health insurance, and costs are rising. CHCE affects the quality of life and the ability to make a living for individuals and families, especially in low- and middle-income countries [2]. This happens when health systems lack sufficient financial protections, leaving many families in extreme poverty [1,2,6,7]. CHCE harms mental health and social cohesion, thereby undermining the quality of life and livelihoods of families [8]. As healthcare costs rise and the gap between rich and poor widens, it is important to develop effective ways to pay for healthcare that improve health, the economy, and social welfare [9].

About 60% of people in India pay for their own medical treatment, which may drive rural and urban families into poverty by 3.6% and 2.9%, respectively [10]. People choose more costly private institutions because the public health system is inadequate, with poor treatment and insufficient healthcare personnel [11]. The average monthly expense of treating chronic diseases like diabetes in India is roughly INR 15,000. This puts significant strain on the family’s finances [12]. Families with a cancer patient need to spend 37% to 49% of their monthly income on cancer treatment. The OOP healthcare expenditures in India constitute approximately 52% of overall health expenditures. It is among the highest worldwide in terms of health expenditure [13]. CHCE affects 12.1% of families, with around 7% of households experiencing it [14]. The National Sample Survey Office (NSSO) reported that over 60% of households in India rely on OOP expenses for healthcare purposes [15]. This condition impacts over 50% of households living below the poverty level in India. In India, the anticipated OOP cost for outpatient care is ₹756 in the lowest wealth quintile. The spending for urban households in the second-lowest wealth quintile is anticipated to be ₹1635. India’s public health spending is around 1.28% of GDP, much lower than that of other nations, hence constraining the government’s capacity to provide sufficient healthcare services [16]. Even with programs like Ayushman Bharat, out-of-pocket costs in India remain high [17].

CHCE has been a significant concern in health economics, especially in low- and middle-income nations where OOP payments predominate in public health funding. Such expenses frequently lead to financial distress, asset depletion, and heightened susceptibility for households without sufficient risk mitigation. Resolving this persistent issue necessitates a thorough analysis of the diverse elements affecting CHCE. Conventional econometric and health system assessments frequently neglect the intricate interconnections among socioeconomic, demographic, and institutional variables. A multi-criteria decision analysis (MCDA) technique provides a systematic framework for addressing this complexity through expert and stakeholder assessments [18] (Biswas et al., 2022). However, the extant literature does not demonstrate clear contributions from a holistic evaluation of the CHCE risk factors.

This research seeks to provide a scientific framework that prioritizes the elements affecting CHCE and their interconnections. The FullEX technique amalgamates expert and household viewpoints to ensure representational validity. The research uses ISM and MICMAC analyses to examine causal relationships among variables such as income, education, job security, and demographics. It addresses deficiencies in health funding policy by advocating inclusive measures to strengthen rural health systems and advance fair health financing that safeguards families while fostering universal healthcare and sustainable development in countries like India. The goals of the current research are:

• To discover the key issues that influence the CHCE of rural households
• To examine the causal relationships among various determinants of the CHCE
• To develop a holistic MCDA framework for comparative evaluation and examining hierarchical relationships

The present research offers several contributions, outlined as follows.

• This study offers a detailed framework for comprehending CHCE in vulnerable groups. It examines numerous determinants, including socioeconomic status, education, employment stability, and demographic characteristics, by integrating expert and household viewpoints.
• The methodology of this paper enhances current theories by elucidating the complex interconnections among financial, institutional, and demographic variables influencing CHCE, thereby advancing the discipline of health economics and management.
• The study utilizes the FullEX methodology to ascertain stakeholder priorities, facilitating a transparent evaluation of the intricate elements influencing CHCE. Furthermore, ISM and MICMAC analyses elucidate hierarchical and causal linkages among critical determinants, thereby providing a comprehensive basis for decision-making.
• The report functions as an evidence-based scientific approach for industry and government leaders to enhance health financing and to ensure social safety in resource-constrained environments. This, in turn, helps formulate inclusive policies for sustainable development.
• This paper incorporates the concept of reliability index and uses the integration and adjustment coefficients. The extended FullEX method allows more flexibility in decision-making.

The next part of this paper is organized as follows. Section 2 delineates the theoretical aspects covering past contributions. Section 3 outlines the research methodology. It discusses the details of the studies, data collection procedure, and steps of the proposed methodological framework. Key findings of data analysis are reported in Section 4. Section 4 also demonstrates the result of the comparison of various MCDA methods and sensitivity analysis. Section 5 discusses the inferences drawn from findings. It highlights key implications of this research. The future scopes are also discussed. Section 6 provides the concluding remarks to end this paper.

2. Theoretical Background

The present research employs a multi-theoretical lens to examine various determinants of CHCE.

2.1. Social Determinants of Health (SDH)

The SDH framework describes how different social factors affect health outcomes. It stresses the need for collaborative care models. It brings together community health workers and social workers to meet complex health needs and improve access to healthcare. Collaboration among different bodies is required to meet desired social needs [19,20,21] (Blebu et al., 2023; Hudon et al., 2022; Rogers et al., 2020). The socio-ecological model puts components into five groups: individual, interpersonal, institutional, community, and policy. It shows how factors such as house size, caring responsibilities, and regional disparities affect healthcare costs. People living in rural areas who know little about health face difficulty paying for medical treatment. This model improves the SDH framework by examining multiple levels [22,23] (Donnelly et al., 2023; Thompson et al., 2021).

2.2. Andersen’s Behavioral Model of Health Service (ABMHS)

This study applies ABMHS to understand CHCEs. It asserts that the utilization of health services is affected by predisposing factors such as age and gender, enabling factors such as income and insurance, need factors, and contextual variables [24,25,26,27] (Alves et al., 2025; Cheng et al., 2025; Halder & Kasemi, 2025; Heider et al., 2014). Predisposing factors influence care-seeking behavior, whereas facilitating factors affect access and expenses, with insufficient resources heightening vulnerability to CHCEs [28,29] (Babitsch et al., 2012; Kabir, 2025). Chronic illnesses significantly increase costs in inadequately safeguarded financial contexts. The present study highlights the relevance of ABMHS for detecting patterns of CHCEs and the necessity of targeted policy interventions to mitigate financial burdens, thereby offering a holistic framework for addressing health finance fairness [30] (Song et al., 2025).

2.3. Health Financing Equity Framework (HFE)

The HFE Framework [31] (Wagstaff & van Doorslaer, 2000) builds on ABMHS by focusing on the equity of healthcare financing and its impact on financial risk, specifically CHCE. This framework emphasizes justice in health funding allocation based on individuals’ capacity to pay, aligning with Andersen’s enabling variables, such as income and insurance coverage [32] (Hassani et al., 2025). While Andersen’s Model identifies factors influencing healthcare consumption, the HFE Framework evaluates the financial burdens of health expenditures, analyzing whether OOP costs are progressive or regressive and the incidence of catastrophic payments that may impoverish households. The integration of these frameworks enhances understanding of the factors affecting healthcare utilization. It addresses the inequities in health funding, ultimately aiming to improve access and protect disadvantaged groups from financial risks associated with CHCE.

2.4. Related Studies

There has been substantial research on the determinants of CHCE, which is a significant financial burden for families, particularly low-income households [33] (Wodniak et al., 2025). Ref. [34] Sarkar et al. (2025) identified income level, household size, education, and disease severity as significant predictors of CHCE in Bangladesh, particularly among uninsured households in the unorganized sector. Ref. [35] Rahman et al. (2024) employed multivariate analysis to demonstrate that rural–urban disparities, structural inequalities, and restricted healthcare access intensify CHCE in rural regions, highlighting persistent challenges with risk pooling and the disadvantages of out-of-pocket financing.

A study conducted by [36] Wright et al. (2025) revealed low socioeconomic status, chronic illness, and family composition as primary determinants of catastrophic health costs. This finding aligns with [37] Enemuwe and Oyibo (2025), who emphasized the unequal risk faced by rural households due to OOP expenses and restricted access to prepayment schemes. Ref. [38] Enuagwuna et al. (2024) highlighted correlations among income quintiles, healthcare-seeking behavior, and treatment expenses, particularly among rural households. The report highlights socioeconomic disparity and insufficient healthcare budget as significant barriers to fair healthcare access in Nigeria.

Ref. [39] Mwinuka and Mwemutsi (2024) emphasized the rising out-of-pocket expenses for Tanzanian households, particularly in rural areas, due to limited service provision and financial constraints. Ref. [40] Tadiwos et al. (2025) observed such patterns in Ethiopia, where economically marginalized populations encounter catastrophic health expenses, especially individuals with chronic illnesses lacking health insurance. Moreover, Ref. [41] Matebie et al. (2024) investigated financial strains on hospitalized cancer patients in Addis Ababa, finding that the exorbitant costs of specialized care frequently compel households to resort to distress financing and to liquidate assets. The correlation between sickness category and CHCE is vital, especially with non-communicable diseases (NCDs) in North India, which profoundly affect uninsured people [42] (Kansra et al., 2025). Ref. [43] Sriram et al. (2024) conducted a study using national survey data, revealing that economic status and healthcare consumption are significant determinants of differences in hospitalization-related catastrophic health expenditures across states. Furthermore, regional studies by [44] Panda et al. (2024) reported a high frequency of CHE among hospitalized patients in impoverished areas of India, underscoring the pressing need for enhanced social protection and subsidized healthcare systems.

Investigations of macro-level variables reveal significant insights. Ref. [45] Gul et al. (2024) found that GDP per capita, government health expenditure, and urbanization influence CHCE in Asian countries, suggesting that economic prosperity and investment in health systems are essential for protecting households from financial hardship. Ref. [46] Abodi et al. (2025) found in Iraq that low-income households, large family sizes, and chronic illnesses are significant predictors of CHCE, illustrating a trend in the Middle East in which insufficient safety nets result in regressive health financing. Research conducted by [47] Mohsin et al. (2024) and [48] Mulupi et al. (2025) emphasized the socio-ecological and social protection determinants of CHCE. Ref. [47] Mohsin et al. (2024) discovered significant factors at the individual, societal, and systemic levels, including demography, inequality, and insurance coverage. Ref. [48] Mulupi et al. (2025) examined chronic respiratory illnesses in Kenya and demonstrated that inadequate social protection and coping mechanisms exacerbate susceptibility. They advocated for disease-specific coverage and extensive welfare measures to bolster resilience.

2.5. Research Gaps

Based on the literature review above, the following significant research gaps are identified.

(a)

Prior studies primarily rely on regression analysis, decomposition techniques, or descriptive statistics to identify the drivers of CHCE. While effective for quantifying associations, these methods fail to assess the relative impacts of the components. Advanced MCDA frameworks, such as FullEX help comprehensive prioritization and structural understanding of the determinants.

(b)

The determinants are complex and interlinked. However, the literature review reveals a lack of understanding of their interrelations and causal connections. It recommends employing ISM-MICMAC methodologies to discern these connections and focus on primary and secondary determinants for efficient policy intervention.

(c)

No prior study has used multiple theories to determine the influencing factors of CHCE comprehensively. The current research draws on three relevant theories. Furthermore, no previous research has considered a combination of responses obtained from experts and households.

3. Materials and Methods

3.1. Description of the Determinants

The determinants of CHCE are identified by revisiting related past studies (

Table 1. Description of the determinants influencing CHCE.

S/L	Factor	Description	References
Aspect: Socioeconomic conditions
F1	Low household income level	Low-income households are profoundly affected by even modest health emergencies, resulting in catastrophic expenses.	[34,36,37,49]
F2	Lack of job security	Uncertain employment, without employer-sponsored health insurance, results in higher out-of-pocket expenditures.	[34,38,50] Xu et al., 2003; Sarkar et al., 2025; Enuagwuna et al., 2024
F3	Low education level	Low educational attainment reduces awareness of health insurance and preventive care, increasing the likelihood of catastrophic health expenditures.	[35,47,51] Van Doorslaer et al., 2007; Mohsin et al., 2024; Rahman et al., 2024
F4	Lack of assets and financial resilience	Insufficient funds or productive assets restrict the ability to cope during times of illness.	[41,48,52] Berman et al., 2010; Mulupi et al., 2025; Matebie et al., 2024
F5	Disparity in rural economic conditions	Rural households frequently experience income reductions and elevated indirect costs, thereby heightening their susceptibility to CHCE.	[35,39,49] Wagstaff et al., 2018; Rahman et al., 2024; Mwinuka & Mwemutsi, 2024
F6	Presence of dependent households	The presence of high-dependency individuals (children, the elderly, non-working members) exacerbates the financial strain during illness.	[36,40,50] Xu et al., 2003; Wright et al., 2025; Tadiwos et al., 2025
Aspect: Demographic and health-related issues
F7	Composition and size of households	Large families with numerous dependents enhance healthcare expenses.	[37,47,50] Xu et al., 2003; Mohsin et al., 2024; Enemuwe & Oyibo, 2025
F8	History of chronic illness and comorbidity	Prolonged management of non-communicable diseases (e.g., diabetes, hypertension) results in persistent out-of-pocket expenditures.	[42,44,51] Van Doorslaer et al., 2007; Kansra et al., 2025; Panda et al., 2024
F9	Presence of elderly people in the family	Elderly family members necessitate more frequent and expensive healthcare interventions.	[34,41,49] Wagstaff et al., 2018; Sarkar et al., 2025; Matebie et al., 2024
F10	Frequent birth rate and child health vulnerability	An increasing child birthrate and pediatric illnesses elevate the likelihood of CHCE	[36,47,53] Reddy et al., 2018; Mohsin et al., 2024; Wright et al., 2025
F11	Long-term disability	Expenditures for the care of disabled members impose a persistent and significant financial strain.	[40,48,52] Berman et al., 2010; Mulupi et al., 2025; Tadiwos et al., 2025
Aspect: Healthcare system-related issues
F12	Limited public health infrastructure	Insufficient local facilities compel patients to seek costly private providers.	[34,38,53] Reddy et al., 2018; Sarkar et al., 2025; Enuagwuna et al., 2024
F13	Costly diagnostic and treatment services	Out-of-pocket expenditures on necessary medications and diagnostics significantly contribute to CHCE.	[42,44,52] Wagstaff et al., 2018; Panda et al., 2024; Kansra et al., 2025
F14	Absence of prepayment or risk-pooling mechanisms	The lack of appropriate insurance or community-based funding exacerbates unplanned out-of-pocket expenditures.	[47,48,50] Xu et al., 2003; Mohsin et al., 2024; Mulupi et al., 2025
F15	Ineffective referral system in healthcare facilities	Patients frequently circumvent referral systems, resulting in worse primary care, longer wait times, and financial difficulties stemming from insufficient communication, rural–urban disparities, administrative barriers, and limited public facilities.	[39,53] Reddy et al., 2018; Mwinuka & Mwemutsi, 2024
F16	Poor quality of public services	Perceived substandard quality in public institutions necessitates dependence on expensive private treatment.	[35,52] Berman et al., 2010; Rahman et al., 2024
F17	Hidden costs in public facilities	Supplementary payments for services, pharmaceuticals, or diagnostics add to concealed out-of-pocket expenditures.	[36,37,51] Van Doorslaer et al., 2007; Wright et al., 2025; Enemuwe & Oyibo, 2025
Aspect: Policy and governance-related issues
F18	Ineffective and inadequate public insurance	Restricted and limited benefit packages within public insurance programs.	[34,46,49] Wagstaff et al., 2018; Sarkar et al., 2025; Abodi et al., 2025
F19	Ineffective resource allocation	Insufficient support in economically disadvantaged states or districts exacerbates susceptibility to CHCE.	[45,53] Reddy et al., 2018; Gul et al., 2024
F20	Ineffective health governance and accountability measures	Leakage, corruption, and inefficiency in budget allocation diminish access to affordable healthcare.	[46,52] Berman et al., 2010; Abodi et al., 2025
F21	Lack of robust regulation for the private healthcare sector	Unregulated pricing, insufficient transparency, and profit-oriented tactics drive up treatment costs.	[35,36,51] Van Doorslaer et al., 2007; Rahman et al., 2024; Wright et al., 2025
Aspect: Cultural and behavioral issues
F22	Poor health-seeking behavior and delay in care	Delayed or unsuitable treatment-seeking exacerbates illness severity and expenses.	[47,48,50] Xu et al., 2003; Mohsin et al., 2024; Mulupi et al., 2025
F23	Lack of trust in public services	Confidence in private healthcare and cultural inhibitions about public services, despite their high costs, lead to a preference for CHCE.	[35,36,49] Wagstaff et al., 2018; Rahman et al., 2024; Wright et al., 2025
F24	Gender disparity in decision-making	The restricted autonomy of women in healthcare expenditure decisions may result in delayed care and higher overall healthcare costs.	[37,47,53] Reddy et al., 2018; Mohsin et al., 2024; Enemuwe & Oyibo, 2025

) and through discussions with various households and experts.

3.2. Data Collection

As stated earlier, the current research has collected opinions from a large section of households in rural and semi-urban areas. We ensured that the sample of respondents was representative of the bottom of the pyramid to the extent possible. We followed a convenience sampling approach. The response rate was approximately 60 percent. We reached out to the respondents and collected their responses through discussions and observations. The sampling technique emphasized characteristics associated with CHE instead of aiming to reflect all socio-demographic factors. The literature [50,54,55] (Wagstaff & Doorslaer, 2003; Xu et al., 2003; Bolongaita et al., 2023) identified key determinants of CHE as income level, household size, age of the earning member, presence of chronic illness, and insurance coverage. Since we use the group decision-making approach to address the problem, the sampling was intended to ensure proportional representation across income categories and insurance status, which were identified as predictors of CHE. Furthermore, MCDM models emphasize knowledge relevance and stakeholder representation rather than demographic quotas [56,57] (Huang et al., 2011; Gao et al., 2018). The respondents and experts were selected based on their competence in the stated domain, engagement in health financing decisions, and knowledge of public health policy.

The number of determinants is 24. Thus, a sample of 250 households is satisfactory in size [58] (Mundfrom et al., 2005). The household respondents’ profiles are reported in

Table 2. Profile of the respondents (Households).

Demographic Variable	Category	Count	Percentage (%)
Age Group	18–25	33	13.2
	26–35	60	24.0
	36–45	85	34.0
	46–55	47	18.8
	56+	25	10.0
	Total	250	100.0
Income Group (monthly)	<10 k	75	30.0
	10 k–20 k	85	34.0
	20 k–30 k	60	24.0
	>30 k	30	12.0
	Total	250	100.0
Education Level	School level	100	40.0
	Undergraduate	87	34.8
	Post-graduate & above	63	25.2
	Total	250	100.0
Gender	Male	192	76.8
	Female	58	23.2
	Total	250	100.0

. We selected the pool of experts through snowball sampling. We obtained responses from 30 experts, which satisfy the requirements for group decision-making [59,60] (Thiel, 2008; Biswas & Pamucar, 2023). The profiles of the experts are reported in

Table 3. Profile of the respondents (Experts).

Category	Segment	Count	Percentage (%)
Experience (Years)	Less than 5 years	5	16.7
	6 to 10	10	33.3
	10 to 15	8	26.7
	More than 15	7	23.3
	Total	30	100.0
Education Level	Masters	18	60.0
	Ph.D.	12	40.0
	Total	30	100.0
Expertise	Rural development	12	40.0
	Public health	4	13.3
	NGOs	7	23.3
	Others	7	23.3
	Total	30	100.0

.

3.3. Methodology

The ongoing research is conducted through two stages: stage 1 (evaluation of the comparative importance of the determinants by experts and households) and stage 2 (examination of the hierarchical relationships among key determinants by experts). Stage 1 utilizes an MCDA approach based on the FullEX method [61] (Bošković et al., 2023). Stage 2 conducts an ISM-MICMAC analysis. The methodological steps are outlined below.

Stage 1. Identification of key determinants utilizing the FullEX method

Step 1.1. Formation of the input response matrix (IRM)

Suppose, $\left(\begin{array}{ccc}{x}_{11}& \dots & x_{1n}\\ ⋮& \ddots & ⋮\\ x_{m1}& \dots & x_{mn}\end{array}\right)$ be the IRM, where $x_{ij}$ is the rating of the determinant $C_j$ by the decision-maker $E_i$.

Step 1.2. Determine the weights of respondents

This study offers a combined analysis of household and expert opinions. Respondents’ weights play a critical role in the FullEX method. The weights of the experts are decided by their experience and education levels [61] (Bošković et al., 2023). However, the weights for rural household respondents are determined by their education level. From the literature review, it is evident that education plays a critical role in relation to CHCE.

Suppose ${\mathsf{\varpi }}_i^{\exp }(i=1,2,..,m)$ is the weight of the $i^{th}$ expert based on his/her experience. It can be obtained as follows.

$${\mathsf{\varpi }}_i^{\exp }={\displaystyle \frac{y_i}{{\displaystyle \sum _{i=1}^m{y}_i}}};{\mathsf{\varpi }}_i^{\exp }>0,{\displaystyle \sum {\mathsf{\varpi }}_i^{\exp }=1}$$

(1)

$y_i$ is the number of years (experience) of the corresponding expert. To simplify the calculation, the more years of experience an expert has, the higher the priority assigned to that expert.

Similarly, the education-based weights can be obtained as follows.

$${\mathsf{\varpi }}_i^{\mathrm{e}du}={\displaystyle \frac{l_i}{{\displaystyle \sum _{i=1}^m{l}_i}}};{\mathsf{\varpi }}_i^{\mathrm{e}du}>0,{\displaystyle \sum {\mathsf{\varpi }}_i^{\mathrm{e}du}=1}$$

(2)

$l_i$ is the education level of the corresponding expert. To simplify the calculation, the higher the education level, the greater the priority assigned to the corresponding experts.

The final weights of the experts can be obtained as a linear combination of experience and education, described below.

$${\mathsf{\varpi }}_i=(1-\alpha ){\mathsf{\varpi }}_i^{\exp }+\alpha {\mathsf{\varpi }}_i^{\mathrm{e}du};{\mathsf{\varpi }}_i>0,{\displaystyle \sum {\mathsf{\varpi }}_i=1};\alpha \in [0,1]$$

(3)

$\alpha$ is the adjustment coefficient, used to vary the priorities on experience and education. It is helpful to examine the sensitivity of the outcome to variations in expert weights. The value of α is set to 0.5 to give equal priority to education and experience.

The weights for the rural household respondents are derived solely using Equation (2).

Step 1.3. Normalize the IRM

The normalized values of the elements of IRM are obtained as follows.

$$x_{ij}^{*}={\displaystyle \frac{x_{ij}}{{\displaystyle \sum _{i=1}^m{x}_{ij}}}}$$

(4)

Step 1.4. Obtain the respondent’s weighted values of the normalized IRM (NIRM)

The weighted values are obtained as follows.

$${\zeta }_{ij}={\mathsf{\varpi }}_j{x}_{ij}^{*}$$

(5)

Step 1.5. Transform the weighted NIRM using the optimum values

The transformation of each element of the weighted NIRM is conducted as follows.

$${\xi }_{ij}={\displaystyle \frac{{\zeta }_{ij}}{\underset{i=1,2,..,m}{Max}\left({\zeta }_{ij}\right)}}$$

(6)

Step 1.6. Calculate the weights of the determinants

The calculation is performed through two steps. First, we obtain the total contribution of each determinant as follows.

$${\gamma }_j={\displaystyle \sum _{i=1}^m{\xi }_{ij}};j=1,2,..,n$$

(7)

The calculated weights of the determinants are derived as follows.

$$w_j={\displaystyle \frac{{\gamma }_j}{{\displaystyle \sum _{j=1}^n{\gamma }_j}}};w_j>0,{\displaystyle \sum w_j=1}$$

(8)

Steps 1.3 to 1.6 are used to determine the weights of the determinants based on experts’ views and households’ opinions separately.

Step 1.7. Obtain the importance scores of the determinants

The IS values are obtained by combining the weights of the determinants, derived by considering experts’ and households’ opinions. Accordingly, the IS values are obtained as follows.

$${\phi }_j=\beta w_j^{\exp ert}+(1-\beta )w_j^{household};\beta \in [0,1]$$

(9)

The integration coefficient β controls the emphasis on households and experts. This helps combine two perspectives, assess their consistency, and derive the determinants’ importance holistically. Further, it helps in the sensitivity analysis. The initial value of β is assumed to be 0.5.

Step 1.8. Calculate the values of the reliability index (RI)

The RI values are computed using the following steps [62] (Bošković et al., 2025) for both experts’ and households’ opinions.

(a)

Obtain the difference between the maximum and minimum values for each determinant in the weighted NIRM as follows.

$${\Delta }_j=\underset{i=1,2,..,m}{Max}\left({\zeta }_{ij}\right)-\underset{i=1,2,..,m}{Min}\left({\zeta }_{ij}\right)$$

(10)

(b)

The weights of the determinants are recalculated as follows.

$$w_j^{*}={\displaystyle \frac{{\Delta }_j}{{\displaystyle \sum _{j=1}^n{\Delta }_j}}};w_j^{*}>0,{\displaystyle \sum w_j^{*}=1}$$

(11)

(c)

Calculate the RI as demonstrated below.

$$RI={\displaystyle \sum _{j=1}^n\left|w_j-w_j^{*}\right|}$$

(12)

If the value of RI is less than 0.1, then the calculated criteria weights ($w_j$) is considered reliable [62] (Bošković et al., 2025).

Stage 2. Establishing the hierarchical relationships among key determinants utilizing the ISM-MICMAC analysis

The ISM-MICMAC analysis is carried out in several sequential steps [63,64,65,66,67] (Sushil, 2012; Khatwani et al., 2015; Ahmad & Qahmash, 2021; Almerino et al., 2024; Chi et al., 2025) as described below. Suppose, $F_k\left(k=1,2,..,k\right)$ denote the identified key determinants (as determined by the FullEX method).

Step 2.1. Establish the contextual relationships among key determinants

The experts discuss and, in consensus, decide the contextual relationships among the determinants. To this end, we first sought individual opinions about contextual relationships among the variables. Then, we follow the dominance theory to set the contextual relationship. Next, we discussed with the expert’s group and reiterate the process until the consensus was reached. The following notations are used to describe the contextual relationships between the determinants in each pair.

V → If the determinant in the row influences the corresponding determinant in the column

A → If the determinant in the row is influenced by the corresponding determinant in the column

X → If both row and column determinants influence each other

O → If there exists no relationship among the corresponding determinants.

Step 2.2. Formulate the structural self-interaction matrix (SSIM)

Based on the decided contextual relationships among the determinants, the SSIM is formulated. It is worth noting that the ISM method uses a single response to determine the SSIM. Thus, the SSIM is formulated through concurrent expert decisions [68] (Mkedder et al., 2024). Since the current study is based on the opinions of 30 experts, we resorted to opinion polls to reach a consensus.

Step 2.3. Obtain the reachability matrix (RM)

The final version of RM is obtained through two steps.

First, we formulate the initial RM based on the convention outlined below.

The IRM portrays the direct linkage between barriers. At this step, the symbols “V,” “A,” “X,” and “O” are replaced with 0 or 1.

• If a cell (i, j) contains “V”, the assigned value is “1” and the corresponding (j, i) cell is assigned with “0”.
• If a cell (i, j) contains “A”, the assigned value is “0” and the corresponding (j, i) cell is assigned with “1”.
• If a cell (i, j) contains “X”, the assigned value is “1” and the corresponding (j, i) cell is assigned with “1”.
• If a cell (i, j) contains “O”, the assigned value is “0” and the corresponding (j, i) cell is assigned with “0”.

In the second step, the transitivity property is applied to IRM to obtain the final RM. The underlying intention is to discover the indirect relationships. According to this property, if any determinant $F_g$ is related to $F_h$ and $F_h$ is related to $F_l$, then $F_g$ is indirectly related to $F_l$. Such a relationship is denoted by “1*”.

Step 2.4. Partitioning of the determinants into various levels

We will now establish the independent levels for the determinants. The segmentation of determinants into distinct, independent levels is contingent on reachability, antecedent, and intersection sets, derived from the application of the final RM. Reachability indicates the determinants that are counted and identified as influenced by a specific one. This is indicated by the 1 or 1* values in a particular row corresponding to the said determinant. On the other hand, the antecedent set represents the dependability of a specific determinant on others. This is by the 1 or 1* values in a particular column corresponding to the said determinant. Accordingly, the row sum indicates the driving power (reachability or influence on others), and the column sum is an indication of dependence power (dependability on others).

The cells that overlap the reachability and antecedent sets constitute the intersection set. The determinant for which the reachability set precisely corresponds to its interaction set is designated a superior level in the hierarchy. Multiple iterations accomplish this until all items are allocated to their respective levels, thus establishing a whole hierarchy. Level “1” represents the pinnacle of the ISM hierarchy.

Step 2.5. Construct the directional graph (Digraph)

The digraph represents the ISM model. It is a structural relationship diagram showing the relationships among various determinants. The digraph is constructed using the final table of the level partitioning. After constructing the initial digraph, the transitivity relations are removed to obtain the final model or digraph. The unidirectional and bidirectional arrows in the final model indicate the unidirectional and reciprocal influence between the determinants.

Step 2.6. Conduct the MICMAC analysis

MICMAC (Matrice d’Impacts Croisés Multiplication Appliquée à un Classement) analysis is a technique for assessing and classifying determinants based on their influence and interdependence within a system. It identifies essential drivers and understands interdependencies by calculating the driving and dependent powers of the determinants. The determinants are categorized into four types: autonomous, dependent, linkage, and independent. This classification assists researchers and decision-makers in identifying critical components and their functions, facilitating strategic planning and risk management in intricate multi-variable situations.

The present research uses widely used and simple Microsoft Excel for MCDM analysis. The tool developed by [65] Ahmad & Qahmash (2021) has been used for ISM analysis owing to its easy implementation and reliable results.

4. Findings

This section reports key findings of the two-stage data analysis. As stated earlier, the first stage is dedicated to evaluating various determinants of CHCE by experts and household respondents. The FullEX method is utilized for this purpose.

4.1. Stage 1: Identification of Key Determinants Utilizing the FullEx Method

The responses of the households and experts are analyzed separately. The responses are provided in the Supplementary File. There are 250 households and 30 experts who rated various determinants on a 7-point Likert scale (7: most important, 1: least important). The responses construct the IRM. The next step is to calculate the weights of the respondents (households and experts). The calculation of the experts’ weights is shown in Appendix A (

Table A1. Calculation of the weights of the experts (α = 0.5).

Expert	Qualification	Rating	Weight (Qual)	Experience (Years)	Rating	Weight (Exp)	Final
Expert 1	Ph.D.	2	0.0476	More than 15	4	0.0519	0.0498
Expert 2	Ph.D.	2	0.0476	10 to 15	3	0.0390	0.0433
Expert 3	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 4	Ph.D.	2	0.0476	More than 15	4	0.0519	0.0498
Expert 5	Masters	1	0.0238	Less than 5	1	0.0130	0.0184
Expert 6	Masters	1	0.0238	10 to 15	3	0.0390	0.0314
Expert 7	Ph.D.	2	0.0476	More than 15	4	0.0519	0.0498
Expert 8	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 9	Ph.D.	2	0.0476	10 to 15	3	0.0390	0.0433
Expert 10	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 11	Ph.D.	2	0.0476	More than 15	4	0.0519	0.0498
Expert 12	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 13	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 14	Ph.D.	2	0.0476	10 to 15	3	0.0390	0.0433
Expert 15	Masters	1	0.0238	Less than 5	1	0.0130	0.0184
Expert 16	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 17	Masters	1	0.0238	10 to 15	3	0.0390	0.0314
Expert 18	Masters	1	0.0238	Less than 5	1	0.0130	0.0184
Expert 19	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 20	Ph.D.	2	0.0476	More than 15	4	0.0519	0.0498
Expert 21	Ph.D.	2	0.0476	More than 15	4	0.0519	0.0498
Expert 22	Masters	1	0.0238	Less than 5	1	0.0130	0.0184
Expert 23	Masters	1	0.0238	Less than 5	1	0.0130	0.0184
Expert 24	Ph.D.	2	0.0476	10 to 15	3	0.0390	0.0433
Expert 25	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 26	Ph.D.	2	0.0476	More than 15	4	0.0519	0.0498
Expert 27	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 28	Masters	1	0.0238	6 to 10	2	0.0260	0.0249
Expert 29	Ph.D.	2	0.0476	10 to 15	3	0.0390	0.0433
Expert 30	Masters	1	0.0238	10 to 15	3	0.0390	0.0314

). Similarly, the weights of the household respondents are calculated. As explained earlier, household weights depend on household education levels. The calculation is shown in the Supplementary File. To calculate the weights of the experts, we set the initial value of the adjustment coefficient, α = 0.5.

Next, we normalize the IRM using Equation (4). The normalized IRMs based on expert responses and household opinions are shown in the Supplementary File. We then obtain the weighted NIRM (Equation (5)). Using the weighted NIRMs and applying Equations (6)–(8), we obtain the weights of the determinants (

Table 4. Calculation of weights of the determinants (Household response).

Determinant	F1	F2	F3	F4	F5	F6
γ	102.9524	100.8571	98.47619	97.52381	98.90476	102.9048
Weight	0.04365	0.04276	0.04175	0.04134	0.04193	0.04363
Rank	1	4	11	15	8	2
Determinant	F7	F8	F9	F10	F11	F12
γ	96.85714	95.14286	100.7143	102.1905	95.80952	97.57143
Weight	0.04106	0.04034	0.04270	0.04332	0.04062	0.04136
Rank	18	23	5	3	20	14
Determinant	F13	F14	F15	F16	F17	F18
γ	95.85714	100.3333	98.52381	98.85714	94.85714	96.90476
Weight	0.04064	0.04254	0.04177	0.04191	0.04021	0.04108
Rank	19	6	10	9	24	17
Determinant	F19	F20	F21	F22	F23	F24
γ	95.14286	99.47619	95.61905	97.42857	97.95238	97.95238
Weight	0.04034	0.04217	0.04054	0.04130	0.04153	0.04153
Rank	22	7	21	16	13	12

and

Table 5. Calculation of weights of the determinants (Expert response).

Determinant	F1	F2	F3	F4	F5	F6
γ	19.71429	17.71739	16.88768	16.44348	16.19928	19.18012
Weight	0.05340	0.04799	0.04575	0.04454	0.04388	0.05196
Rank	1	4	5	9	10	2
Determinant	F7	F8	F9	F10	F11	F12
γ	13.52174	13.43478	15.43478	18.46584	13.0942	14.43478
Weight	0.03663	0.03639	0.04181	0.05002	0.03547	0.03910
Rank	19	21	11	3	23	14
Determinant	F13	F14	F15	F16	F17	F18
γ	13.68841	16.6646	16.54348	15.35507	13.47826	14.21739
Weight	0.03708	0.04514	0.04481	0.04160	0.03651	0.03851
Rank	18	7	8	12	20	15
Determinant	F19	F20	F21	F22	F23	F24
γ	13.26087	16.70186	14.01449	11.47826	15.31159	13.90833
Weight	0.03592	0.04524	0.03796	0.03109	0.04148	0.03768
Rank	22	6	16	24	13	17

). We then obtain the importance scores for the determinants using Equation (9). As stated earlier, the integration coefficient β is set to 0.5.

Table A2. Consolidated weighting and ranking of the determinants.

		Ranks		Weights		Final
S/L	Determinants of Catastrophic Health Expenditure	Experts	Household	Experts	Household	Weight	Rank
F1	Low household income level	1	1	0.0534	0.0436	0.0485	1
F2	Lack of job security	4	4	0.0480	0.0428	0.0454	4
F3	Low education level	5	11	0.0457	0.0417	0.0437	6
F4	Lack of assets and financial resilience	9	15	0.0445	0.0413	0.0429	9
F5	Disparity in rural economic conditions	10	8	0.0439	0.0419	0.0429	10
F6	Presence of dependent households	2	2	0.0520	0.0436	0.0478	2
F7	Composition and size of households	19	18	0.0366	0.0411	0.0388	19
F8	History of chronic illness and comorbidity	21	23	0.0364	0.0403	0.0384	20
F9	Presence of elderly people in the family	11	5	0.0418	0.0427	0.0423	11
F10	Frequent birth rate and child health vulnerability	3	3	0.0500	0.0433	0.0467	3
F11	Long-term disability	23	20	0.0355	0.0406	0.0380	23
F12	Limited public health infrastructure	14	14	0.0391	0.0414	0.0402	14
F13	Costly diagnostic and treatment services	18	19	0.0371	0.0406	0.0389	18
F14	Absence of prepayment or risk-pooling mechanisms	7	6	0.0451	0.0425	0.0438	5
F15	Ineffective referral system in healthcare facilities	8	10	0.0448	0.0418	0.0433	8
F16	Poor quality of public services	12	9	0.0416	0.0419	0.0418	12
F17	Hidden costs in public facilities	20	24	0.0365	0.0402	0.0384	21
F18	Ineffective and inadequate public insurance	15	17	0.0385	0.0411	0.0398	15
F19	Ineffective resource allocation	22	22	0.0359	0.0403	0.0381	22
F20	Ineffective health governance and accountability measures	6	7	0.0452	0.0422	0.0437	7
F21	Lack of robust regulation for the private healthcare sector	16	21	0.0380	0.0405	0.0393	17
F22	Poor health-seeking behavior and delay in care	24	16	0.0311	0.0413	0.0362	24
F23	Lack of trust in public services	13	13	0.0415	0.0415	0.0415	13
F24	Gender disparity in decision-making	17	12	0.0377	0.0415	0.0396	16

reports the final calculated weights of the determinants and their ranking. We calculate the RI values for both cases (experts and households) using Equations (10)–(12). The values are obtained as follows.

RI (expert opinions) = 0.032 < 0.1; RI (household response) = 0.0421 < 0.1

Thus, we contend that the calculated weights are satisfactorily reliable.

4.1.1. Validation of Results

The validation is performed through three steps.

Step 1. Comparison of several MCDA models

Comparing many MCDM models is crucial for validating a particular model and ensuring generalizability, dependability, and consistent decision-making outcomes. Diverse MCDA methods may yield disparate rankings due to distinct aggregation processes, normalization strategies, and weighting schemes. Evaluating results reveals method-specific biases or computational inaccuracies, enhancing confidence in the ultimate decision [69,70] (Chaki et al., 2025; Biswas et al., 2026). This validation promotes the acceptance and implementation of the selected MCDA model. This paper uses three widely applied MCDA methods for group decision-making, such as LIWA, SWARA, and CIMAS, to rank the determinants. Then, it performs Spearman’s rank correlation test (

Table 6. Result of Spearman’s rank correlation test: Comparison of MCDA methods.

Method	LBWA	SWARA	CIMAS
FullEX	0.982 *	0.988 *	0.984 *

(* All coefficients of correlation are significant at the 0.05 level).

). The results show that FullEX maintains a statistically significant and high correlation with all other methods. This confirms the reliability of the results obtained by using the FullEX method.

Step 2. Calculation of the RI values

As stated earlier, the RI values for both household and expert decision-making are less than 0.1, suggesting that the decision-making is reliable. Hence, the outcome is dependable.

Step 3. Checking the consistency in the evaluation of the weights of the determinants. Since a large number of respondents are involved in the decision-making process, it is important to examine the deviation from consistency (DFC) in the decision-making process. The FullEX method does not offer any inherent consistency checking built into the model itself. Thus, we use the Full Consistency Method (FUCOM) [71] (Pamučar et al., 2018). We use the calculated weight values for the determinants (by the FullEX method) to formulate the MINLP further model using the FUCOM. The objective function of this model is DFC. We solve the models (for experts and households separately) to obtain the DFC values and recalculated weights. We find that the DFC value ≈ 0.0000 and the total absolute difference in the weights < 0.1 for both cases (experts and households).

Thus, it may be contended that the results obtained by applying the FullEX method are dependable, reliable, and valid.

4.1.2. Sensitivity Analysis

Sensitivity analysis in MCDA is crucial for assessing the robustness of decision outcomes. It helps analyze the impact of variations in input parameters on MCDA outcomes. It identifies the critical criteria that significantly influence outcomes. The sensitivity analysis helps decision-makers assess model stability, identify inconsistent assessments, and mitigate bias. Thus, sensitivity analysis assesses the model’s robustness [72,73] (Biswas et al., 2025, 2024). This paper examines the model’s sensitivity at two levels: household and expert.

First, we vary households’ responses. This is achieved through ten experimental cases. Under each case, we randomly remove the responses of five households. We examine its impact on the final ranking of the determinant (Figure 1).

Second, we vary experts’ opinions. We keep household respondents unchanged. We vary the values of the coefficients, α and β, respectively, one at a time. We examine whether these changes affect the outcome (Figure 2 and Figure 3)

The sensitivity analysis findings (Figure 1) demonstrate the stability and robustness of the model. Most of the determinants, including F1–F6 and F10, demonstrate consistent ranking across all cases. Nevertheless, several determinants, such as F8, F9, F12, F13, F17, F18, F19, F21, F23, and F24, exhibit variation under specific circumstances (e.g., deletions of household respondents). Overall, it is observed that deleting household respondents does not affect the extreme performers. The middle segment is affected to some extent. Figure 2 and Figure 3 show that changes in experts’ weights (variations in α values) and in the overall integration of households’ and experts’ opinions (variations in β values) visibly affect the final ranking. However, the extreme positions remain unchanged, reflecting the reliability and stability of the decision-making. We observe that the effect of changes in β values is more pronounced than that of α values. Overall, the model shows considerable stability.

4.2. Stage 2: Identification of the Causal Relationships Among Key Determinants (ISM-MICMAC)

The FullEX analysis identifies the top ten critical determinants of CHCE (

Table 7. Key determinants of CHCE.

S/L	Code	Determinants of Catastrophic Health Expenditure	Rank
1	F1	Low household income level	1
2	F2	Lack of job security	4
3	F3	Low education level	6
4	F4	Lack of assets and financial resilience	9
5	F5	Disparity in rural economic conditions	10
6	F6	Presence of dependent households	2
7	F10	Frequent birth rate and child health vulnerability	3
8	F14	Absence of prepayment or risk-pooling mechanisms	5
9	F15	Ineffective referral system in healthcare facilities	8
10	F20	Ineffective health governance and accountability measures	7

). We now proceed to discover the hierarchical relationships among these ten determinants using ISM-MICMAC analysis. To this end, first, experts decide the contextual relationships among the determinants through discussions and formulate the SSIM (

Table 8. SSIM.

Determinants	1	2	3	4	5	6	7	8	9	10
1		A	V	V	V	O	V	O	O	O
2			A	V	V	O	O	O	O	O
3				V	V	O	V	V	O	O
4					V	A	V	A	O	O
5						A	A	A	A	A
6							A	V	O	O
7								A	O	A
8									A	X
9										X
10

).

Based on the SSIM, the initial reachability matrix is constructed (

Table 9. Initial reachability matrix.

Determinants	1	2	3	4	5	6	7	8	9	10	Driving Power
1	1	0	1	1	1	0	1	0	0	0	5
2	1	1	0	1	1	0	0	0	0	0	4
3	0	1	1	1	1	0	1	1	0	0	6
4	0	0	0	1	1	0	1	0	0	0	3
5	0	0	0	0	1	0	0	0	0	0	1
6	0	0	0	1	1	1	0	1	0	0	4
7	0	0	0	0	1	1	1	0	0	0	3
8	0	0	0	1	1	0	1	1	0	1	5
9	0	0	0	0	1	0	0	1	1	1	4
10	0	0	0	0	1	0	1	1	1	1	5
Dependence Power	2	2	2	6	10	2	6	5	2	3

). Then, by applying the transitivity property, the final reachability matrix is obtained (

Table 10. Final reachability matrix.

Determinants	1	2	3	4	5	6	7	8	9	10	Driving Power
1	1	1*	1	1	1	1*	1	1*	1*	1*	10
2	1	1	1*	1	1	1*	1*	1*	1*	1*	10
3	1*	1	1	1	1	1*	1	1	1*	1*	10
4	0	0	0	1	1	1*	1	1*	1*	1*	7
5	0	0	0	0	1	0	0	0	0	0	1
6	0	0	0	1	1	1	1*	1	1*	1*	7
7	0	0	0	1*	1	1	1	1*	1*	1*	7
8	0	0	0	1	1	1*	1	1	1*	1	7
9	0	0	0	1*	1	1*	1*	1	1	1	7
10	0	0	0	1*	1	1*	1	1	1	1	7
Dependence Power	3	3	3	9	10	9	9	9	9	9

).

We observe that determinants F1, F2, and F3 hold the same driving power, while F5 has the highest dependence. It is observed that several determinants share the same dependence or driving power.

Based on the final RM, we proceed to partition the determinants into various levels. This is achieved through three iterations. The result of the level partitioning is reported in

Table 11. Final level partitioning and allocation of the determinants.

Determinants	Reachability Set	Antecedent Set	Intersection Set	Level
1	1, 2, 3,	1, 2, 3,	1, 2, 3,	3
2	1, 2, 3,	1, 2, 3,	1, 2, 3,	3
3	1, 2, 3,	1, 2, 3,	1, 2, 3,	3
4	4, 6, 7, 8, 9, 10,	1, 2, 3, 4, 6, 7, 8, 9, 10,	4, 6, 7, 8, 9, 10,	2
5	5,	1, 2, 3, 4, 5, 6, 7, 8, 9, 10,	5,	1
6	4, 6, 7, 8, 9, 10,	1, 2, 3, 4, 6, 7, 8, 9, 10,	4, 6, 7, 8, 9, 10,	2
7	4, 6, 7, 8, 9, 10,	1, 2, 3, 4, 6, 7, 8, 9, 10,	4, 6, 7, 8, 9, 10,	2
8	4, 6, 7, 8, 9, 10,	1, 2, 3, 4, 6, 7, 8, 9, 10,	4, 6, 7, 8, 9, 10,	2
9	4, 6, 7, 8, 9, 10,	1, 2, 3, 4, 6, 7, 8, 9, 10,	4, 6, 7, 8, 9, 10,	2
10	4, 6, 7, 8, 9, 10,	1, 2, 3, 4, 6, 7, 8, 9, 10,	4, 6, 7, 8, 9, 10,	2

.

Table 11 shows that F5 is the ultimate dependent variable, and F1, F2, and F3 are independent determinants. The result of level partitioning aligns with the final RM. The other variables (i.e., determinants) are classified as the linkage variables. Hence, the ISM model consists of three levels. Removing the transitivity or indirect relations, we obtain the final ISM model (Digraph) shown in Figure 4.

To further validate the final ISM model, we proceed to conduct the MICMAC analysis. The result of the MICMAC analysis is shown in Figure 5. The MICMAC analysis also reflects that F1, F2, and F3 have the same driving power. The linkage variables also remain in the same place. F5 remains the standalone dependent variable. Hence, it confirms the level partitioning result.

5. Discussion

The MCDA framework identifies critical factors that drive CHCE. The principal variables identified are poor household income, dependent households, and elevated birth rates associated with child health concerns. It unveils the significant impact of socioeconomic and demographic vulnerabilities on health expenditure shocks. It also highlights the absence of employment security and prepayment or risk-pooling mechanisms, which intensify people’s susceptibility to catastrophic OOP expenses.

The ISM analysis illustrates the hierarchical linkages among factors, emphasizing F5 (disparity in rural economic situations) as a manifested driver, contributed by several intermediate and fundamental determinants. This demonstrates how macro-level disparities result in familial vulnerabilities and systemic inefficiencies. The interconnections among F6, F10, F14, and F20 indicate reinforcing feedback loops related to demographic composition, healthcare system inadequacies, and governance issues, suggesting that actions focused on a single factor may have extensive repercussions across multiple domains.

The MICMAC analysis examines factors within a driving-dependency matrix, identifying low household income, job security, and education level as independent variables that significantly affect CHCE risk with minimal dependence. Conversely, healthcare governance, risk-pooling mechanisms, and referral systems are interrelated factors with considerable interdependence and impact. Focused improvements in these interconnected variables can expedite initiatives to reduce catastrophic healthcare costs.

Overall, the findings necessitate comprehensive policy measures to mitigate structural socioeconomic challenges and systemic healthcare inefficiencies, and to alleviate catastrophic expenditures, particularly for marginalized populations. A comprehensive strategy to address these issues is essential to maintaining enduring financial protection in healthcare.

The FullEx method offers several advantages. The FullEX method is an objective weighting approach to criteria prioritization. However, it is built on expert evaluations. Thus, it offers the advantages of expert evaluation grounded in experience and adaptability to diverse MCDA challenges. It is a straightforward process that promotes transparency and accountability [61,74] (Bošković et al., 2023, 2024). However, the advantage of the FullEx method is sometimes undermined by a lack of expertise, improper expert selection, and overreliance on biased responses.

5.1. Research Implications

The research findings provide several actionable insights across managerial, social, policy, and technical perspectives.

• The MCDA findings assist healthcare administrators in prioritizing resource allocation and interventions by identifying critical characteristics such as low household income and elevated birth rates that impact child health. Managers need to concentrate on specific financial and wellness initiatives for at-risk populations. Furthermore, by comprehending linking characteristics such as insufficient governance, hospital administrators and insurers can formulate policies that enhance service delivery. This necessitates the inclusion of socioeconomic factors in health spending models to improve forecasts and deliver customized treatments.
• The study highlights the need to control inequities and social vulnerabilities that render households susceptible to financial shocks from healthcare expenses. It recognizes insufficient education and rural economic disparities as critical determinants for community-oriented programs to improve health literacy and economic empowerment for at-risk communities. The interconnections among these factors indicate that social policies should employ comprehensive, multi-sectoral strategies that address education, employment, and healthcare access. The findings support cross-sector collaborations among healthcare providers, local governments, and social organizations to create social safety nets and empower disadvantaged populations against catastrophic costs.
• The findings provide insights for public health policy formulation by classifying determinants into driving, connecting, and dependent elements, facilitating strategic interventions. It urges governments to expand risk pooling, enhance transparency in governance, and optimize referral networks to assist disadvantaged households. The results align with global standards that promote integrated systems for social protection, rural development, and healthcare reform.
• The study emphasizes the scientific usefulness of MCDA-based approaches in healthcare decision-making, offering a framework relevant to diverse policy matters. Using ISM-MICMAC analysis, it assesses interdependencies among variables and enhances decision transparency. The MCDA methodology enhances flexible modeling, hence advancing health economics and policy approaches. This comprehensive framework helps mitigate exorbitant healthcare costs, enabling systematic, sustainable actions across the healthcare system [75,76] (Marsh et al., 2017; Gongora-Salazar et al., 2023)

The current study focuses on India as a case study. India is a vast country, with a substantial portion of its population residing in rural areas. According to a recent government report [77] (PIB, 2021), about 65 percent of its population lives in rural areas. Although per capita income shows promising growth, enabling phenomenal economic growth, it is worth studying the BoP in rural areas. Thus, the current study stands as a significant contribution to the literature. We have also noticed that some studies [78] (e.g., Quintal, 2019) have also worked on developed countries.

5.2. Future Scopes

However, the current study offers several scopes for further extensions, outlined below.

• The success of the FullEX method largely depends on experts’ evaluations and context. Future studies may conduct a large-scale empirical analysis (through exploratory and confirmatory factor analysis) to validate the findings.
• The current work was focused on the rural segment. However, future studies may consider regional comparisons to examine the common factors and heterogeneous requirements.
• Future studies may be designed to examine the intermediate role of NGOs in implementing appropriate measures to reduce CHCE. A mediation model can be built and tested. In this regard, the pivotal role of technology in mitigating CHCE can also be examined.
• Future studies may investigate the efficacy of the government schemes and initiatives in removing the OOP financial burden of the people. In this context, the usefulness of the PPP models may be examined.
• Cross-cultural applications and comparative analysis may yield significant insights into the adaptability of MCDA approaches. Furthermore, integrating MCDA findings with qualitative and political analyses will provide a solid basis for policy and administrative decisions.
• The FullEX model may be further extended using various uncertainty measures (using fuzzy and rough numbers) to offset the subjective bias and improve its robustness.
• The current study involves a limited number of experts having 20+ years of experience. Although several studies on MCDM applications (e.g., [79] Guan et al., 2024) have considered 5 years of experience as a benchmark, we recognize this as a limitation of our work. Nevertheless, the current work draws on the opinions of a large group of residents and experts. Thus, this limitation does not undermine the usefulness and reliability of the current work. Future studies can validate the findings of this work by considering more experts having 20–25 years of experience.

6. Conclusions

The critical factors influencing the CHCE have been identified in the current study. A two-stage approach has been designed to evaluate the determinants and establish hierarchical relationships among the critical variables. A focus group of 30 experts and 250 household respondents participated in the research. The opinions of the respondents and experts have been synthesized to identify ten critical determinants using the FullEX method. Then, an ISM-MICMAC analysis was performed to investigate the structural relationships among the determinants. In conclusion, the study emphasizes the need for comprehensive policy interventions to address socioeconomic challenges and structural inefficiencies, while advocating greater methodological standardization, stakeholder involvement, and qualitative insights. The use of the FullEx method, integrated with the ISM-MICMAC analysis, fosters a scientific and comprehensive examination of the determinants of CHCE to formulate appropriate action plans.