Is publicly funded health insurance associated with healthcare utilisation and financial protection in India? Evidence from a cross-sectional household survey (CAMS 2022–2023)

Abstract

Objectives

To examine the association of India’s publicly funded health insurance (PFHI), mainly Ayushman Bharat–Pradhan Mantri Jan Arogya Yojana, with healthcare utilisation, out-of-pocket expenditure (OOPE) and catastrophic health expenditure (CHE).

Design

Cross-sectional study using nationally representative data from the Comprehensive Annual Modular Survey 2022–2023. Multivariable logistic regression and Heckman two-stage models were applied, adjusting for socioeconomic and demographic covariates and survey design.

Setting

All states and union territories of India.

Participants

302 086 households (173 096 rural; 128 990 urban) surveyed between July 2022 and June 2023.

Results

PFHI coverage was associated with higher outpatient (adjusted OR 1.404, 95% CI 1.328 to 1.484, p<0.01) and inpatient care use (adjusted OR 1.168, 95% CI 1.128 to 1.210, p<0.01). At the 20% threshold, PFHI coverage significantly reduced the likelihood of CHE for both outpatient (adjusted OR 0.869, 95% CI 0.810 to 0.932, p<0.01) and inpatient care (adjusted OR 0.790, 95% CI 0.737 to 0.846, p<0.01). Heckman model estimates showed a significant reduction in inpatient OOPE (coefficient –0.205, 95% CI −0.258 to −0.153, p<0.01), while outpatient OOPE reductions were modest (–0.047, 95% CI −0.087 to −0.006, p<0.05). However, outpatient care remained inadequately covered, with high OOPE and CHE persisting even among insured households. Financial vulnerability was higher among the uninsured, socioeconomically disadvantaged groups and rural residents.

Conclusions

PFHI was associated with improved access to inpatient services and partial financial protection, but association with outpatient care indicates a major source of financial burden. Policy reforms should expand PFHI benefits to outpatient services, strengthen primary care through Ayushman Arogya Mandirs and tailor implementation to state contexts and vulnerable populations to achieve equitable and comprehensive financial protection.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The study uses nationally representative data from the Comprehensive Annual Modular Survey 2022–2023, covering more than 300 000 households, which enhances the generalisability of findings to the Indian population.

• Household-level insurance coverage and expenditure data are well aligned with the design of India’s publicly funded health insurance schemes, which extend benefits to all household members once enrolled.

• The analysis employs multivariable logistic regression and Heckman two-stage models, accounting for survey design, clustering and potential selection bias in estimating out-of-pocket expenditures.

• The survey does not collect data on disease-specific morbidity or type of healthcare provider (public vs private), limiting disaggregation of effects and adjustment for health needs.

• As a cross-sectional survey relying on self-reported expenditure, the study is subject to recall bias and cannot establish causal relationships.

Background

India’s health insurance landscape has undergone substantial reforms in recent years, primarily driven by the introduction of the publicly funded health insurance (PFHI) scheme, namely Ayushman Bharat–Pradhan Mantri Jan Arogya Yojana (AB-PMJAY) in 2018, alongside complementary state-level initiatives.^{1 2} AB-PMJAY, one of the world’s largest PFHI programmes, replaced Rashtriya Swasthya Bima Yojana (RSBY), which was launched in 2008 with a limited secondary care coverage of 30 000 Indian rupees (INR) ($350) per family per year.²

PFHI schemes in India, including AB-PMJAY and most legacy state schemes, are household based by design. Eligible households receive a single scheme card that entitles all members to cashless treatment for a defined package of inpatient services across empanelled public and private hospitals. Hence, coverage is not individual but rather automatically extended to all household members enrolled in the scheme.² AB-PMJAY substantially increased the annual health coverage to INR500 000 (approximately $6000) per family per year for over 120 million poor households (approximately 500 million individuals) who are automatically eligible, with no enrolment fees for secondary and tertiary-level hospitalisation across public and empanelled private hospitals nationwide.²,⁸

India’s National Health Policy (2017) emphasises financial risk protection for achieving universal health coverage.⁹ Health insurance plays a vital role in shielding households from the economic burden of healthcare expenses.¹⁰,¹⁴ Evidence on whether PFHI improves healthcare utilisation is generally favourable across the globe.¹⁵ Many studies in India reported increased healthcare use among insured individuals, while a few found no significant association.^{57 16},²¹

Several studies have documented the incidence of catastrophic health expenditure (CHE) across diseases, demographic groups and regions in India.²²,³⁰ Reddy³⁰ reported substantial variation in CHE, with incidence ranging from 5.1% to 69.9%. Gaddam and Rao identified a peak CHE incidence of 23.45% across India at a 10% threshold in 2014, in line with the findings of Sriram and Albadrani and Pandey et al, followed by a decline to 16.7% in 2018–2019.^{23 25 31} Higher CHE rates have been reported for chronic conditions, injuries and maternal health. Karan et al²⁴ and Dhankhar et al²² found CHE rates exceeding 45% for non-communicable diseases and multimorbidity cases. Similarly, injuries and accidents, especially those treated in private hospitals, were associated with higher CHE.^{26 29} Institutional and caesarean deliveries were linked to distress financing.²⁷ Sriram et al²⁸ identified wealth-related inequalities in CHE for inpatient care.

Studies on the role of PFHI in reducing CHE have produced mixed findings.¹⁸ Many studies have found RSBY ineffective in reducing the financial burden on poor households.²⁴,^{6 8 32} Garg et al³³ observed no reduction in CHE among elderly beneficiaries of PFHI. Similarly, Garg et al reported no reductions in CHE or out-of-pocket expenditure (OOPE) following AB-PMJAY enrolment in Chhattisgarh, with CHE exceeding 70% at a 10% threshold for individuals using private hospitals.^{16 34} Garg et al³⁵ found that PFHI did not reduce OOPE or distress financing for caesarean and non-caesarean institutional deliveries across India. However, Parmar et al¹⁹ found that AB-PMJAY was associated with reduced CHE in a survey of 57 000 eligible individuals across six Indian states. Sriram and Albadrani found lower CHE rates among insured individuals, while Harish et al observed reduced CHE among insured patients in public hospitals in Kerala.^{31 36} Aashima and Sharma noted significant reductions in CHE among the insured in 21 out of 32 states studied.³⁷

All the recent nationally representative cited studies are based on dated data, much of which were collected before or shortly after the launch of AB-PMJAY, with the majority drawing on the National Sample Survey (NSS) 75th round (2017–2018) or Fifth National Family Health Survey (2019–2021).2324 28 29 37,⁴⁴ Although the NSS 75th round (2017–2018) provided richer and more granular information on morbidity, treatment-seeking patterns and provider type, its timing coincided with the nascent stages of AB-PMJAY implementation, limiting its utility for assessing the scheme’s medium-term effects.

In contrast, this study uses the NSS 79th round (Comprehensive Annual Modular Survey (CAMS) 2022–2023), which was conducted 4 years after AB-PMJAY’s roll-out, making it far more suitable for evaluating the scheme’s effectiveness at scale. CAMS employed a much larger nationally representative sample of over 302 000 households, thereby improving statistical power and representativeness.⁴⁵ While it sacrifices some granularity—such as morbidity profiles and provider-type details—and uses a slightly longer recall period for outpatient care (30 days vs 15 days), its value lies in timing and scope. By capturing household-level health expenditure and insurance coverage after 4 years of PFHI operation, CAMS provides timely evidence to assess the functioning of India’s largest PFHI programme, with direct implications for fine-tuning design and implementation to protect more than 500 million individuals in the lower half of India’s socioeconomic distribution. This approach aligns with PFHI policy design, where once a household is enrolled, all members are covered. AB-PMJAY and state-run PFHI schemes do not require individual enrolment and provide coverage to the entire eligible household.²

Research questions

The study addresses the following questions: (1) Are households in India covered under PFHI schemes more likely to use outpatient and inpatient care services than uninsured households? (2) Is PFHI coverage associated with reduced OOPE for healthcare, particularly inpatient services in India? (3) Does PFHI coverage lower the incidence of CHE, and if so, to what extent across different thresholds and population subgroups in India?

Methods

Data

This study draws on the CAMS 2022–2023 (NSS 79th round) conducted by the National Statistical Office.⁴⁶ Using a stratified multistage design, CAMS 2022–2023 sampled 302 086 households (173 096 in rural areas and 128 990 in urban areas).⁴⁷ ((Dataset) CAMS 2022–2023 data and related resources can be downloaded from the following website: https://microdata.gov.in/NADA/index.php/catalog/220/related-materials.) The survey elicited information at the household level on whether any household member was covered under one or more health insurance schemes. Data collected included whether any household member sought ‘outpatient care’ (that consists of all non-hospitalised care, including preventive measures and self-care) within the past 30 days of the survey and ‘inpatient care’ (including hospitalisation and institutional deliveries) within the past 365 days of the survey. Details were also obtained regarding the OOPE, amount reimbursed by medical insurance providers or employers for outpatient and inpatient care.

Design

This is a cross-sectional, observational study based on secondary analysis of anonymised, nationally representative household data from CAMS 2022–2023. The study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cross-sectional studies.

CAMS 2022–2023, conducted between July 2022 and June 2023, captures health insurance-related data after 4 years of AB-PMJAY implementation, making it valuable for evaluating the medium-term impacts of PFHI schemes in India. Although less granular, CAMS 2022–2023 maintains national representativeness, allowing for policy-relevant macro-level insights.

To ensure comparability and reproducibility, we applied standard methods recommended for health expenditure analysis. Household consumption expenditure was normalised using per capita equivalence scales that adjust for household size and composition following Wagstaff and van Doorslaer.¹³ Missing or inconsistent expenditure values were handled using mean imputation within state–residence strata, which preserves distributional characteristics while minimising bias. (Item non-response on core expenditure variables was negligible, attributable to the Computer-Assisted Personal Interviewing-enabled data collection of the 79th round which incorporated real-time validation and consistency checks. Household consumption expenditure, used as the denominator for CHE calculations, had no missing values. Insurance reimbursement values were set to zero for households that used care but had no insurance coverage and therefore received no reimbursement, reflecting the substantively correct value. Given the very low rate of item non-response, the choice of imputation method has negligible impact on estimates) ^{48 49} Outpatient expenditures reported for a 30-day recall period were annualised, while inpatient expenditures collected over a 365-day recall period were retained, in line with international conventions.^{11 50} These adjustments allow CHE to be consistently calculated as the share of annualised household OOPE relative to total household consumption thresholds using various thresholds.¹¹,¹⁴

Outcome variables

Schemes in India cascade through several steps, including enrolment, awareness about enrolment and associated benefits, access to and selection of participating healthcare providers, and use of services.^{51 52} Within this broad framework, this study seeks to assess the association of PFHI with the following outcome variables:

1. Healthcare use (binary: 1=yes, 0=no): (a) use of outpatient care in the last 30 days before the survey by any member of the household; and (b) use of inpatient care in the last 365 days before the survey by any member of the household to evaluate whether PFHI increases the likelihood of seeking care.

2. CHE (binary: 1=yes, 0=no): whether households incur CHE while seeking healthcare to evaluate the association of PFHI with CHE separately for outpatient and inpatient care. CHE is defined as health expenditure exceeding a specified threshold of household consumption expenditure. Multiple definitions of CHE exist in the literature. Xu et al⁵³ proposed a ‘capacity-to-pay’ threshold of 40% of non-subsistence expenditure, while others recommend using a proportion of total household consumption, commonly 10%, to enable international comparability.¹¹,¹³ In the Indian context, several studies using NSS and National Family Health Survey datasets have operationalised CHE at 10% of household consumption expenditure (eg, Gaddam and Rao²³; Sriram and Albadrani³¹), making these thresholds widely recognised benchmarks. Accordingly, this study adopted 10% as the standard global and Indian cut-off and 20% as a stricter threshold for robustness checks, thereby distinguishing moderate from severe financial shocks while ensuring comparability across both global and national research.

• Although PFHI schemes in India do not directly cover outpatient services, their potential indirect influence on outpatient utilisation can be situated within Andersen’s behavioural model of health service use,⁵⁴ which emphasises the interaction of predisposing, enabling and need factors in shaping care-seeking behaviour. Insurance coverage functions as an enabling resource, reducing perceived financial barriers and increasing households’ propensity to seek care even for services not explicitly covered. PFHI enrolment may strengthen awareness of entitlements, foster trust in formal health systems and enhance contact with health facilities through referrals and continuity of care—all of which increase the likelihood of outpatient use. We examined outpatient care utilisation to provide a complete picture of healthcare access under universal health coverage goals and investigate any indirect influence of PFHI on outpatient care, such as increased health awareness, referrals or willingness to seek primary care due to overall engagement with the health system.⁶ Outpatient expenditure was annualised for comparability and to estimate total household OOPE over a year. CHE incidence was calculated as the proportion of households whose annual OOPE exceeded thresholds of total consumption expenditure of more than 10% or 20%.

• 3 aOOPE (continuous variable, INR spent): expenditure incurred by households while seeking outpatient or inpatient care to study the association of PFHI with OOPE.

Although CAMS does not include information on disease type or diagnosis, which is a limitation of CAMS 2022–2023, this does not compromise the analysis of PFHI financial protection, as all inpatient care up to a ceiling per household, irrespective of disease type, is covered. The eligibility of coverage is not disease based but instead determined by the nature of care (ie, hospitalisation) and empanelment of the provider. PFHI schemes cover a wide range of hospital-based procedures irrespective of disease classification. Therefore, inpatient OOPE and CHE outcomes remain valid indicators of scheme effectiveness, regardless of diagnosis data.

Independent variables

PFHI coverage is the key independent variable, which is categorical: 0=uninsured (100% self-pay); 1=PFHI schemes, including AB-PMJAY and those funded by state governments (typically no copayments, target low income and cover all members of the household); 2=Employees’ State Insurance Corporation (ESIC)/Central Government Health Scheme (CGHS) and other employer-provided insurance; 3=private insurance.

Covariates

The other independent variables capture socioeconomic, demographic and regional disparities based on previous studies and data availability.^{2 17 18 23 42} Rural/urban residence indicates whether a household is rural or urban. A state or union territory (UT) is a categorical variable that accounts for regional variations in policy implementation, economic development and health infrastructure. Social groups include scheduled tribes, scheduled castes, other backward classes and others. Religion is a categorical variable, including Hindu, Muslim, Christian and Others. Household size has three categories: small households (one to two members), medium households (three to five members) and large households (more than 5 members). Households are divided into expenditure deciles based on the household’s monthly per capita expenditure, a proxy for income. The education of the household head is categorised as illiterate or less than primary, primary, middle, secondary/senior secondary, and graduate and above.

The CAMS 2022–2023 survey does not capture whether care was sought in public or private hospitals. While this is a limitation, our analysis accounts for underlying variation in healthcare infrastructure and service delivery environments through the inclusion of state fixed effects in all multivariable models. This approach helps control for unobserved, time-invariant differences across states—such as the relative reliance on public versus private providers—thereby enhancing the validity of estimated associations between PFHI coverage and health outcomes. This is discussed further in the Limitations and data gaps section.

Statistical analysis

Descriptive statistics are used to profile sampled households, PFHI coverage and outpatient and inpatient care use. Cross-tabulations present the association between independent variables and CHE and OOPE. All analyses were conducted using Stata ‘survey’ (svy) commands to account for the complex survey design of the CAMS 2022–2023 data, including stratification, clustering and unequal sampling probabilities. The survey design variables—strata, primary sampling units and population weights—were specified before estimation.⁵⁵ This approach ensures that SEs and CIs are appropriately adjusted for the multistage sampling design and enables valid population-level inferences.⁵⁵ Statistical significance is assessed at conventional levels (p<0.05, p<0.01).

Logistic regression models attempt to quantify the relationship between PFHI and the use of healthcare and CHE after controlling for various covariates in the model. Logistic regression was used for the following binary outcome variables: (a) outpatient care use in the past 30 days, (b) inpatient care use in the past 365 days, (c) CHE during outpatient care and (d) CHE during inpatient care. In the logistic regression, the outcome (Y₁) is modelled as a binary outcome (‘1’ if healthcare is used or ‘0’ otherwise, or ‘1’ if the household undergoes CHE or ‘0’ otherwise). The logistic regression equation is expressed as:

$${\displaystyle {\displaystyle P(Y_1=1)=\frac{\exp ({\beta }_0+{\beta }_1{X}_1+{\beta }_2{X}_2+{\beta }_3{X}_3+{\beta }_4{X}_4+{\beta }_5{X}_5+{\beta }_6{X}_6+{\beta }_7{X}_7+{\beta }_8{X}_8)}{[1+\exp ({\beta }_0+{\beta }_1{X}_1+{\beta }_2{X}_2+{\beta }_3{X}_3+{\beta }_4{X}_4+{\beta }_5{X}_5+{\beta }_6{X}_6+{\beta }_7{X}_7+{\beta }_8{X}_8)]}}}$$

where:

• P(Y₁=1) represents the probability of healthcare use or CHE.

• β₀ is the intercept, and β₁ to β₈ are coefficients for the covariates. The covariates include: X₁—health insurance status (categorical); X₂—residence: urban/rural; X₃—state/UT (categorical); X₄—religion (categorical); X₅—social group (categorical); X₆—household size (categorical); X₇—education of household head (categorical); and X₈—expenditure deciles (categorical).

Logistic regression output reports adjusted ORs instead of coefficients explained by the equation.

$${\displaystyle {\displaystyle \text{OR}=e^{\left({\beta }_0+{\beta }_1{X}_1+{\beta }_2{X}_2+\cdots +{\beta }_k{X}_k\right)}}}$$

To model household OOPE conditional on healthcare utilisation, we estimated a naïve ordinary least squares regression of OOPE and a log of OOPE. OOPE data are log transformed to handle skewed distributions observed in expenditure data.^{10 50} A Heckman two-stage selection model addresses sample selection bias because OOPE is only observed for individuals who accessed outpatient or inpatient services.^{10 50} The two-step process includes:

The selection equation (probit model), which determines whether an individual incurs healthcare expenditure (binary outcome):

S*=Xγ+u

s=1 if S*>0 (households use health service either outpatient or hospitalised)

S=0 otherwise

where:

• S represents the selection decision (whether the household uses health services).

• X includes the covariates: ownership of television (TV), access to road and public transport and broadband availability

• γ represents the coefficients, and u is the error term.

The outcome equation (expenditure model), which models the expenditure amount, conditional on incurring expenditure (S=1):

$${\displaystyle {\displaystyle Y=X\beta +\rho \lambda +\epsilon }}$$

where:

• Y represents healthcare expenditure (continuous outcome).

• X includes the covariates defined above.

• β represents the coefficients.

• λ is the inverse Mills ratio, derived from the selection equation to adjust for selection bias.

• ρ represents the coefficient of λ, capturing the correlation between errors in the selection and outcome equations.

• ε is the error term.

In the Heckman two-stage model, TV ownership, broadband access and road/transport availability were used as exclusion restrictions because they capture household-level awareness, connectivity and mobility that influence the likelihood of seeking care but not the level of OOPE once care is accessed. For example, TV and broadband exposure increase awareness of illness and entitlements. At the same time, transport access reduces mobility barriers, yet it neither directly affects medical costs, which are driven by provider type, disease severity and insurance coverage. This logic follows Andersen’s behavioural model, and similar instruments have been used in low- and middle-income country settings to predict utilisation without biasing cost equations. Their use thus provides theoretically sound and empirically supported validity for the Heckman specification.^{54 56 57} The second stage (outcome equation) models the log of OOPE for those who used services, controlling for the inverse Mills ratio derived from the first stage. The significance of the selection term (lambda) will indicate the presence of non-random selection and justify using this correction.

We applied a log(x+1) transformation to OOPE values to accommodate the substantial skewness in expenditure data and retain households reporting zero spending. In this specification, regression coefficients can be interpreted as semielasticities: for households incurring positive expenditure, a one-unit change in the independent variable is associated with an approximate percentage change in OOPE. For example, a coefficient of –0.20 implies roughly a 20% lower OOPE relative to the reference group. This approach is widely recommended in health expenditure studies to stabilise variance, reduce the influence of extreme values and provide interpretable percentage effects while preserving zero observations.⁵⁰

Model fit was evaluated using Akaike’s information criterion and Bayesian information criterion. Variance inflation factor was computed for all covariates included in the models; all variance inflation factor (VIF) values were below 2.5, indicating no serious collinearity issues. Tests were employed to check for model specification errors. These diagnostics provide confidence in the model structure’s robustness and the validity of the inferences.

This cross-sectional study, based on self-reported expenditure data, is constrained by the absence of morbidity and clinical information, provider type and indirect costs, with insurance measured at the household level. These are discussed further in the Discussion section. All estimated coefficients should be interpreted as associations rather than causal effects, given the cross-sectional nature of the data.

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research. The analysis is based entirely on anonymised, publicly available secondary survey data.

Results

Descriptive and bivariate analysis

60% of households had health insurance coverage. 49.7% reported coverage under different PFHI schemes. 8.2% were covered under ESIC/CGHS/employer-provided schemes (table 1). A small proportion (2.0%) reported having private insurance. Reported PFHI coverage reflects household-level enrolment, consistent with PFHI policy that automatically covers all members once the household is enrolled.

Table 1. Sample and PFHI distribution and incidence of outpatient and inpatient utilisation^*.

Variable	n (weighted)	Insurance coverage (%)	Outpatient care use (%)	Inpatient care use (%)
All India	302 086	59.9	32.4	13.1
Insurance status
No insurance	121 060	40.1	29.7***	12.4***
PFHI	150 243	49.7	35.1	13.7
ESIC/CGHS/employer proposed/others	24 680	8.2	29.0	12.5
Private	6103	2.0	33.6	13.3

Significance levels: ***p<0.01, **p<0.05, *p<0.1.

^*Refer to online supplemental table A1 for a detailed table of all the covariates: residence, household size, religion, social group, expenditure decile, education of the head of household and state/UT of residence.

CGHS, Central Government Health Scheme; ESIC, Employees’ State Insurance Corporation; PFHI, publicly funded health insurance; UT, union territory.

At the national level, 32.4% of households sought outpatient care 30 days before the survey (table 1). Households with PFHI reported the highest utilisation of outpatient care (35.1%) compared with those without insurance (29.7%) (p<0.01). At the national level, 13.1% of households reported inpatient care in the 365 days preceding the survey (table 1). Households with PFHI reported higher hospitalisation rates (13.7%) than those without insurance (12.4%) (p<0.01) (online supplemental figure A1).

Online supplemental table A1 gives the sample distribution by urban/rural, state/UT residence, social, demographic and economic background. Significant variations in PFHI coverage by rural/urban, household size, religion, social group, expenditure deciles and state/UT were found. The bubble chart (online supplemental figure A2) illustrates the state-level variation in outpatient and inpatient healthcare use, with bubble size representing the extent of PFHI coverage.

The average outpatient health expenditure in the last 30 days among households that accessed outpatient care was INR1732 nationally, with AB-PMJAY or state scheme households spending relatively less (INR1640), which is statistically significant (table 2). Households without insurance incurred the highest average inpatient spending (INR37 572), while those covered under AB-PMJAY or state health insurance schemes incurred comparatively lower costs (INR29 875), suggesting some financial protection, which is statistically significant. The burden was particularly high among privately insured households (INR56 596) and those covered by employer-based schemes (INR44 557), likely reflecting access to costlier private facilities (table 2).

Table 2. Incidence of CHE (%) and amount of OOPE (INR)^*.

Variables	Monthly household expenditure (INR)	Outpatient health expenditure (95% CI) in the last 30 days in households with outpatient care (INR)	Inpatient health expenditure (95% CI) in the last 365 days in households with inpatient care (INR)	Outpatient CHE (10%) in households with outpatient care (%)	Outpatient CHE (20%) in households with outpatient care (%)	Inpatient CHE (10%) in households with inpatient care (%)	Inpatient CHE (20%) in households with inpatient care (%)
All India	12 660	1732	33 928	38.2	19.4	46.3	27.8
Insurance status
No insurance	11 373	1755 (1676 to1833)	37 572 (35 893 to 39 250)	38.1***	20.7***	54.2***	32.7***
AB-PMJAY or state health insurance	12 154	1640 (1591 to 1688)	29 875 (28 608 to 31 141)	39.2	19.3	41.3	24.9
ESIC/CGHS/employer/others	19 040	2058 (1904 to 2211)	44 557 (39 511 to 49 602)	32.7	15.4	42.0	24.3
Private insurance	24 827	2542 (2302 to 2779)	56 596 (42 952 to 70 238)	34.1	14.7	42.6	23.7

Significance levels: ***p<0.01, **p<0.05, *p<0.1.

^*Refer to online supplemental table A2 for a detailed table of all the covariates: residence, household size, religion, social group, expenditure decile, education of the head of household and state/UT of residence.

AB-PMJAY, Ayushman Bharat–Pradhan Mantri Jan Arogya Yojana ; CGHS, Central Government Health Scheme; CHE, catastrophic health expenditure; ESIC, Employees’ State Insurance Corporation; INR, Indian rupee; OOPE, out-of-pocket expenditure; UT, union territory.

38.2% of such households face outpatient CHE at the 10% threshold and 19.4% at the 20% threshold. The incidence was high among uninsured households, 38.1% and 20.7%, respectively (p<0.01). In contrast, households covered under AB-PMJAY or state government insurance schemes reported a higher rate for CHE at 10% and a lower rate at 20% compared with uninsured (39.2% at 10% and 19.3% at 20%) (p<0.01). The lowest CHE incidence was observed among households with employer-based or CGHS/ESIC coverage (32.7% and 15.4%) and private insurance (34.1% and 14.7%) (table 2) (p<0.01).

Inpatient CHE was 46.3% of households at the 10% threshold and 27.8% exceeding the 20% threshold. The financial burden was most severe among uninsured households, where 54.2% and 32.7% crossed the 10% and 20% CHE thresholds (p<0.01). Households covered under AB-PMJAY or state health insurance schemes fared better, with 41.3% facing CHE at 10% and 24.9% at 20% (p<0.01). Employer-based and government schemes, such as ESIC or CGHS, showed an incidence of 42.0% at 10% and 24.3% at 20%, while private insurance holders also reported reduced CHE rates of 42.6% at 10% and 23.7% at 20%.

Online supplemental table A2 describes the distribution of CHE and OOPE for other covariates. The average monthly household consumption expenditure ranged from INR6003 among the poorest decile to INR21 768 among the richest, while outpatient expenditure in the last 30 days varied from INR1225 (poorest decile) to INR2583 (richest decile).

Multivariable analysis

Results of five multivariable models are presented.

Use of outpatient and inpatient care (model 1 and model 2)

Compared with individuals without any insurance, those covered under AB-PMJAY or state health insurance schemes had significantly higher odds of using both outpatient care (adjusted OR 1.404, 95% CI 1.328 to 1.484, p<0.01) and inpatient care (adjusted OR 1.168, 95% CI 1.128 to 1.210, p<0.01). Private insurance was also associated with significantly higher odds of using outpatient (adjusted OR 1.217, 95% CI 1.081 to 1.370, p<0.01) and inpatient services (adjusted OR 1.110, 95% CI 1.002 to 1.229, p<0.05). Individuals covered by employer-based or government insurance schemes (eg, ESIC/CGHS) showed modest but statistically significant increases in utilisation for both outpatient (adjusted OR 1.090, 95% CI 0.994 to 1.195, p<0.1) and inpatient care (adjusted OR 1.098, 95% CI 1.025 to 1.176, p<0.01). These findings indicate that health insurance, particularly publicly funded programmes, enables access to outpatient and inpatient services (table 3).

Table 3. Logistic regression of outpatient and inpatient care utilisation^*.

	Model 1				Model 2
Variables	OR of outpatient utilisation				OR of inpatient utilisation
	Unadjusted OR	95% CI	Adjusted OR	95% CI	Unadjusted OR	95% CI	Adjusted OR	95% CI
Insurance coverage (none)
AB-PMJAY or state health insurance	1.283***	1.221 to 1.348	1.404***	1.328 to 1.484	1.129***	1.092 to 1.168	1.168***	1.128 to 1.210
ESIC/CGHS/employer/others	0.967	0.884 to 1.058	1.090*	0.994 to 1.195	1.016	0.943 to 1.095	1.098***	1.025 to 1.176
Private insurance	1.199***	1.071 to 1.343	1.217***	1.081 to 1.370	1.088*	0.983 to 1.205	1.110**	1.002 to 1.229

Significance levels: ***p<0.01, **p<0.05, *p<0.1. The p<0.1 threshold is reported to indicate marginal significance, following standard econometric practice in large-sample survey analyses.

^*Refer to online supplemental table A3 for a detailed table of all the covariates: residence, household size, religion, social group, expenditure decile, education of the head of household and state/UT of residence.

AB-PMJAY, Ayushman Bharat–Pradhan Mantri Jan Arogya Yojana; CGHS, Central Government Health Scheme; ESIC, Employees’ State Insurance Corporation; UT, union territory.

Online supplemental table A3 provides the results of logistic regression of the full model. Larger households were significantly more likely to use both outpatient and inpatient services. Religious minorities, particularly those classified as ‘Others’, showed higher odds of using both types of care. Scheduled tribes had consistently lower odds of utilisation across both models.

CHE during outpatient care (model 3)

At the 10% threshold, ESIC/CGHS/employer-based insurance was associated with a significantly lower likelihood of CHE (adjusted OR 0.879, 95% CI 0.789 to 0.983, p<0.05) (table 4). In contrast, AB-PMJAY or state insurance showed no significant association (adjusted OR 1.048). At the 20% threshold, CHE was statistically lower in AB-PMJAY or state insurance (adjusted OR 0.869, 95% CI 0.810 to 0.932, p<0.01) and ESIC/CGHS/employer-based insurance (adjusted OR 0.811, 95% CI 0.715 to 0.919, p<0.01), while private insurance (adjusted OR 0.851, p>0.1) showed no association (table 3). Online supplemental figure A3, a forest plot, shows the odds of outpatient CHE compared with the uninsured for all three insurance categories.

Table 4. Logistic regression of CHE for outpatient and inpatient care^*.

	Model 3
	CHE outpatient (10%)				CHE outpatient (20%)
Variable	Unadjusted OR		Adjusted OR		Unadjusted OR		Adjusted OR
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Insurance coverage (none)
AB-PMJAY or state health insurance	1.048	0.985 to 1.114	0.999	0.936 to 1.066	0.916**	0.855 to 0.982	0.869***	0.810 to 0.932
ESIC/CGHS/employer/others	0.791***	0.710 to 0.881	0.879**	0.786 to 0.983	0.699***	0.617 to 0.792	0.811***	0.715 to 0.919
Private insurance	0.839**	0.716 to 0.984	1.006	0.854 to 1.186	0.663***	0.544 to 0.809	0.851	0.693 to 1.046

	Model 4
	CHE inpatient (10%)				CHE inpatient (20%)
	Unadjusted OR		Adjusted OR		Unadjusted OR		Adjusted OR
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Insurance coverage (none)
AB-PMJAY or state health insurance	0.596***	0.560 to 0.634	0.721***	0.676 to 0.769	0.684***	0.641 to 0.730	0.790***	0.737 to 0.846
ESIC/CGHS/employer/others	0.613***	0.537 to 0.700	0.620***	0.550 to 0.699	0.661***	0.569 to 0.767	0.693***	0.603 to 0.796
Private insurance	0.629***	0.523 to 0.756	0.628***	0.514 to 0.768	0.641***	0.532 to 0.773	0.664***	0.543 to 0.812

Significance levels: ***p<0.01, **p<0.05, *p<0.1. The p<0.1 threshold is reported to indicate marginal significance, following standard econometric practice in large-sample survey analyses.

^*Refer to online supplemental table A4 for a detailed table of all the covariates: residence, household size, religion, social group, expenditure decile, education of the head of household and state/UT of residence.

AB-PMJAY, Ayushman Bharat–Pradhan Mantri Jan Arogya Yojana; CGHS, Central Government Health Scheme; CHE, catastrophic health expenditure; ESIC, Employees’ State Insurance Corporation; UT, union territory.

CHE during inpatient care (model 4)

Compared with uninsured households, all three insurance types—AB-PMJAY/state insurance, ESIC/CGHS/employer-based insurance and private insurance—were significantly (p<0.01) associated with lower odds of inpatient CHE (table 4). At the 10% threshold, adjusted ORs were 0.721 (95% CI 0.676 to 0.769, p<0.01) for AB-PMJAY/state insurance, 0.620 (95% CI 0.550 to 0.699, p<0.01) for ESIC/CGHS/employer-based insurance and 0.628 (95% CI 0.514 to 0.768, p<0.01) for private insurance, all statistically significant at the 1% level. Similarly, at the 20% threshold, adjusted ORs remained significant and below 1 for AB-PMJAY/state insurance (0.790, 95% CI 0.737 to 0.846, p<0.01), ESIC/CGHS/employer-based insurance (0.693, 95% CI 0.603 to 0.796, p<0.01) and private insurance (0.664, 95% CI 0.543 to 0.812, p<0.01), indicating a consistent protective effect of insurance against high inpatient financial burden. Online supplemental figure A4, a forest plot, shows log odds of facing inpatient CHE compared with the uninsured for all three insurance categories.

Online supplemental table A4 presents findings of the whole model 3 and model 4. Larger households had consistently lower odds of CHE. Scheduled tribes had among the lowest adjusted odds of inpatient CHE (0.416 at 20%), while scheduled castes also showed lower risk across outcomes.

OOPE in outpatient care (model 5)

Compared with the uninsured, AB-PMJAY or state insurance has a significant effect on reducing outpatient OOPE (coefficient –0.047, 95% CI −0.087 to −0.006, p<0.05) and a substantial reduction in inpatient OOPE (–0.205, 95% CI −0.258 to −0.153, p<0.01) (table 5). ESIC/CGHS and other employment-based insurance schemes significantly reduced OOPE for both outpatient (–0.090, 95% CI −0.157 to −0.023, p<0.05) and inpatient care (–0.265, 95% CI −0.361 to −0.169, p<0.01). In contrast, private insurance significantly reduced inpatient OOPE (–0.186, 95% CI −0.331 to −0.041, p<0.05). These results suggest that employment-based insurance and PFHI offer strong financial protection for inpatient and outpatient care. Online supplemental figure A5 is a forest plot comparing the log OOPE of outpatient and inpatient care by different categories.

Table 5. Heckman second-stage (OOPE model) coefficients for log outpatient and log inpatient OOPE^*.

	Model 5
Variables	Log OOPE: outpatient		Log OOPE: inpatient
	Coefficient	95% CI	Coefficient	95% CI
Insurance coverage (none)	–	–	–	–
AB-PMJAY or state health insurance	−0.047**	−0.087 to −0.006	−0.205***	−0.258 to −0.153
ESIC/CGHS/employer/others	−0.090***	−0.157 to −0.023	−0.265***	−0.361 to −0.169
Private insurance	0.074	−0.018 to 0.166	−0.186**	−0.331 to −0.041
Diagnostic tests
athrho	−1.877***	−1.913 to −1.842	−2.602***	−2.634 to −2.572
lnsigma	1.006***	0.985 to 1.027	1.729***	1.711 to 1.749
rho	−0.95425	−0.957 to −0.951	−0.98909	−0.990 to −0.988
sigma	2.73385	2.677 to 2.792	5.63842	5.531 to 5.746
lambda	−2.60879	−2.671 to −2.547	−5.57691	−5.686 to −5.468

Significance levels: ***p<0.01, **p<0.05, *p<0.1. The p<0.1 threshold is reported to indicate marginal significance, following standard econometric practice in large-sample survey analyses.

^*Refer to online supplemental table A5 for the full model, including all covariates and diagnostics. Please also refer to online supplemental Table A6 for the results of the naïve model (linear regression of OOPE and log OOPE).

AB-PMJAY, Ayushman Bharat–Pradhan Mantri Jan Arogya Yojana; CGHS, Central Government Health Scheme; ESIC, Employees’ State Insurance Corporation; OOPE, out-of-pocket expenditure.

Online supplemental table A5 presents the results from the Heckman two-stage model. The model diagnostics validate the relevance of the Heckman approach: the statistically significant and lambda (λ=−2.609 for outpatient; λ=−5.576 for inpatient) indicates that selection into healthcare utilisation is non-random and correlated with the error term in the expenditure equation. This affirms that conventional regression would underestimate the effects of key variables and that the Heckman correction improves the robustness of financial burden estimates in health expenditure studies. Online supplemental table A6 presents the results from naïve linear regression models estimating OOPE in INR and log-transformed form.

Socioeconomic disparities

The online supplemental appendix highlights marked socioeconomic disparities between uninsured households and those covered under AB-PMJAY. Across the tables, coverage under the scheme is more common among socially and economically disadvantaged groups, yet patterns of access and protection remain uneven. The magnitude of protection varies substantially by social group, income decile and state context. These findings (see online supplemental tables A1–A5) underscore that insurance expansion alone does not eliminate inequities: socioeconomic differences continue to shape who benefits most from coverage and the extent of financial risk protection.

Discussion

The study provides new insights into the association of PFHI with outpatient and inpatient utilisation, CHE and OOPE in India.

Healthcare use

The study demonstrates PFHI’s effectiveness in improving access to hospital care and confirming findings from previous studies.^{18 19} The paradoxical finding of higher use of outpatient care among insured individuals, although outpatient care is not covered, is in line with earlier studies suggesting that PFHI can indirectly influence health-seeking behaviour through increased awareness, trust in formal providers, improved health systems and perceived financial protection.^{615 58},⁶⁰ Andersen’s behavioural model of health service use⁵⁴ theoretical framing also helps explain why insured households demonstrate significantly higher outpatient utilisation despite the absence of explicit coverage, underscoring the importance of considering indirect behavioural pathways when evaluating PFHI’s impact on health service use in India.

In addition to indirect behavioural pathways, the observed higher outpatient utilisation among insured households may partly reflect reverse causality—where families with greater healthcare needs are more likely to be insured—or reporting bias, underscoring the need for longitudinal or quasiexperimental studies to disentangle these mechanisms. From a policy perspective, this shift results in continued financial vulnerability, as outpatient care remains a major driver of OOPE and CHE in India.^{8 10 42}

Catastrophic health expenditure

PFHI does not cover outpatient care costs, so the observed reduction in outpatient CHE at the 20% threshold is both intriguing and somewhat counterintuitive. One possible explanation is substitution from outpatient to inpatient care: patients who might otherwise have sought repeated outpatient visits could instead receive hospital-based treatment under PFHI, thereby lowering reported outpatient spending. Another mechanism may be reduced outpatient need following inpatient interventions—for example, effective inpatient care for chronic or acute conditions could diminish subsequent outpatient requirements.

At the same time, there is a possibility of scheme gaming, where providers convert cases that could be managed in outpatient settings into inpatient admissions to access PFHI reimbursements. Johnson and Krishnaswamy raised this concern in early evaluations of RSBY, and Bauhoff and Sudharsanan also cautioned about such dynamics in assessing health insurance schemes.^{5 51} More operational research is needed to disentangle these pathways and assess whether the reduction reflects genuine financial protection benefits or unintended distortions in care-seeking and provider behaviour. Future national health surveys should therefore integrate morbidity and provider-type information to allow researchers to examine whether outpatient reductions stem from improved care efficiency, substitution effects or provider gaming, ensuring a more accurate understanding of PFHI’s impact.

AB-PMJAY and state health insurance schemes are significantly associated with reduced incidence of CHE for inpatient care. The protective effect was stronger for ESIC, CGHS and employer-based schemes. These findings align with earlier studies documenting PFHI’s role in reducing CHE.^{6 8} However, the persistence of CHE among insured households highlights gaps in coverage depth, benefit design and private sector engagement.

Many studies in the past have stressed the importance of CHE due to outpatient care and the need to improve public health facilities and primary care to reduce outpatient OOPE.^{5 6 10 24 27 28 36 37 40} Some have argued for including outpatient care and preventive services in PFHI.^{31 37} The government’s policy response strengthened the primary healthcare network as part of AB-PMJAY.^{1 2} More operational research and evidence-based policy responses will be required to address the persistent CHE in outpatient care.

Out-of-pocket expenditure

The Heckman two-stage model results show that publicly funded and employer-based health insurance schemes significantly reduce household OOPE, particularly for inpatient care. Findings corroborate those of Harish et al and Aashima and Sharma, who documented reductions in CHE among insured groups, although inconsistently across regions.^{36 37} The findings, however, diverge from those of Garg et al,^{16 34} who reported no significant association between AB-PMJAY and reducing OOPE and CHE.

In absolute terms, the OOPE is still burdensome, even for those with insurance coverage. For inpatient care, PFHI households that sought inpatient care spent INR29 875, 20% of the annual household expenditure of INR145 848. Previous studies have highlighted the need to expand coverage, raise awareness about procedures to avail benefits, increase coverage limits and consider multiple morbidities, among others, to reduce inpatient OOPE.^{31 37 52}

Socioeconomic and regional disparities

The study reveals stark inequalities in the coverage and financial protection offered by PFHI, which aligns with previous studies.^{19 22 23 31 36} The average household monthly expenditure of the uninsured households is the lowest (INR11 373) compared with INR12 154 for PFHI, INR19 040 for ESIC/CGHS/employer-proposed insurance and INR24 827 for private health insurance.²³ This indicates that wealth-based inequalities in access to PFHI persist.

PFHI coverage is significantly higher in rural areas (54.5%) than in urban areas (30%). However, OOPE and CHE remain high in both contexts. Rural households incur lower inpatient OOPE (INR30 688) compared with urban (INR42 970) but experience higher inpatient CHE incidence (46.8% vs 45.2% at the 10% threshold), indicating greater vulnerability. Socioeconomically marginalised groups—scheduled tribes and the poorest deciles—report the lowest OOPE yet suffers disproportionately high relative CHE, underscoring the inadequacy of even low absolute spending.

Historical evolution of the insurance landscape

Most previous studies on PFHI in India focused on the earlier RSBY scheme, while only a few have assessed the initial phases of AB-PMJAY implementation. This study is the first to investigate the CHE and OOPE outcomes of AB-PMJAY after 4 years of the programme’s introduction using a large-scale, nationally representative survey. As summarised earlier, the study findings are largely positive in reducing CHE and OOPE for inpatient care (use, CHE and OOPE) and, to a limited extent, for outpatient care (use, CHE at 20% and OOPE). In contrast, evaluations of RSBY’s impact on healthcare utilisation, CHE and OOPE have yielded largely negative results.^{5 6 43 61} The enhanced design and implementation of PFHI under AB-PMJAY may account for some of the positive departures from earlier trends.^{1 2} Moreover, AB-PMJAY’s design included alignment with Ayushman Bharat Arogya Mandirs to offer primary care, diagnostics and essential medicines.⁶²

Policy implications and recommendations

This study confirms that India’s PFHI schemes, particularly AB-PMJAY, have expanded access to inpatient care and provided meaningful financial protection. Yet, outpatient care remains a glaring gap, with persistently high OOPE driving catastrophic expenditures even among insured households. The urgency of reform is clear.

First, integrating outpatient coverage into PFHI benefit packages is essential to address the largest source of household health spending and to protect the poor from everyday financial shocks. A phased approach could involve piloting outpatient coverage schemes in select states or districts to assess feasibility, costs and impacts before nationwide scale-up.^{8 19}

Second, strengthening primary healthcare through Ayushman Arogya Mandirs offers a cost-effective pathway to deliver outpatient and preventive services, reduce avoidable hospitalisations and anchor financial protection in a strong continuum of care.^{1 2} Linking Mandirs with PFHI reimbursements for defined outpatient services could enhance both utilisation and equity.

Third, targeted equity-focused interventions are needed to close socioeconomic gaps that persist despite broad PFHI coverage. These could include expanding benefit packages to cover high-cost chronic diseases, subsidising transportation and indirect costs in remote or tribal areas and tailoring outreach to disadvantaged groups through simplified enrolment, real-time helpdesks and user-friendly communication strategies.^{7 23 24 51 52 62}

Fourth, adaptive, state-specific implementation frameworks should be institutionalised to respond to wide interstate variations in health system capacity, private sector engagement and population needs. This requires not only stronger referral linkages but also systematic monitoring, evaluation and accountability mechanisms to track coverage depth and financial protection outcomes across states.^{16 18 21 31 40 60 63}

Finally, aligning PFHI reforms with broader health system strengthening—through provider payment reforms, improved contracting with the private sector and enhanced benefit awareness programmes—can help ensure that financial protection translates into equitable and efficient service delivery.^{17 32}

Limitations and data gaps

The study has several limitations that should be acknowledged to contextualise the findings and inform the design of future research.

Health insurance status is collected at the household level and not at the individual level. However, this is consistent with the policy architecture of PFHI schemes. Hence, the outcome and exposure variables used are both valid and policy aligned.

CAMS 2022–2023 lacks morbidity and clinical information, limiting the ability to adjust for health needs or disease severity, which are key confounders in estimating utilisation and OOPE. Since PFHI enrolment is automatic and free for eligible households, the distortions resulting from information asymmetry and moral hazard are limited. However, PFHI coverage is not condition specific; it covers all inpatient procedures up to a defined ceiling. Therefore, variation in disease type does not meaningfully affect financial protection metrics.

The study does not include stratified or subgroup analyses (eg, by rural-urban residence or income deciles) or interaction terms to empirically test for heterogeneity in the association between PFHI coverage and outcomes across socioeconomic groups. While we present descriptive evidence of socioeconomic disparities in the online supplemental tables, the absence of formal subgroup modelling limits the ability to identify differential effects of PFHI across population segments. Future studies should address this gap through stratified models, interaction terms or concentration indices.

Nonetheless, the absence of morbidity and diagnosis data constrains the analysis of utilisation patterns, treatment-seeking behaviour and the distribution of financial burden across communicable and non-communicable diseases. Rich morbidity data would allow researchers to explore heterogeneity in PFHI’s effectiveness across disease categories, better identify high-cost conditions and assess equity implications for chronic and catastrophic illnesses. Future all-India health surveys should therefore incorporate morbidity and clinical modules to ensure that the richness of such insights is not missed and to enable more nuanced evaluation of PFHI’s role in advancing universal health coverage.

CAMS 2022–2023 does not record the type of healthcare provider (public or private) from which care was sought. Given known differences in OOPE across provider types, this restricts the ability to disaggregate effects by sector. However, our analysis addresses this issue by including state fixed effects in all regression models. This controls for unobserved, time-invariant characteristics of state health systems, including the public-private mix, thereby improving the robustness of the estimated associations.^{27 31 33 37}

The survey relies on self-reported expenditure data for outpatient and inpatient care, which is inherently subject to recall bias, particularly for events occurring over longer recall periods such as hospitalisations. In addition, reliance on self-reported insurance status introduces the possibility of misclassification, as households may conflate different types of schemes (eg, state-funded PFHI, employer-provided insurance or private insurance) or lack full awareness of their eligibility. Both recall bias and misclassification could attenuate or exaggerate associations between PFHI coverage, utilisation and financial protection.

As a cross-sectional study, the analysis cannot establish causal relationships. The possibility of reverse causality remains, but the effect is limited due to the nature of PFHI, as explained earlier. Finally, the study does not include indirect costs, such as wage loss, caregiving burden or transportation expenditures beyond those recorded, potentially underestimating the full economic burden of illness.

Future research and surveys

These limitations underscore the need for future household surveys to collect individual-level insurance linkage, morbidity profiles, provider type and broader cost components. To strengthen causal inference, future research should employ panel data or quasiexperimental designs—for instance, difference-in-differences exploiting phased roll-outs of PFHI across states or natural experiments leveraging staggered eligibility criteria—to provide more robust evidence on PFHI’s impact on financial risk protection in India. Future studies should examine whether the observed outpatient utilisation patterns under PFHI reflect behavioural pathways, reverse causality or reporting bias, using longitudinal or quasiexperimental designs.

Future studies may also apply concentration indices and interaction models to CAMS and related datasets to generate more nuanced evidence on socioeconomic inequalities in healthcare utilisation and financial protection under PFHI. In addition, future studies should build on our findings by conducting subgroup analyses—such as stratification by rural-urban residence, income deciles or interaction models—to provide deeper empirical insights into the inequalities already highlighted in this paper.

Future studies with higher rates of missing expenditure data should consider multiple imputation or predictive mean matching to preserve the full variance of the expenditure distribution.

Conclusion

This study confirms that India’s PFHI schemes, particularly AB-PMJAY, have expanded access to inpatient care and provided meaningful financial protection. Yet, outpatient care remains a glaring gap, with persistently high OOPE driving catastrophic expenditures even among insured households.

The urgency of reform lies in addressing both coverage gaps and systemic inequities. Expanding PFHI to include outpatient services, anchored in strengthened primary healthcare through Ayushman Arogya Mandirs, would reduce everyday financial shocks and avoidable hospitalisations. At the same time, targeted support for disadvantaged groups and state-specific adaptive strategies are essential to ensure equitable benefits and responsive implementation across diverse contexts.

Taken together, these findings underscore that incremental improvements are no longer sufficient. India must act decisively to expand the depth of PFHI coverage, align it with primary care reforms and strengthen adaptive governance. Doing so would reduce financial vulnerability for millions of households and move the country closer to its goal of universal health coverage and health system resilience.

Future studies should employ longitudinal or quasiexperimental designs, such as difference-in-differences exploiting phased AB-PMJAY roll-out, to better establish causal pathways. Household surveys should collect morbidity, provider type and indirect costs to strengthen evaluation.

Data availability statement

Data are available in a public, open access repository.