Financial protection failure among urban poor populations in Karachi, Pakistan: a cross-sectional household survey of catastrophic health expenditure prevalence, determinants and insurance effectiveness Authors: Ahsanullah Khan, Shahzad Ali Khan, Ali Akbari-Sari, Mostafa Hosseini, Amirhossein Takian Published: 2026 ABSTRACT AbstractIntroductionPakistan has one of South Asia’s highest burdens of out-of-pocket health expenditure. The disaggregated evidence for urban poor populations in major cities remains limited and hidden under the national data. Among the big cities in Pakistan, Karachi’s informal settlements concentrate the country’s most financially vulnerable households where formal health insurance coverage is negligible. This study measured catastrophic health expenditure prevalence among urban poor households in Karachi, determined its socioeconomic distribution and evaluated the financial protective effect of existing health insurance.MethodsA cross-sectional household survey was conducted across six administrative districts of Karachi between March and August 2023, covering informal settlements and low- and middle-income neighbourhoods. Multi-stage stratified cluster probability sampling yielded 1211 households (response rate 94.7%), with household heads or primary economic decision-makers as respondents. Catastrophic health expenditure was assessed at 10%, 25% and 40% WHO thresholds using standardised methodology. Concentration index analysis quantified socioeconomic distribution of the burden. Multivariate logistic regression identified determinants and assessed insurance effectiveness. Health expenditure-induced impoverishment and financial coping mechanisms were also examined.ResultsCatastrophic health expenditure affected 91.7% (95% CI 90.1% to 93.1%) of households at the 10% threshold, 78.2% (95% CI 75.8% to 80.5%) at 25% and 66.1% (95% CI 63.4% to 68.8%) at 40%. Prevalence was uniformly high across income tertiles (lowest 95.0%, middle 93.8%, highest 86.4%), yielding a near-zero concentration index of −0.048 that contradicts the expected pro-poor gradient. Health expenditures impoverished 342 households (28.2%), pushing them below the national poverty line. Hospitalisation (adjusted OR 4.8, 95% CI 3.2 to 7.1) and chronic illness (adjusted OR 2.9, 95% CI 2.1 to 4.0) were the strongest determinants. Insurance coverage, present in only 68 households (5.6%), provided no significant financial protection (adjusted OR 0.82, 95% CI 0.43 to 1.57, p=0.553).ConclusionsNear-universal catastrophic burden distributed uniformly across income groups argues against means-tested targeting and supports universal financial protection approaches. Existing insurance provides no meaningful protection, pointing to systemic failure requiring fundamental health financing reform rather than incremental programme expansion. WHAT IS ALREADY KNOWN ON THIS TOPIC Pakistan reports one of South Asia’s highest national catastrophic health expenditure rates, yet disaggregated data for urban poor populations in major cities remain sparse. Pakistan reports one of South Asia’s highest national catastrophic health expenditure rates, yet disaggregated data for urban poor populations in major cities remain sparse. Existing studies have not tested whether catastrophic spending in Pakistani urban poor households follows the conventional pro-poor gradient or affects all income groups uniformly. Prior household-level quantitative evidence on whether health insurance protects against catastrophic expenditure in this population has been absent. Existing studies have not tested whether catastrophic spending in Pakistani urban poor households follows the conventional pro-poor gradient or affects all income groups uniformly. Prior household-level quantitative evidence on whether health insurance protects against catastrophic expenditure in this population has been absent. WHAT THIS STUDY ADDS Catastrophic health expenditure affects 91.7% of urban poor households in Karachi, two to three times the rates documented in comparable South Asian urban settings. Catastrophic health expenditure affects 91.7% of urban poor households in Karachi, two to three times the rates documented in comparable South Asian urban settings. A near-zero concentration index (−0.048) demonstrates that catastrophic burden is distributed almost uniformly across income tertiles, going against the assumed pro-poor pattern. A near-zero concentration index (−0.048) demonstrates that catastrophic burden is distributed almost uniformly across income tertiles, going against the assumed pro-poor pattern. Existing insurance coverage provides no statistically significant financial protection. Existing insurance coverage provides no statistically significant financial protection. Health payments pushed 28.2% of previously non-poor households below the national poverty line, with informal borrowing and asset sales being the dominant coping strategies. Health payments pushed 28.2% of previously non-poor households below the national poverty line, with informal borrowing and asset sales being the dominant coping strategies. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY The near-universal burden pattern argues against the approach for income-based targeting and supports universal financial protection for urban poor populations. The near-universal burden pattern argues against the approach for income-based targeting and supports universal financial protection for urban poor populations. Extending the Sehat Sahulat Programme’s benefit package to outpatient care and essential medicines would address the cost categories that drive catastrophic spending. Extending the Sehat Sahulat Programme’s benefit package to outpatient care and essential medicines would address the cost categories that drive catastrophic spending. Federal coordination of health financing standards is needed so that mobile urban populations are not excluded by provincial implementation gaps such as the one observed in Sindh. Federal coordination of health financing standards is needed so that mobile urban populations are not excluded by provincial implementation gaps such as the one observed in Sindh. Comparable low-governance, high-informality settings should test whether their catastrophic expenditure also shows near-universal rather than pro-poor distribution. Comparable low-governance, high-informality settings should test whether their catastrophic expenditure also shows near-universal rather than pro-poor distribution. Introduction Recent systematic evidence indicates that while the global average catastrophic health expenditure (CHE) rate stands at approximately 13%, the burden is considerably heavier in low- and middle-income countries (LMICs), where average prevalence reaches 16% with pronounced regional variation.12 South Asia bears a disproportionate share: Bangladesh reports CHE of around 24%, and India approximately 17%, both substantially exceeding figures documented across much of sub-Saharan Africa.13 When the analysis narrows to growing urban areas with large populations living in informal settlements and working in the informal economy, the protective effects of insurance and other policy interventions vary further still. Several Southeast Asian countries have achieved notably lower rates following comprehensive insurance reforms, while others with structurally similar economies have stalled. Pakistan occupies a distinctive and instructive position within this landscape. The country’s national CHE rate is approximately 54%, already among the highest globally, though disaggregated data for urban poor populations in major metropolitan areas remain limited.4–6 Pakistan represents an example of stalled insurance expansion among informal sector workers, in contrast to other nations that have successfully developed large-scale insurance programmes for similar populations.7–12 With 73% of its workforce in the informal sector, Pakistan sits between countries that have successfully expanded coverage, such as Thailand at 55% informality and Indonesia at 60%, and those where expansion has stalled, such as Bangladesh at 82%.13–16 Yet Pakistan’s health financing outcomes resemble those of the stalled cases far more closely than its structural position alone would predict, challenging the assumption that the size of the informal sector mechanically determines the difficulty of insurance expansion (table 1).1718 Table 1Pakistan within comparative health financing landscape, contrasting with successful UHC cases and peer contextsDimensionThailandIndonesiaPakistanBangladeshNigeriaUHC timeline2002–2009 (7 years→99%)2014–2019 (5 years→220M)2015–2025 (10 years→40M)2001–present (stalled 2%)2005–present (fragmented)GovernanceUnitary, National Health Security Office (NHSO)Coordinated decentralFederal fragment. Post-2010Unitary, weakFederal fragmentCHE rate (%)11.415.654.025.131.8Informal (%)5560738265Coverage (%)99.582.45.62.13.5CHE, catastrophic health expenditure; UHC, universal health coverage. Pakistan within comparative health financing landscape, contrasting with successful UHC cases and peer contexts CHE, catastrophic health expenditure; UHC, universal health coverage. Several gaps in the existing evidence base identified the need for this study. First, earlier research on CHE in Pakistan has relied on national household income and expenditure surveys with limited precision of estimates for the urban poor specifically.11920 Second, no study to our knowledge has tested whether catastrophic spending among Pakistan’s urban poor follows the conventional pro-poor gradient or whether the burden might be distributed more uniformly, a distinction with direct implications for targeting policy. Third, while qualitative and programmatic assessments have questioned the protective capacity of Pakistan’s health insurance, household-level quantitative evidence on whether insurance coverage reduces catastrophic expenditure risk among the urban poor has been limited. Fourth, the interplay between catastrophic spending, impoverishment depth and the cascading financial coping strategies by affected households has not been examined in an integrated way for this population. Within Pakistan, Karachi offers an appropriate study setting as the country’s largest city and economic centre; it contains substantial socioeconomic, cultural and ethnic diversity within a single metropolitan area.21–24 Its health ecosystem has public facilities offering subsidised care with a dominant private sector. The city has a large population of urban poor living in informal settlements with variable access to formal health services. Notably, the government-financed Sehat Sahulat Programme, Pakistan’s primary social health protection scheme, is not implemented by the Sindh provincial government in Karachi, though beneficiaries registered in other provinces and residing in the city can access services at empanelled hospitals. This creates a natural setting in which to examine financial protection failure under conditions of minimal formal insurance infrastructure. Research aims This study aimed to: (1) measure the prevalence of CHE among urban poor populations across Karachi’s six administrative districts using WHO-standardised methodology; (2) determine whether catastrophic spending follows the expected pro-poor distribution or a near-universal pattern through concentration index analysis; (3) identify factors independently associated with CHE through multivariate analysis; and (4) document health expenditure-induced impoverishment and the financial coping mechanisms deployed by affected households. Methods Study design and setting We conducted a cross-sectional household survey in Karachi, Pakistan, between March and August 2023. Data collection covered informal settlements (katchi abadis) and adjacent low- and middle-income neighbourhoods across all six administrative districts including Central, East, Korangi, Malir, South and West. Karachi’s socioeconomic heterogeneity, diverse urban poor populations and health indicator disparities allowed coverage of a broad sociocultural and religious spectrum.21 Study population and eligibility The target population included households residing in identified katchi abadis (informal settlements) and adjoining low- and middle-income neighbourhoods, identified using local government records at district and union council levels and family folders maintained by Lady Health Workers (community health workers). Household heads or primary economic decision-makers aged 18 years or older served as respondents. Households currently enrolled in any health insurance plan were not excluded from the survey; their insurance status was recorded as an independent variable. Sample size calculation We applied standard formulae for population proportions with design effect adjustment for cluster sampling.25 Assuming a CHE proportion of 50% (maximum variance, yielding the most conservative sample size), 95% confidence level, 5% margin of error and a design effect of 2.5 to account for intra-cluster correlation, the minimum required sample was calculated as n=(1.96²×0.5×0.5/0.05²)×2.5=960 households. We increased the target to 1200 to account for potential non-response and to enable district-level disaggregation. The realised sample of 1211 completed interviews provided adequate statistical power for citywide estimates. Sampling strategy Multi-stage stratified cluster probability sampling generated representative estimates while addressing the absence of comprehensive household sampling frames in informal urban settlements. In the first stage, sampling was stratified by administrative district, with probability-proportional-to-size selection of 24 union councils (four per district) as primary sampling units. Union councils, the smallest administrative units containing several thousand households, serve as natural clusters with available population data from the 2017 Census.21 In the second stage, field teams developed comprehensive household lists within each selected union council by combining Lady Health Worker programme family registers with walking-list verification procedures. Within each union council, 50–51 households were systematically sampled using a fixed skip interval calculated by dividing the total eligible households on the consolidated list by the required cluster sample size, with a random start point selected using a random number generator. This approach yielded self-weighting samples where each eligible household within sampled union councils possessed equal selection probability. The complete sampling, from target population through recruitment to analysis, is given in online supplemental file 5. Questionnaire development and content validation The structured survey questionnaire was adapted from WHO household health expenditure survey instruments,2627 modified for the local context through expert review and pilot testing. The instrument comprised six modules including household sociodemographic characteristics, health expenditure patterns over a 4-week recall period capturing direct medical costs and indirect expenses, existing health insurance coverage and utilisation, financial resilience indicators and social capital measures using validated cohesion scales.27 Content validity was assessed by two independent panels comprising health financing experts and field implementation specialists. Items achieving content validity index values of 0.78 or above across both panels were retained. All items achieved content validity index scores of 0.75–1.00 with inter-rater agreement (Cohen’s kappa) of 0.75–1.00. Pilot testing with 50 households in one non-study union council identified difficulties related to the delivery and understanding of the survey and language in it, leading to simplified phrasing for economic questions. The refined questionnaire required 45–60 min for administration (Summary of elements of the survey are given in table 2). The full survey questionnaire is provided in online supplemental file 1. Table 2Summary of elements of the surveyModuleContent domainsPrimary measuresRecall periodSociodemographic characteristicsHousehold composition, education levels, occupation types, housing conditionsHousehold size, dependency ratio, years of education, employment category, dwelling typeCurrent statusEconomic statusIncome sources, consumption expenditure, asset ownership, wealth indicatorsMonthly income (PKR), total expenditure, asset ownership index, poverty proxy means testPast monthHealth expenditureOut-of-pocket spending on all healthcare services and productsConsultation fees, medications, procedures, hospitalisation, diagnostics, transport4 weeksInsurance coverageCurrent enrollment status, premium payments, benefit utilisationCoverage type, premium amount (PKR), utilisation frequency, satisfaction ratingsCurrent statusFinancial resilienceSavings capacity, borrowing access, social support networksSavings amount (PKR), credit availability, family support, coping strategiesCurrent capacitySocial capitalCommunity cohesion, organisational participation, institutional trustCohesion scale (1–5), group membership (yes/no), trust indices, civic engagementGeneral perceptionAll modules were administered in Urdu with translation protocols for local dialects. Content validity index for all items ranged 0.75–1.00, indicating excellent validity. Questionnaire required 45–60 min for completion. Data were collected via tablet-based SurveyCTO with built-in skip logical and validation checks.PKR, Pakistani rupees. Summary of elements of the survey All modules were administered in Urdu with translation protocols for local dialects. Content validity index for all items ranged 0.75–1.00, indicating excellent validity. Questionnaire required 45–60 min for completion. Data were collected via tablet-based SurveyCTO with built-in skip logical and validation checks. PKR, Pakistani rupees. Data collection procedures District-specific field teams comprising 8–10 enumerators and two supervisors implemented data collection. Enumerators were Lady Health Workers or community health workers with prior household survey experience and local language fluency. Female enumerators proved essential for accessing households in conservative neighbourhoods. Comprehensive 2-day training sessions were conducted before the fieldwork in each district, covering research ethics, questionnaire content and skip patterns, standardised probing techniques for sensitive economic questions, sampling implementation and tablet-based data capture using SurveyCTO with built-in skip logic and validation checks. Field implementation followed standardised protocols with three revisit attempts for initially unavailable households. Supervisors validated 10% of interviews through callbacks within 48 hours. Outcome definitions and measurement CHE was measured using the WHO capacity-to-pay methodology.2829 Households reported out-of-pocket payments for all health services during a 4-week recall period. Total household expenditure, including all health and non-health spending, was calculated over the same period. Capacity to pay was defined as total household expenditure minus subsistence spending, where subsistence spending was estimated as food expenditure for households whose food share fell between the 45th and 55th percentiles of the sample distribution. A household was classified as experiencing CHE when out-of-pocket health spending exceeded 10%, 25% or 40% of capacity to pay. Impoverishment was assessed by comparing pre-payment and post-payment household income against Pakistan’s national poverty line of Pakistani rupees (PKR) 3705 per capita per month.30–32 Reported national catastrophic expenditure rates for Pakistan and comparator countries vary in part with denominator choice. The estimates calculated against capacity to pay (non-subsistence expenditure) are systematically higher than those calculated against total household consumption (the Sustainable Development Goal 3.8.2 indicator), because exclusion of food spending shrinks the denominator. Where international comparisons are drawn in the Results and Discussion sections, we indicate the underlying methodology to enable like-for-like interpretation. Statistical analysis Data analysis used IBM SPSS Statistics version Version 24.0.33 Descriptive statistics characterised household demographics and expenditure patterns. CIs for prevalence estimates used the Wilson score method with adjustment for clustering.34 Catastrophic expenditure distribution across socioeconomic strata was examined using household income tertiles. The Erreygers-corrected concentration index quantified whether catastrophic expenditure exhibited pro-poor or pro-rich concentration, calculated following the procedures described by O’Donnell and colleagues.2935 Analyses proceeded on a complete-case basis. The multivariate logistic regression model included the 1176 households with complete data on all covariates; this was equivalent to 97.1% of the completed sample. Multivariate logistic regression identified factors independently associated with CHE at the 10% threshold. Candidate independent variables considered in the initial full model comprised district, household size, education level, occupation, income tertile, asset ownership, presence of elderly members or children in the household, chronic illness, recent hospitalisation, distance to the nearest health facility, household savings and insurance status. Backward stepwise elimination with a retention threshold of p<0.10 yielded the final parsimonious model, which retained hospitalisation, chronic illness, income tertile, household size, distance to the nearest health facility, insurance status, education level and household savings (table 3). Variables that did not meet the retention threshold (district, occupation, asset ownership and presence of elderly members or children) were excluded from the reported model. Model fit was assessed using the Hosmer-Lemeshow goodness-of-fit test (p=0.34, indicating adequate fit). All tests were two-sided with significance level alpha=0.05. Full variable definitions and coding are provided in online supplemental file 2, and expanded analytical procedures, including the concentration index computation and instrument content-validity results, in online supplemental file 3. Table 3Multivariate logistic regression: factors independently associated with catastrophic health expenditure (10% threshold)VariableCategory/unitAdjusted OR95% CIP valueHospitalisation (past 12 months)Yes vs no4.83.23 to 7.13 <0.001 Chronic illness in householdYes vs no2.92.11 to 3.99 <0.001 Income bandsMiddle vs lowest0.480.31 to 0.730.001Highest vs lowest0.410.26 to 0.65<0.001Household sizePer additional member1.181.05 to 1.330.006Distance to nearest facility>5 km vs ≤5 km1.520.97 to 2.380.068Has health insuranceYes vs no0.820.43 to 1.570.553Education levelPer level increase0.880.74 to 1.050.163Has savingsYes vs no0.680.43 to 1.070.090n=1176 households (complete-case analysis); overall catastrophic health expenditure (CHE) prevalence 1111/1211 (91.7%); model: backward stepwise logistic regression.Income tertiles: lowest=bottom 33% (Pakistani rupees (PKR) 8000–22 000/month), middle=middle 33% (PKR 22 001–36 000/month), highest=top 33% (PKR 36 001–85 000/month). All tertiles within urban poor classification. ORs indicate middle (OR 0.48) and highest (OR 0.41) tertiles show statistically lower odds of CHE, but absolute prevalence remains catastrophically high across all groups (86%–95%), indicating weak protective effect of higher income within this vulnerable population. Multivariate logistic regression: factors independently associated with catastrophic health expenditure (10% threshold) n=1176 households (complete-case analysis); overall catastrophic health expenditure (CHE) prevalence 1111/1211 (91.7%); model: backward stepwise logistic regression. Income tertiles: lowest=bottom 33% (Pakistani rupees (PKR) 8000–22 000/month), middle=middle 33% (PKR 22 001–36 000/month), highest=top 33% (PKR 36 001–85 000/month). All tertiles within urban poor classification. ORs indicate middle (OR 0.48) and highest (OR 0.41) tertiles show statistically lower odds of CHE, but absolute prevalence remains catastrophically high across all groups (86%–95%), indicating weak protective effect of higher income within this vulnerable population. Patient and public involvement Urban poor community representatives participated in formative research informing questionnaire development and sampling strategy design, particularly in identifying katchi abadi (informal settlements) boundaries and culturally appropriate phrasing for economic questions. Community members were not directly involved in statistical analysis or manuscript preparation. Results Characteristics of the study population The final sample comprised 1211 households across Karachi’s six administrative districts, achieving a 94.7% response rate from 1280 contacted households. The mean age of household heads was 41.1 years with 88.9% male. Mean household size was 5.2 members. Educational attainment varied widely, ranging from 31.2% with no formal education, 38.5% completed primary, 23.7% secondary and 6.6% reached higher education. Mean monthly household income was PKR 28 450 (approximately US$104 at 2023 exchange rates). Only 68 households (5.6%) reported any existing health insurance coverage, substantially below Pakistan’s national average of approximately 14%.5736 Of these, the majority held private sector policies or were registered under the Sehat Sahulat Programme through provinces other than Sindh, as the programme is not implemented by the Sindh provincial government in Karachi (Summary of key sociodemographic and health characteistics are given in table 4). Full sociodemographic and health characteristics disaggregated by catastrophic health expenditure status are provided in online supplemental table 1. Table 4Summary of key sociodemographic and health characteristics of study populationCharacteristicTotal SampleN=1211 n (%)P value(CHE vs non-CHE)District distribution (range)189–215 per district (15.6%–17.8%)0.86–0.96Household size, mean (SD)5.2 (2.1)0.27Household head age (years), mean (SD)41.1 (14.7)<0.001Male household head1077 (88.9%)<0.001Education level No formal education378 (31.2%)0.16 Primary completed466 (38.5%)0.91 Secondary completed287 (23.7%)0.08 Higher education80 (6.6%)0.50Employment status Informal sector828 (68.4%)0.06 Formal sector204 (16.8%)0.16 Unemployed179 (14.8%)0.36Monthly household income, PKR mean (SD)28 450 (12 380)<0.001Income tertile Low (PKR 8000–22 000/month)404 (33.4%)0.003 Lower-middle (PKR 22 001–36 000/month)403 (33.3%)0.30 Middle (PKR 36 001–85 000/month)404 (33.4%)0.05Chronic illness present547 (45.2%)<0.001Recent hospitalisation (past 12 months)438 (36.2%)<0.001Any health insurance coverage68 (5.6%)0.29Full data disaggregated by CHE status are provided in online supplemental table S1.Data are n (%) unless otherwise specified. CHE (out-of-pocket health spending ≥10% of total household capacity to pay). P values from χ² tests for categorical variables and independent samples t-tests for continuous variables comparing CHE versus non-CHE households. PKR (exchange rate: PKR 285=US$1, 2023). Income tertiles defined within the study sample: low=bottom 33% (PKR 8000–22 000/month); lower-middle=middle 33% (PKR 22 001–36 000/month); middle=top 33% (PKR 36 001–85 000/month). All tertiles represent households within the urban poor classification. Full district-level and CHE-disaggregated data are provided in online supplemental file 1.CHE, catastrophic health expenditure; PKR, Pakistani rupees. Summary of key sociodemographic and health characteristics of study population Full data disaggregated by CHE status are provided in online supplemental table S1. Data are n (%) unless otherwise specified. CHE (out-of-pocket health spending ≥10% of total household capacity to pay). P values from χ² tests for categorical variables and independent samples t-tests for continuous variables comparing CHE versus non-CHE households. PKR (exchange rate: PKR 285=US$1, 2023). Income tertiles defined within the study sample: low=bottom 33% (PKR 8000–22 000/month); lower-middle=middle 33% (PKR 22 001–36 000/month); middle=top 33% (PKR 36 001–85 000/month). All tertiles represent households within the urban poor classification. Full district-level and CHE-disaggregated data are provided in online supplemental file 1. CHE, catastrophic health expenditure; PKR, Pakistani rupees. CHE prevalence CHE prevalence at the 10% threshold was 91.7% (95% CI 90.1% to 93.1%), affecting 1111 of the 1211 surveyed households. At the 25% threshold, prevalence was 78.2% (947 households, 95% CI 75.8% to 80.5%), and at the 40% threshold, 66.1% (801 households, 95% CI 63.4% to 68.8%). The distribution of catastrophic expenditure across income tertiles revealed a pattern that departs from expectations. Prevalence remained uniformly high: 95.0% (95% CI 92.6% to 97.0%) in the lowest tertile (PKR 8000–22 000 per month), 93.8% (95% CI 91.2% to 95.8%) in the middle tertile (PKR 22 001–36 000) and 86.4% (95% CI 82.8% to 89.5%) in the highest tertile (PKR 36 001–85 000). The difference between lowest and middle tertiles did not reach statistical significance (p=0.43), while the difference between lowest and highest tertiles did (p<0.001). The Erreygers-corrected concentration index was −0.048, indicating a near-zero socioeconomic gradient and a pattern approximating universal catastrophic burden rather than the pro-poor concentration typically documented in comparable settings (table 5).3738 Table 5Catastrophic health expenditure prevalence by income tertileIncome tertileIncome range (PKR/month)NCHE casesCHE prevalence (%)95% CI lower95% CI upperP valueLowest (bottom 33%)8000–22 00040438495.092.697.0RefMiddle (middle 33%)22 001–36 00040337893.891.295.80.43Highest (top 33%)36 001–85 00040434986.482.889.5<0.001Total8000–85 0001211111191.790.193.1–CHE (≥10% household expenditure). Tertiles=three equal groups. All households within urban poor classification. P values from χ2 test versus lowest tertile.CHE, catastrophic health expenditure; PKR, Pakistani rupees. Catastrophic health expenditure prevalence by income tertile CHE (≥10% household expenditure). Tertiles=three equal groups. All households within urban poor classification. P values from χ2 test versus lowest tertile. CHE, catastrophic health expenditure; PKR, Pakistani rupees. Catastrophic expenditure also persisted across education levels. Among household heads with no formal education, 93.5% experienced catastrophic burden, compared with 89.0% among those holding a bachelor’s degree or higher (p=0.08). District variation was modest, ranging from 89.5% in Karachi Central to 94.0% in Malir (p=0.21). The typical pattern across South and Southeast Asia shows urban populations experiencing lower catastrophic spending than their rural counterparts. Pakistan’s urban poor data showing 91.7% prevalence with minimal gradients across tertiles represents a reversal of this expected protective pattern, suggesting not insufficient urban infrastructure requiring incremental development but rather a more fundamental health financing system failure. Regional and global comparisons of CHE Pakistan’s CHE prevalence among the urban poor exceeded the rates documented for Bangladesh’s urban poor (35%–42%, Dhaka slums) by a factor of 2.2–2.6, and for comparable Indian urban settings (28%–45%) by a factor of 2.0–3.3 (table 6). To position this finding within the global distribution, we compared the current estimate against published CHE prevalence for lower-middle-income countries available in the WHO Global Health Expenditure Database for the period 2015–2022, using the standardised 10% of total household consumption threshold.52839 Applying this consistent threshold, Pakistan’s urban poor prevalence of 91.7% represents 8.5 times the lower-middle-income country’s mean of 10.8% (range 1.4%–57.3%). We note that some comparator country surveys define catastrophic expenditure against non-food consumption or capacity-to-pay denominators rather than total consumption, and studies applying such alternative denominators tend to yield higher national estimates. Our comparisons are therefore conservative; had an alternative denominator been applied to Pakistan’s data, the relative position would be at least as extreme. Table 6Regional and global comparisons of catastrophic health expenditure prevalence and concentration indicesCountryNational catastrophic health expenditure prevalenceUrban poor CHEConcentration indexPakistan54.0%91.7% (current study)−0.048Bangladesh24.6% (2022)—Bangladesh HIES35%–42% (Dhaka slums)—BRAC studies−0.18 to −0.22India17.3% (2019–2021) NSSO 75th Round (varies by state)28%–45% (varies by city)—local slum studies−0.25 to −0.30Indonesia3.5% (2019)—Susenas SurveyData not available−0.28Thailand2.1% (2021)—Thailand Health Welfare Survey~3%–5% (estimated)−0.31 to −0.35Nigeria7.3% (2018–2019) Nigeria Living Standards SurveyData not availableNot calculatedEthiopia4.2% (2019)—Ethiopia HH Health Exp SurveyData not available−0.15 to −0.20AfghanistanData not availableData not availableNo dataData period: 2021–2023. Sources: WHO GHED, World Bank, Surveys.HIES = Household Income and Expenditure Survey (Bangladesh Bureau of Statistics); BRAC = Bangladesh-based development organisation, formerly Bangladesh Rural Advancement Committee; NSSO = National Sample Survey Office, Government of India (75th Round, 2017–18); HH = householdCHE, catastrophic health expenditure. Regional and global comparisons of catastrophic health expenditure prevalence and concentration indices Data period: 2021–2023. Sources: WHO GHED, World Bank, Surveys. HIES = Household Income and Expenditure Survey (Bangladesh Bureau of Statistics); BRAC = Bangladesh-based development organisation, formerly Bangladesh Rural Advancement Committee; NSSO = National Sample Survey Office, Government of India (75th Round, 2017–18); HH = household CHE, catastrophic health expenditure. Threshold sensitivity analysis demonstrates that the catastrophic burden persists across all threshold levels, including 91.7% at the standard 10% threshold, 78.2% at the 25% threshold and 66.1% even at the stringent 40% threshold (figure 1). This pattern contrasts sharply with regional comparators, where prevalence falls steeply at higher thresholds. Thailand reports a national catastrophic expenditure rate of approximately 2.1% at the 10% threshold, with prevalence approaching zero at higher thresholds. Statistical tests confirm Pakistan as an extreme outlier, with prevalence substantially exceeding the global average and the lower-middle-income country average. The financial burden in Pakistan was also significantly greater than in other South Asian countries, including Bangladesh, India, Nepal and Sri Lanka. Notably, Pakistan’s prevalence exceeded the average for fragile and conflict-affected states, suggesting that extreme financial burden persists independently of active conflict. Prevalence of catastrophic health expenditure (CHE) across threshold levels in Pakistan and comparator countries. Factors associated with CHE Multivariate logistic regression identified several factors independently associated with catastrophic health expenditure at the 10% threshold after controlling for demographic, economic and health-related variables. Recent hospitalisation emerged as the strongest predictor (adjusted OR 4.8, 95% CI 3.2 to 7.1, p<0.001), followed by the presence of chronic illness in the household (adjusted OR 2.9, 95% CI 2.1 to 4.0, p<0.001). Households in the lowest income tertile had higher odds of catastrophic spending compared with the highest tertile (adjusted OR for highest vs lowest: 0.41, 95% CI 0.26 to 0.65, p<0.001), though prevalence remained catastrophically high even in the highest group. Each additional household member increased the odds by 18% (adjusted OR 1.18, 95% CI 1.05 to 1.33, p=0.006). Distance to the nearest health facility greater than 5 km was marginally associated (adjusted OR 1.52, 95% CI 0.97 to 2.38, p=0.068). On the other hand, the existing health insurance coverage showed no statistically significant protective effect (adjusted OR 0.82, 95% CI 0.43 to 1.57, p=0.553). Among the 68 insured households, 60 (87.9%) still experienced catastrophic expenditure. This non-protective finding reflects that the Sehat Sahulat Programme is not formally implemented by the Sindh provincial government in Karachi, and insurance coverage among the sample largely reflects either private sector policies with limited benefit packages or Sehat Sahulat registration through other provinces, with constrained access to empanelled facilities. Education level (adjusted OR 0.88 per level increase, p=0.163) and savings (adjusted OR 0.68, p=0.090) showed marginally protective but non-significant associations (table 3). Financial coping mechanisms Among the 1111 households experiencing CHE, most of the coping strategies employed were impoverishing. The most frequently reported mechanism was borrowing from family or friends (455 households, 41.0%), followed by borrowing from informal moneylenders (387 households, 34.8%). Asset sales occurred in 226 households (20.3%), including productive assets such as livestock, jewellery serving as savings stores and income-generating equipment. Only 197 households (17.7%) used existing savings, a figure consistent with the finding that 83.7% of all surveyed households lacked any savings whatsoever. The dependence on informal borrowing and asset liquidation points to a financial coping cascade in which households exhaust increasingly damaging strategies in sequence, a pattern resulting in long-term poverty traps.3640–43 Health expenditure-induced impoverishment Beyond the catastrophic financial burden, health expenditures directly contributed to the impoverishment of substantial numbers of households. Among the 1211 surveyed, 342 (28.2%, 95% CI 25.7% to 30.9%) were classified as newly impoverished, that is, non-poor before health payments but falling below the poverty line of PKR 3705 per capita per month after subtracting out-of-pocket health spending. An additional 487 households (40.2%, 95% CI 37.5% to 43.0%) were already poor pre-payment but became further impoverished. Cumulatively, 829 households (68.5%) experienced health expenditure-induced poverty worsening. The median impoverishment gap among newly impoverished households was PKR 1847 per capita per month, equivalent to 49.8% of the poverty line. Households with chronic illness or recent hospitalisation experienced impoverishment rates of 35.4% and 41.7%, respectively. Middle-income households showed the highest impoverishment rate at 32.6%, suggesting that the financial vulnerability extends beyond the conventionally targeted poorest stratum (table 7). Table 7Health expenditure-induced impoverishment analysis (N=1211)Impoverishment categoryn%95% CINon-poor (pre and post health payment)38231.528.9 to 34.2Newly impoverished34228.225.7 to 30.9Already poor, further impoverished48740.237.5 to 43.0Already poor, not further impoverished–––Total experiencing poverty worsening82968.565.8 to 71.0Mean impoverishment gap (newly impoverished)PKR 1847/capita/month49.8% of poverty line–Mean additional gap (further impoverished)PKR 1235/capita/month33.3% of poverty line–Poverty line=PKR 3705 per capita per month (Pakistan Bureau of Statistics 2023, equivalent to US$13.00). Impoverishment gap=mean shortfall below poverty line among affected households, expressed as absolute amount and percentage of poverty line.PKR, Pakistani rupees. Health expenditure-induced impoverishment analysis (N=1211) Poverty line=PKR 3705 per capita per month (Pakistan Bureau of Statistics 2023, equivalent to US$13.00). Impoverishment gap=mean shortfall below poverty line among affected households, expressed as absolute amount and percentage of poverty line. PKR, Pakistani rupees. Discussion Principal findings This study found that CHE affects 91.7% of urban poor households in Karachi, a prevalence that substantially exceeds documented rates in comparable South and Southeast Asian settings. Three findings together define the nature of the problem. First, the near-zero concentration index (−0.048) indicates that catastrophic spending is distributed almost uniformly across income groups within this population, in contrast to the pro-poor gradient commonly observed in regional comparators. Second, existing health insurance and health coverage programmes provide no measurable financial protection, with insured households experiencing catastrophic expenditure at nearly the same rate as uninsured households. Third, the financial coping strategies deployed by affected households, dominated by informal borrowing (41.0%) and productive asset sales (20.3%), indicate that catastrophic spending is not absorbed through buffers but transmitted into deeper poverty, with 28.2% of households pushed below the national poverty line by health payments alone. Taken together, these findings point to a systemic financial protection failure rather than a coverage gap amenable to incremental programme expansion. Comparison with existing evidence Pakistan’s catastrophic expenditure prevalence among the urban poor stands out against regional benchmarks. Indonesia, which undertook comprehensive insurance expansion between 2014 and 2019, has brought its overall catastrophic expenditure rate to approximately 15.6%.14 Thailand, following the introduction of universal coverage in 2002, reports rates around 11.4%.1344 Even Bangladesh, which shares many structural similarities with Pakistan, including high informality, weak governance infrastructure, and low insurance coverage, reports a national catastrophic expenditure rate of approximately 25%.1345 Afghanistan, another regional comparator with weak governance and low insurance coverage, has been documented to face similar catastrophic expenditure burdens in provincial studies.33 India’s national average sits at around 17%, though urban poor populations in some Indian states experience higher rates.2839–41 The magnitude of the gap between Pakistan’s urban poor and these regional figures cannot be explained by differences in measurement methodology alone, given that all employ WHO-standardised thresholds and capacity-to-pay definitions. Rather, the gap reflects structural features of Pakistan’s health financing architecture that amplify financial vulnerability beyond what poverty alone would predict. Structural determinants of financial protection failure Four interconnected structural features help explain why Pakistan’s urban poor face such extreme health financing burdens. First, government health expenditure remains at approximately 1.2% of gross domestic product, among the lowest in the region and well below the 5% threshold recommended by the WHO for achieving meaningful progress towards universal coverage.3146–48 This funding shortfall translates into chronic deficits in essential medicines, diagnostic capacity, and healthcare workforce at public facilities, which in turn drives poor households towards higher-cost private providers. Analysis of Pakistan’s National Health Accounts indicates that outpatient service providers in the private sector account for 62.9% of overall health spending, with hospitals accounting for a further 31.2%, making outpatient care and associated medication costs the dominant driver of out-of-pocket household spending.519 Second, the Sehat Sahulat Programme, Pakistan’s primary social health protection scheme, is designed around hospitalisation coverage and does not extend to outpatient consultations, medications and diagnostic tests that constitute the majority of healthcare encounters in this population. Even where the programme operates, co-payments and provider balance billing erode its protective effect. Critically, the programme is not implemented by the Sindh provincial government, leaving Karachi’s urban poor largely outside its coverage envelope. The 5.6% insurance coverage rate found in our sample, consisting primarily of private sector policies or cross-provincial Sehat Sahulat registrations, reflects this structural gap.49 Third, the unregulated private sector operates without standardised pricing or quality assurance mechanisms, meaning that poor households seeking care face unpredictable and often substantial costs without the information needed to make informed choices. Fourth, the 18th Constitutional Amendment devolved health to the provinces without establishing mechanisms for federal coordination of health financing standards, creating fragmentation that particularly disadvantages mobile urban populations such as internal migrants in Karachi who may retain provincial registrations from their areas of origin.811 Coping mechanisms and poverty dynamics The coping mechanism data reveals a financial distress cascade with implications that extend well beyond the immediate health expenditure episode. When 41.0% of affected households borrow from family networks and 34.8% turn to informal moneylenders, often at interest rates that far exceed formal lending, a single catastrophic health event generates downstream debt burdens that compound over time. Estimated 20.3% of households selling productive assets, including livestock and income-generating equipment, are depleting the very resources that might otherwise enable economic recovery. Wagstaff and colleagues, in their 122-country analysis, documented the impoverishing effect of health payments but noted that the depth of impoverishment varies substantially with the coping resources available.41 Our data show that 83.7% of Karachi’s urban poor lack any savings whatsoever, which compresses the coping sequence and accelerates the transition from catastrophic expenditure to poverty deepening. The 28.2% newly impoverished rate found in this study is not merely a financial statistic, it represents households whose capacity for future health-seeking, educational investment and economic participation has been structurally compromised. Implications for health financing theory These findings carry implications for two areas of health financing theory. Classical insurance demand theory, following Arrow and Pauly, assumes that risk-averse individuals will purchase actuarially fair insurance in environments with reliable contract enforcement.5051 In Karachi’s urban poor population, this prediction fails as the insurance status does not correlate with reduced catastrophic expenditure risk, suggesting that available insurance products do not or moderately provide the financial protection that the theory assumes. The finding does not necessarily invalidate the theory but rather identifies a boundary condition under which it may not hold, specifically in settings where benefit packages exclude the predominant cost categories and institutional trust is low. Another theory implication is that the traditional targeting assumes that the most vulnerable groups can be identified and therefore should receive preferential treatment. The near-zero concentration index challenges income-based targeting theory, which highlights the means-tested approach used in many social protection programmes.52 If catastrophic burden affects 86%–95% of households across all income tertiles within the urban poor and middle-income population, then income-based targeting excludes households with essentially identical vulnerability. The data support a universal approach to financial protection in settings where the concentration index approaches zero, consistent with the arguments made by Reich and colleagues and Rizvi and colleagues regarding the political and practical advantages of universalism over targeting.5253 Policy implications The evidence from this study points towards three priority areas for reform. Priority 1 recommends that the government develop a government-financed universal healthcare system, similar to Thailand, requiring allocation of approximately 3%–4% of gross domestic product to health, up from the current 1.2%. This fiscal expansion would provide the foundation for extending universal health insurance to all citizens.1354 Priority 2 proposes implementing comprehensive primary care and essential medicine coverage under the Sehat Sahulat Programme benefit package. Since outpatient service providers absorb approximately 63% of private sector health spending in Pakistan and hospitals account for a further 31%, extending coverage to outpatient care would address the dominant cost categories driving catastrophic expenditure. Evidence from Brazil’s Rio de Janeiro urban poor cohort demonstrates that primary healthcare expansion is associated with reduced emergency hospital admissions among vulnerable populations, supporting the phased-in development of outpatient services.5384355 Priority 3 advocates for developing a National Health Insurance Authority to develop a mechanism for the federal government to coordinate and address the geographic disparities created by the fragmentation of health services as a result of the 18th Constitutional Amendment.3356–58 The theoretical frameworks developed for Pakistan may also apply to settings characterised by weak government institutions, fragmented federal systems, and large informal sectors. In countries that are similar to Nigeria and Ethiopia (ie, have weak institutional capacity), effective purchasing functions will likely degrade absent appropriate coordination. Community-based insurance systems are most applicable in low-institutional trust environments (eg, Bangladesh and Yemen). Restoration of trust and the quality of purchasing functions, simultaneously, will also be required in countries that have stalled universal health coverage efforts. Transferability will highlight the relevant factors, but not necessarily which policy mechanisms should be used. Strengths and limitations This study offers methodological strengths including the probability sampling ensuring population representativeness, WHO-standardised catastrophic expenditure measurement enabling international comparison, comprehensive coverage across all six Karachi districts, high response rate (94.7%) and rigorous quality assurance procedures. However, important limitations must also be considered in future research. Cross-sectional design captures expenditure burden at single time points rather than tracking longitudinal poverty trajectories. Four-week recall periods, while standard for LMIC contexts,1920 may underestimate chronic disease expenditures accumulating gradually. The sample restricted to Karachi limits generalisability to rural populations, though Karachi’s status as Pakistan’s largest metropolis makes urban poor experiences particularly policy-relevant. Conclusions CHE affects 91.7% of Karachi’s urban poor households, substantially exceeding global and regional comparators. The near-universal burden across income tertiles (with no meaningful protective gradient), the 28.2% impoverishment rate, and the absence of measurable protection from existing insurance coverage together demonstrate a systemic financial protection failure that requires comprehensive reform. The structural mismatch between hospitalisation-focused insurance design and the outpatient and medication costs driving catastrophic expenditure identifies a specific, addressable reform target for the Sehat Sahulat Programme and its provincial counterparts. The human cost evident in the coping mechanisms documented in this study, with households borrowing at high interest rates, selling productive assets and falling below the poverty line because of illness, demands a policy response that extends beyond incremental programme adjustments towards systemic reform of Pakistan’s health financing architecture, including expanded benefit packages, increased government health expenditure and federal coordination across provinces. Future research should track longitudinal poverty trajectories following catastrophic health expenditure episodes and evaluate the impact of outpatient coverage expansion on financial protection outcomes. Supplementary material