Cost-to-baby as a percentage of median household income predicts population-level access to ART: a global health economics analysis

Infertility affects approximately one in six people of reproductive age worldwide, with comparable prevalence across high-income and low-and-middle-income settings ( WHO Infertility Prevalence Estimates 1990–2021, 2023 ). Infertility carries substantial psychological, social, and economic consequences for affected individuals and communities ( Inhorn and Patrizio, 2015 ) ART, principally IVF, represents the most effective treatment for many forms of infertility, and global ART activity has expanded considerably over the past two decades. The most recent International Committee for Monitoring Assisted Reproductive Technologies world report documented approximately 3.3 million treatment cycles performed across 83 countries in 2017–2018 ( Baker et al. , 2025 ). Yet this impressive aggregate figure conceals profound inequities in access. ART utilization varies by more than an order of magnitude across countries of comparable economic development. Within Europe alone, treatment rates range from 142 to 3230 cycles per million inhabitants, and when expressed per million women of reproductive age, the spread widens further, from 753 (Malta) to 18 715 (Greece) ( Smeenk et al. , 2025 ). Denmark, Belgium, and Israel consistently achieve the highest utilization globally, while countries including the USA, despite substantial clinical capacity, report rates well below estimated needs, resulting in an underserved population ( Chambers et al. , 2009 ; Peipert et al. , 2023 ). Already in 2007, Boivin et al. (2007) estimated that among 72.4 million infertile women worldwide, 56% sought medical help but only 22.4% received treatment, a disparity that Dyer et al. (2020) subsequently proposed quantifying through national ART utilization rates as a standardized indicator of access to infertility care. The International Federation of Fertility Societies (IFFS) surveillance programme has documented across successive editions that fewer than half of countries offering ART provide any public financial support, and only approximately one in five offers full reimbursement ( International Federation of Fertility Societies’ Surveillance (IFFS), 2019 , 2022 ). These data collectively describe a persistent global access gap of extraordinary scale. Among the multiple factors contributing to this gap, including regulatory restrictions, cultural attitudes, clinical infrastructure, and physician availability; cost is consistently identified as the single most important barrier. Adamson (2009) identified economic factors and the absence of public funding as the principal determinants of international variation in ART access. Connolly et al. (2010) , in a review for the ESHRE Reproduction and Society Task Force, concluded that only countries with funding arrangements that minimize out-of-pocket expenses approach estimated demand. In low-and-middle-income countries, the barrier is particularly stark: Njagi et al. (2023) found that a single IVF cycle frequently exceeds annual average income, constituting catastrophic health expenditure by standard definitions, while Chiware et al. (2021) reported ART services entirely absent or functionally inaccessible in many low- and middle-income countries due to prohibitive costs. The most recent IFFS consensus document reinforces the position plainly: affordability is the major barrier to access ( Fauser et al. , 2024 ). The foundational cross-national analysis in this area was performed by Chambers et al. (2009) , who in 2009 compared IVF costs across six developed countries and found that a single cycle consumed ∼50% of annual disposable income in the USA but only 6% in Australia, a distinction driven largely by funding structures rather than raw clinic prices. This work was extended by the same group in 2014 by assessing national ART utilization against affordability defined as cycle cost expressed as a percentage of gross national disposable income per capita and demonstrated a statistically significant affordability gradient across countries ( Chambers et al. , 2014 ). That analysis, now over a decade old, remains the closest precedent to a formal quantitative model linking affordability to utilization at the population level. Several important limitations characterize the existing evidence base. First, prior analyses have relied on per-cycle cost as the measure of economic burden, yet patients typically require multiple cycles to achieve a live birth. Cumulative live birth rates are substantially higher than per-cycle rates. For example, Smith et al. (2015) demonstrated that first-cycle live birth rate in the UK was 29.5%, rising to 65.3% after six complete cycles, meaning the true cost to achieve a baby is two to three times higher than single-cycle cost, depending on patient age and prognosis. Keller and Chambers (2022) argued that cost-per-live-birth is the appropriate outcome metric for economic evaluation of fertility treatments, superseding both per-cycle cost and conventional Quality-Adjusted Life Year thresholds. Despite this, no cross-national affordability analysis has adopted cost-to-baby as the independent variable. Second, existing studies have normalized treatment costs to disposable income per capita or gross domestic product, neither of which directly captures the economic capacity of the typical household. Median household income (HHI), being less distorted by income inequality than per-capita averages, more accurately reflects the affordability pressure experienced by the relevant decision-making unit. Third, and perhaps most critically, no study has tested the functional form of the affordability–utilization relationship. In 2014, Chambers et al. (2014) employed ordinary least-squares regression implying a linear relationship; however, given that ART utilization spans more than an order of magnitude across countries, a power-law specification may provide a more parsimonious and empirically accurate description of the data. Finally, the distinction between gross cost-to-baby (the full fee-for-service cost of achieving a live birth) and net out-of-pocket cost-to-baby (after insurance, subsidies, and tax benefits) has not been formally examined, despite the fact that two countries with identical gross costs may exhibit vastly different utilization rates if one subsidizes treatment heavily. The present study aims to quantify the relationship between ART affordability defined as both gross and net cost-to-baby expressed as a percentage of median HHI and population-level ART utilization across 22 countries and regions representing more than 95% of global ART cycles. We hypothesize that this relationship follows a power-law function, implying constant elasticity of utilization with respect to affordability across the observed range. As a secondary aim, we seek to identify affordability thresholds associated with high and low utilization tiers, offering a data-driven framework for policymakers evaluating the likely impact of financial interventions on ART access within their populations.

Gross cost-to-baby varied more than 12-fold across the 22 countries and regions studied, ranging from 66% of median HHI in Israel to 833% in Africa (excluding Egypt, Tunisia, and South Africa). Net out-of-pocket cost-to-baby showed a similar spread, from 13% of median HHI in Israel to 825% in the same African grouping. The gap between gross and net costs was substantial in countries with comprehensive public funding or insurance coverage (e.g. Israel, Japan, Australia) and negligible in regions where patients bear the full fee-for-service cost. Table 1 presents gross and net cost-to-baby (absolute and as % of median HHI), ART births as a percentage of total births, and ART births per 1000 women aged 15–49 years for all 22 countries and regions. Power-law regression revealed strong inverse relationships between affordability and ART utilization. The strongest associations were observed between gross cost-to-baby and ART births per 1000 women of reproductive age ( R 2 = 0.84) and between net out-of-pocket cost-to-baby and the same absolute utilization measure ( R 2 = 0.84). Both gross and net cost-to-baby also showed strong predictive value for the relative proportion of total births achieved via ART ( R 2 = 0.83 and R 2 = 0.77, respectively). In all four models, the regression slope in log-transformed space was negative, confirming that higher cost burden is associated with lower utilization across the full range of observed values. The analysis is reported in Fig. 1 . Power-law regression analysis for all reproductive age women (15–49). There was a strong inverse relationship between affordability and ART utilization for reproductive age women 15–49. This was evident when considering either gross or estimated net cost-to-baby. ( A ) Proportion of ART births versus Gross cost-to-baby as % of median HHI. ( B ) Proportion of ART births versus Net cost-to-baby as % of median HHI. ( C ) ART Babies per 1000 Females (15–49) versus Gross cost-to-baby as % of median HHI. ( D ) ART Babies per 1000 Females (15–49) versus Net cost-to-baby as % of median HHI. Blue continuous line = power-law regression line. Gold dotted line = 95th percentile confidence band. Red dotted line = 5th percentile confidence band. HHI, household income. The power-law relationship was preserved in women aged 38–49 years, with R 2 values ranging from 0.76 to 0.85 across the four model specifications. The cost-to-baby burden in this age group was ∼1.5 to 2 times higher than in the overall study population, reflecting the lower cumulative live birth rates and greater number of cycles required to achieve a live birth in older women. The affordability–utilization gradient of the older sub-group was of comparable magnitude to the observed on the overall (all-age) population, indicating that the relationship is robust across the reproductive age spectrum. The analysis is reported in Fig. 2 . Power-law regression analysis for women aged 38–49 years. The inverse relationship between affordability and ART utilization is maintained. ( A ) Proportion of ART births versus Gross cost-to-baby as % of median HHI for women aged 38–49 years. ( B ) Proportion of ART births versus Net cost-to-baby as % of median HHI for women aged 38–49 years. ( C ) ART Babies per 1000 Females (38–49) versus Gross cost-to-baby as % of median HHI. ( D ) ART Babies per 1000 Females (38–49) versus Net cost-to-baby as % of median HHI. HHI, household income. Three utilization tiers emerged from the observed clustering of countries along the affordability–utilization gradient. These tiers are descriptive groupings along a continuous affordability–utilization gradient rather than discrete categories with sharp boundaries: there is no distinct interface between high, moderate, and low utilization, and the bands therefore overlap, with the moderate-access zone spanning 80–120% of median HHI. Regions where gross cost-to-baby fell below 100% of median HHI and net out-of-pocket cost-to-baby fell below 50% of median HHI achieved the highest ART utilization. Israel reported 8.7% of total births achieved via ART and 7.0 ART births per 1000 women of reproductive age; Japan 9.3% and 3.05; Spain 11.7% and 3.76; and South Korea 11.8% and 2.45. Moderate-access regions, typified by gross costs of 80–120% and net costs of 40–75% of median HHI, included Australia/New Zealand (5.6% of births via ART; 2.70 per 1000 women) and Western Europe (3.9%; 1.82). Taiwan, China was a notable outlier in this tier, achieving high-tier utilization (11.9%; 3.10) despite net costs in the moderate range, suggesting that additional policy or cultural factors may amplify utilization beyond what affordability alone would predict. In low-access regions, where gross costs exceeded 120% of median HHI and net costs exceeded 75%, utilization was markedly lower. The USA reported 2.5% of all births via ART (0.97 ART births per 1000 women of reproductive age), and Latin American countries fell below 1% of all births via ART (∼0.15–0.20 ART births per 1000 women). The power-law specification yields constant-elasticity estimates of the affordability–utilization relationship. For net out-of-pocket cost-to-baby, the model implies that a 10% reduction in patient cost burden is associated with ∼16–17% higher utilization (measured as ART births per 1000 women of reproductive age), and that halving the cost burden is associated with a 2.67-fold increase in absolute utilization. The corresponding estimates for gross cost-to-baby are larger in magnitude, reflecting the steeper slope: a 10% reduction in gross cost is associated with an ∼28% increase in absolute utilization, and a 50% reduction predicts a 5.12-fold increase. The full elasticity estimates for all four model specifications are presented in Table 2 . Elasticity estimates. These estimates were derived from the observed cross-sectional relationship and should not be interpreted as causal predictions of the effect of a specific policy intervention. Using the power-law models, we calculated the reduction in gross and net cost-to-baby required for each country/region to reach 90th-percentile (best-in-class) utilization levels, defined as 11.5% of total births achieved via ART and 3.1 ART births per 1000 women of reproductive age. These country-specific estimates are presented in Table 3 . As an example, shifting the USA from its current absolute utilization of 0.97 ART births per 1000 women to the 90th-percentile target of 3.1 would require reducing gross cost-to-baby by 30%, from 123% to 86% of median HHI. The corresponding model-implied net out-of-pocket cost-to-baby target for the USA is 34% of median HHI. Reaching this target would require action on both cost components: a reduction in gross cost-to-baby (a supply-side efficiency gain) would need to be accompanied by an expansion of insurance and subsidy coverage to bring net out-of-pocket cost-to-baby sufficiently down (a demand-side intervention). Country-specific affordability targets to reach 90th-percentile ART utilization (≥3.1 ART births per 1000 women of reproductive age). The table presents the required % reduction in gross cost-to-baby, and then the model-implied net out-of-pocket cost as a percentage of the reduced (target) gross cost. The last column expresses the same net target as a percentage of current gross cost-to-baby, capturing the combined effect of supply-side cost reduction and demand-side coverage expansion. Excluding Egypt, Tunisia, South Africa. Excluding Mexico and Brazil. Excluding Turkey, Israel, Egypt. Excluding Spain, UK.

This is a cross-national observational study comparing country/region-level ART affordability metrics against population-level utilization outcomes. ART cycle volumes and live birth outcomes were extracted from the International Committee for Monitoring Assisted Reproductive Technologies (ICMART) world reports ( Chambers et al. , 2021 ; Baker et al. , 2025 ), the European Society of Human Reproduction and Embryology (ESHRE) ( Smeenk et al. , 2025 ), the United States Center for Disease Control and Prevention National ART Surveillance System (CDC NASS) ( CDC, 2022 ), the Canadian Assisted Reproductive Technologies Register (CARTR) ( CARTR-Plus & BORN Ontario, 2025 ), UK’s Human Fertilisation and Embryology Authority (HFEA) ( HFEA, 2025 ), the Latin American Network of Assisted Reproduction (REDLARA) ( Zegers-Hochschild et al. , 2025 ), the Australian and New Zealand Assisted Reproduction Database (ANZARD) ( Kotevski et al ., 2023 ), and national registries where available (see Table 1 ), covering reporting years 2021–2023. The dataset encompasses 22 countries and regions representing more than 95% of estimated global ART cycle volume. Gross and net cost-to-baby (absolute and as % of median HHI), ART births as a percentage of total births, and ART births per 1000 women aged 15–49 years. Archary et al. (2024) Dyer et al. (2026) Hu et al. (2020) Suzhou (2025) Archary et al. (2024) Dyer et al. (2026) Archary et al. (2024) Dyer et al. (2026) Archary et al. (2024) Dyer et al. (2026) Excluding Egypt, Tunisia, South Africa. Excluding Mexico and Brazil. Excluding Turkey, Israel, Egypt. Excluding Spain, UK. Data Quality: High = government, ART/Healthcare regulatory authorities and professional societies and 10 or more published 3rd party and clinic fee schedules sampled in country/region. Mod = Moderate. Grey literature and 3—5 published 3rd party and clinic fee schedules sampled in country/region. Estim = Estimate. Expert opinion based on relative comparison to similar countries with higher data quality. Source details are offered in Supplementary Tables S2 and S4 . HHI, Household income; USD, US dollars. Countries with sufficiently complete data from a dedicated national registry were analysed individually, whereas countries with merged, sparse or heterogeneous data were aggregated into regional groupings (Australia and New Zealand, Western Europe, Eastern Europe, the Middle East, South-East Asia, the rest of Latin America, and the rest of Africa). The full composition of each regional grouping is given in Supplementary Table S1 . ICMART estimation methods for total cycle volumes and live births, including extrapolation from participating to non-participating centres and confidence categorization based on cycle participation rates, were applied as described in the ICMART 2014 and ICMART 2017–2018 world reports ( Chambers et al. , 2021 ; Baker et al. , 2025 ). Countries with fewer than 33% of cycles reported to ICMART and without an informed regional estimate were not included in utilization calculations. Median after-tax disposable HHI was obtained from the Organisation for Economic Co-operation and Development (OECD) ( OECD, 2026 ) and the World Bank (2021) . In cases where only average HHI was available, the typical ratio of median to average (70%) from comparable countries was used to estimate the median HHI. Female population aged 15–49 years and total live births were sourced from the United Nations Population Division ( United Nations, Department of Economic and Social Affairs, Population Division, 2024 ). Gross cost-to-baby was defined as the mean total fee-for-service cost of achieving one ART live birth, before any insurance reimbursement, government subsidy, or tax benefit. It was calculated as the product of mean cost per IVF cycle (from published clinic fee surveys and national fee schedules within each country/region), including embryo transfer, preimplantation genetic testing (PGT) (adjusted for usage rate in the given country/region), and medication costs, and the weighted mean number of cycles required to achieve a live birth, adjusted for regional maternal age distribution. Age adjustment was performed by applying age-specific cumulative live birth rates derived from registry data to the maternal age distribution observed in each country/region, yielding an age-weighted cycles-to-baby estimate. Net out-of-pocket cost-to-baby was defined as the residual cost borne by the patient after accounting for private and government insurance coverage, direct subsidies, and applicable tax deductions or credits. For countries with published government reimbursement schedules including Japan (National Health Insurance), Australia (Medicare Benefits Schedule/Pharmaceutical Benefits Scheme), Israel (Health Basket), South Korea (National Health Insurance Service), and several European national health systems, subsidy values were derived directly from official sources. For countries where comprehensive published data were not available, net costs were estimated through structured triangulation of published literature against estimates obtained from grey literature, comparable countries, expert consultation, and interrogating language models GPT (version 4.0, OpenAI, San Francisco, CA, USA), Claude (version 4.5 Opus, Anthropic, San Francisco, CA, USA), and Gemini (version 1.5Pro, Google, Mountain View, CA, USA) to discover and synthetize information sources; these estimates are flagged in the results. The underlying data and data sources are presented in Supplementary Tables S2, S3, S4 , and S5 . Both gross and net cost-to-baby were expressed as a percentage of median HHI in the corresponding country/region. Treatment costs and HHI were expressed in nominal local-currency terms. Purchasing power parity adjustment was not applied because the affordability metric is constructed as a within-country ratio of local treatment cost to local HHI; it inherently reflects the economic reality faced by the household contemplating treatment. Such conversion would re-weight income against a standardized international consumption basket that does not correspond to the actual cost structure of fertility care. ART utilization was assessed using two complementary measures. The relative measure was defined as the percentage of total national live births achieved via ART, capturing the technology’s contribution to the overall birth landscape. The absolute measure was defined as the number of ART-conceived live births per 1000 women of reproductive age (15–49 years), providing a population-standardized rate that is not confounded by differences in national birth rates. Both measures were calculated from the same reporting period as the cost data. The relationship between affordability (cost-to-baby as percentage of median HHI) and utilization was assessed using power-law regression. This specification was selected on the basis that ART utilization spans more than an order of magnitude across the countries studied, suggesting a multiplicative rather than additive relationship: a constant proportional change in affordability produces a constant proportional change in utilization. In log-transformed space, this corresponds to a linear model of the form log(utilization) = α + β.log(affordability), where the slope coefficient β represents the elasticity of utilization with respect to affordability. Log-linear models were fitted via ordinary least squares on transformed variables; the analyses were performed in Excel (version 16.0, Microsoft 365 MSO, Microsoft, Redmond, WA, USA). Pearson correlation coefficients were calculated from natural-log-transformed variables. Model fit was assessed by R 2 values. Alternative functional forms were evaluated. Linear regression on untransformed variables yielded consistently lower R 2 values across all four model specifications, confirming that the power-law (log–log) specification provides superior fit to the observed data. A subgroup analysis was performed for women aged 38–49 years to evaluate whether the affordability–utilization relationship is amplified in an age group where cumulative live birth rates are lower and the number of cycles required to achieve a live birth and therefore cost-to-baby is substantially higher. To quantify the impact of uncertainty in input estimates on the primary regression findings, a Monte Carlo probabilistic sensitivity analysis was performed in Microsoft Excel, following established practice in health economic evaluation ( Claxton et al. , 2005 ; Briggs et al. , 2006 ; Cassettari et al. , 2016 ). Four key input variables for each of the 22 countries and regions: mean gross IVF cycle cost (inclusive of one frozen embryo transfer, PGT weighted by national usage rate, and medications) mean number of cycles to live birth median HHI out-of-pocket cost as a percentage of gross cost-to-baby were each assigned a PERT (Program Evaluation and Review Technique) probability distribution. The PERT distribution, a four-parameter beta distribution defined by minimum, mode, and maximum values ( Vose, 2008 ), was selected because it accommodates asymmetric uncertainty around a central estimate while producing a smooth, unimodal curve that concentrates probability near the most likely value, properties well suited to expert-derived cost and utilization inputs where full empirical distributions are unavailable. For each variable, the mode was set to the baseline point estimate. Uncertainty bands were assigned to each country/region based on the quality and provenance of the underlying cost data, using a three-tier classification. High-quality sources (±5% uncertainty band) were defined as government publications, ART or healthcare regulatory authority reports, professional society data, and cost inputs corroborated by 10 or more independently published third-party or clinic fee schedules sampled within the country/region. Moderate-quality sources (±7.5% uncertainty band) comprised a combination of published and grey literature with three to five independently sampled fee schedules available for cross-validation. Estimated sources (±10% uncertainty band) relied on expert opinion informed by relative comparison with economically or structurally similar countries for which higher-quality data were available, supplemented by AI-assisted triangulation using large language models (OpenAI GPT, Anthropic Claude, Google Gemini) to discover and synthetize information sources. The tier assigned to each country/region is reported alongside the corresponding cost inputs in Table 1 . A standard shape parameter of λ = 4 was applied throughout, yielding PERT shape parameters: For each of the four input variables in each of the 22 countries and regions (88 distributions in total), 10 000 independent random values were drawn from the corresponding PERT distribution. Each draw was generated by sampling a uniform random number between 0 and 1 and transforming it through the inverse cumulative distribution function of the beta distribution with the country-specific shape parameters (α 1 , α 2 ), then rescaling to the range defined by the minimum and maximum values ( Vose, 2008 ). In each simulation run, the two derived affordability metrics (gross cost-to-baby as a percentage of median HHI, and net out-of-pocket cost-to-baby as a percentage of median HHI) were recalculated for every country/region. The 5th and 95th percentile values (P5, P95) of each derived affordability metric were then extracted across the 10 000 runs for each country/region. Power-law regressions were fitted to the P5 and P95 affordability values against both utilization measures (percentage of births via ART, and ART births per 1000 women aged 15–49 years), generating 90% simulation envelopes around each of the four baseline regression models. These envelopes are presented graphically alongside the baseline regressions. All analyses were performed in Microsoft Excel.

To our knowledge, this is the first study to quantify the relationship between ART affordability and population-level utilization using a cost-to-baby metric normalized to median HHI, and the first to identify a power-law functional form for this relationship. The consistency of the association across all four model specifications ( R 2  = 0.77–0.84), and across both gross and net affordability metrics, suggests that the cost of achieving a live birth, not the cost of a single cycle, is the economically relevant quantity governing treatment access at the population level. This metric also reflects the decision-making reality of patients, for whom the relevant economic question is not the cost of a single treatment cycle, but the total investment required to achieve a live birth. Our analysis extends previous work ( Chambers et al. , 2009 , 2014 ) in three specific ways: adopting cost-to-baby rather than per-cycle cost as the affordability metric, normalizing to median HHI rather than per-capita disposable income, and testing a power-law functional form that yields constant-elasticity estimates directly comparable to the price-elasticity literature ( Connolly et al. , 2009b ). The result is a predictive model linking the funding landscape documented by the IFFS surveillance programme to the utilization outcomes tracked by ICMART and regional registries, through the mechanism of household-level affordability ( Allan et al. , 2019 ; Horton et al. , 2022 ). The identified thresholds (gross cost-to-baby below 100% of median HHI, net below 50%) are observed characteristics of the highest-utilization countries. That gross and net cost-to-baby independently predict utilization is among the most policy-relevant findings of this analysis, because the two metrics respond to fundamentally different interventions. Gross cost reflects the efficiency of the healthcare delivery system: clinic capacity, workforce models, technology adoption, and economies of scale. Net out-of-pocket cost reflects the policy environment: the extent to which insurance, subsidies, and tax incentives shield patients from the full fee-for-service burden. While affordability is the overall binding constraint on ART access, neither lever alone appears sufficient to increase utilization. A country may subsidize treatment generously, but if gross costs are high, the fiscal burden limits coverage scope or eligibility; conversely, efficient low-cost clinics without public funding still place the full burden on household budgets. The countries achieving the highest utilization (Israel, Japan, Spain, South Korea) have acted on both levers simultaneously. In each high-utilization country, the combination of low gross costs and generous coverage brings net cost-to-baby below 50% of median HHI, crossing the threshold associated with high utilization in our model. The French experience illustrates this asymmetry from the opposite direction. Despite near-complete public reimbursement, France does not achieve top-tier utilization and records among the highest rates of cross-border reproductive care in Europe, an outcome attributed to chronic underfunding of laboratory infrastructure and capacity constraints that effectively inflate the gross cost of treatment delivery ( Chaillot et al. , 2026 ). In the framework of the present analysis, France is a case where the demand-side lever has been maximally deployed while the supply-side lever remains largely unaddressed. Indeed, expanding access to fertility treatment is also a capacity problem. Subsidies and insurance mandates cannot help patients if the industry lacks the infrastructure to treat them ( Sable, 2026 ). Taiwan presents an instructive outlier: despite net costs in the moderate range (40–75% of median HHI), it achieved utilization rates comparable to the highest-access tier (11.9% of births via ART), suggesting that cultural acceptance, regulatory accessibility, and the momentum of the 2021 subsidy expansion can amplify utilization beyond what affordability alone would predict. This does not undermine the model but illustrates that the power-law relationship captures dominant variance while leaving room for country-specific factors to shift. The affordability targets presented here present a rationale for increased public expenditure. Whether such investment is fiscally defensible is addressed by a growing body of accounting analyses across different fiscal regimes, which was pioneered by Connolly (2026) and which project the lifetime fiscal profile of an ART-conceived individual. Across 12 countries studied, 11 show consistently positive returns: from 2.7× in Australia ( Connolly et al. , 2025 ), to 5.7× in Taiwan, China ( Chen et al. , 2025 ), and ∼16× in Spain ( Matorras et al. , 2015 ), with comparable findings in middle-income settings like South Africa, 5.6 × ( Connolly et al. , 2021 ), and earlier estimates for the USA of ∼7 × ( Connolly et al. , 2008 ), UK 8.5 × ( Connolly et al. , 2009a ), Denmark of ∼14 × ( Connolly et al. , 2011 ), and Sweden at 1.2 × ( Svensson et al. , 2008 ). The sole exception, the Netherlands, reflects a welfare system that produces negative fiscal balances for all citizens, not specifically ART-conceived individuals ( Moolenaar et al. , 2014 ). The consistency of positive returns across diverse settings supports the conclusion that ART represents a high-return public health investment in the context of demographic contraction ( Connolly et al. , 2023 ; Fauser et al. , 2024 ). This urgency is underscored by the observation that more than half of the world’s nations already exhibit sub-replacement total fertility rates, with projections suggesting ∼97% will do so by the end of the century ( Bhattacharjee et al. , 2024 ; Aitken, 2026 ). In countries where ART-conceived births already account for 7–10% of all newborns, ART has become de facto demographic infrastructure, yet the affordability barriers identified in the present analysis mean that this capacity remains inaccessible to most. Readers should note that this is a cross-national, cross-sectional observational study based on aggregated, country-level data. The observed associations cannot support individual-level causal inference: relationships that hold across populations need not hold within them (ecological fallacy), and unmeasured country-level factors correlated with both affordability and utilization may contribute to the observed gradient. Accordingly, the elasticity estimates presented in Table 2 describe the observed cross-sectional relationship and should not be interpreted as forecasts of the effect of any specific policy intervention. Registry data are heterogeneous across countries in reporting standards, cycle definitions, and completeness, as documented by ICMART and ESHRE ( Baker et al. , 2025 ; Smeenk et al. , 2025 ). Unmeasured confounders include cultural and religious attitudes, regulatory restrictions on access, clinic capacity constraints, and cross-border reproductive care, which may inflate utilization estimates in destination countries and deflate them in origin countries. The analysis likewise did not capture the broader societal context of family formation, including family-friendly employment practices, parental leave, childcare provision, and the healthcare and educational support available to families which shape household reproductive decision-making independently of the direct cost of treatment. A reinforcement loop, where high utilization drives down costs through economies of scale, represents a potential source of reverse causation that a cross-sectional design cannot resolve. Moreover, the macro-level framework cannot resolve within-country racial or ethnic inequities in ART access, and this represents an important direction for future research. Transparency regarding the provenance and quality of input data is also essential to interpreting these findings. ART utilization data derive from established international and national registries with documented quality assurance processes, and economic and demographic data from authoritative agencies (OECD, World Bank, UN). Gross cost inputs were drawn from published surveys and national fee schedules for the majority of countries. However, for a subset of countries and regional aggregates where comprehensive published data were not available from a single authoritative source, cost estimates—particularly net out-of-pocket costs—were derived through structured triangulation of expert consultation, grey literature, and AI-assisted synthesis using large language models (OpenAI GPT-4o, Anthropic Claude, Google Gemini Pro). These estimates are identified in the results section and supplementary tables and should be interpreted with appropriate caution. This heterogeneity in input quality is the principal reason we performed Monte Carlo probabilistic sensitivity analysis with uncertainty bands explicitly calibrated to data provenance: ±5% for countries with well-documented published sources, ±7.5% for those with moderate-quality evidence, and ±10% for countries relying on estimated inputs. The assignment of countries to these tiers involved a semi-qualitative interpretation, and the specific band widths are not empirically derived; the three-tier scheme may understate uncertainty for the weakest sources and overstate it for the strongest, but it represents a structured approach to differential uncertainty that avoids embedding additional unverifiable assumptions. The critical finding is that the power-law relationship and the associated R 2 values are not sensitive to plausible variation in any single country’s input estimates: the 90% simulation envelopes remain narrow across all four model specifications, indicating that the affordability–utilization gradient is a robust structural feature of the data and is not dependent on any individual point estimate. Two further caveats apply. The simulation evaluates uncertainty in the independent variables (affordability metrics) while holding the dependent variables (utilization outcomes) fixed; utilization data also carry measurement uncertainty, and the reported envelopes should therefore be interpreted as sensitivity bands on the affordability estimates rather than full probabilistic confidence intervals on the fitted relationship. The practical implication is that even if specific country-level cost estimates carry meaningful uncertainty the overall relationship and the policy-relevant thresholds derived from it are stable. Nevertheless, we call for the development of a standardized international ART affordability monitoring framework, analogous to the ICMART cycle reporting infrastructure, to replace estimated inputs with verified data over time.

Affordability measured as the cost of achieving an ART live birth relative to median HHI follows a power-law relationship with population-level utilization. The metric proposed here offers a standardized framework for benchmarking fertility care access globally. Two complementary economic levers can increase utilization: reducing gross cost-to-baby below 100% of median HHI through provider-side efficiencies and reducing net out-of-pocket cost-to-baby below 50% of median HHI through coverage expansion, subsidies, or tax incentives. Countries that have acted on both levers simultaneously achieve the highest utilization globally. The power-law nature of the relationship implies that the marginal return on affordability improvements is greatest where costs are currently highest relative to income.