Reproductive healthcare use before and after infertility diagnosis: a population-based study in South Korea.

Infertility, defined as the inability to achieve pregnancy after 12 months of regular unprotected intercourse, is recognised as a major public health concern. 1 2 Recent global estimates from the WHO indicate that approximately 17.5% of adults experience infertility over their lifetimes, with prevalences of 17.8% in high-income countries and 16.5% in low and middle-income countries. 2 Biomedical research has identified specific pathological risk factors and disease entities, such as endometriosis, 3 polycystic ovary syndrome 4 and tubal occlusion 5 as well as metabolic abnormalities that influence assisted reproductive technology (ART) outcomes. 6 Alongside these biomedical perspectives, a growing body of research has drawn attention to social and structural factors associated with infertility and its consequences, including intimate-partner violence, 7 socioeconomic status 8 and broader contexts related to reproductive health. 9 10 In parallel, accumulated studies have emphasised infertility as a condition that is clinically recognised, socially experienced and increasingly incorporated into policy responses. 11 13 Research on reproductive stratification and reproductive justice demonstrates how age, class, gender, race/ethnicity and region shape experiences of infertility and access to treatment. 14 16 Qualitative studies further suggest that infertility may become experienced at different points along the life course, shaped by family expectations, institutional arrangements and insurance eligibility criteria, even in the absence of identifiable biomedical causes. 13 14 17 These perspectives have encouraged consideration of infertility as a process unfolding over time rather than solely as a discrete diagnostic event. 18 20 Against this background, policy debates in many settings have increasingly foregrounded ART as a key response to infertility. ART is often presented as a means to alleviate psychological distress, enhance reproductive choice and secure individual freedom, even when the broader social and structural conditions that shape infertility remain unchanged. 21 24 In South Korea, where fertility rates have fallen to below one, the government has progressively expanded financial support for infertility treatment, culminating in the inclusion of ART in the National Health Insurance benefit package and the introduction of targeted financial assistance. 25 Driven by these expanded benefit packages and policy incentives, the annual number of women diagnosed with infertility nationwide remained remarkably stable from 151 489 in 2018 to 149 353 in 2020, showing no substantial decline even during the COVID-19 pandemic, and further rose to 162 938 in 2021. 26 However, research conducted in response to these policy expansions has focused primarily on evaluating the performance of individual institutions or specific ART units. 27 29 In this institutional context, a diagnosis of infertility functions more as a practical criterion for accessing public assistance services than as a medical condition requiring treatment. In other words, formal diagnostic codes are prerequisites for financial assistance, the clinical diagnosis represents an administrative entry point rather than the biological onset of subfertility. Consequently, this pooling mechanism aggregates individuals with highly heterogeneous clinical profiles into a single baseline category. For example, the newly diagnosed cohort simultaneously includes individuals with prior obstetric histories and those with prolonged, unassisted primary infertility. These patterns suggest that initial administrative records reflect strategic adaptation to institutional incentives rather than inappropriate diagnoses at the individual level. 30 31 Despite growing interest in these issues, quantitative evidence has largely relied on self-reported measures of mental health, emotions and quality of life, leaving patterns of healthcare utilisation before and around infertility diagnosis relatively underexamined in population-level data. 32 33 To address this gap, we analyse nationwide healthcare utilisation records from the Korean National Health Insurance Service (NHIS). By utilizing the International Classification of Diseases (ICD) to define the scope of our analysis as a single diagnostic code for female infertility (ICD-10: N97), we aim to minimise the aetiological ambiguity associated with male infertility (N46). This approach aligns with the objective of this study, which is not to establish causal relationships regarding infertility diagnosis and treatment, but rather to examine the overall status of women’s health and healthcare utilisation before and after an infertility diagnosis within the context of expanding treatment and policy support for infertility. We document heterogeneity in prediagnostic disease profiles and healthcare utilisation patterns among women receiving a first infertility diagnosis and examine how sociodemographic characteristics, prior pregnancy-related care and infertility-related interventions are associated with having a pregnancy identified without recorded loss within 1 year after diagnosis. This analysis demonstrates that an infertility diagnosis is part of an ongoing process of healthcare utilisation rather than an isolated episode. Ultimately, these findings redirect attention towards patient-centred care focused on women’s reproductive health.

Table 1 summarises the baseline prevalence of the 20 most common disease groups observed during the year prior to the first infertility diagnosis among Korean women aged 20–49 years in 2019 and 2020, stratified by whether a pregnancy was identified without recorded loss during the 1-year follow-up period. The prevalence of pregnancy-related conditions was substantially higher among women who later had a pregnancy identified without recorded loss, while diseases of the genitourinary system showed minimal variation between groups. Antenatal screening and supervision of pregnancy (Z34–Z36) exhibited the highest overall prevalence at 48.4%, followed by encounters for health services for other reasons (Z31–Z33, Z37, Z55–Z99) at 35.5%. When stratified by follow-up pregnancy status, women who later met the outcome definition had a markedly higher prevalence of antenatal screening (97.2%), whereas among women who did not, the most frequent category was other examinations, including assisted reproductive procedures, at 32.7%. Disease codes were grouped according to the 298-category classification system presented in the Health Insurance Statistics Yearbook, published by the National Health Insurance Service and the Health Insurance Review and Assessment Service in the Republic of Korea. This system was adopted in place of the International Classification of Diseases (ICD) codes because it reflects the standardised grouping scheme used for nationwide statistical reporting and enables consistent identification of the most common disease categories across the entire Korean population. Complications of pregnancy and delivery (O20–O29, O60–O63, O67–O71, O73–O75, O81–O84), maternal care related to the fetus and amniotic cavity and possible delivery problems (O30–O43, O47, O48) and postpartum care and examination (Z39) were recorded in 69.7%, 42.3% and 39.2% of women who subsequently had a pregnancy identified without recorded loss, compared with 4.1%, 1.2% and 0.5% among those who did not meet this outcome definition. Disorders of the genitourinary tract (N82, N84–N90, N93, N94, N96, N98, N99) and disorders of menstruation (N91, N92) accounted for 25.6% and 21.1%, respectively. Inflammatory diseases of female pelvic organs (N71, N73–N77) and inflammatory diseases of the cervix uteri (N72) were less common, with prevalence of 15.7% and 11.3%. Disease groups unrelated to sexual and reproductive health, such as respiratory infections, thyroid disorders and musculoskeletal conditions, showed little difference by follow-up pregnancy status. Table 2 presents differences in demographic characteristics, socioeconomic status, healthcare utilisation patterns and baseline health conditions by pregnancy identified without recorded loss during the 1-year follow-up period. Age emerged as a strong correlate of subsequent reproductive healthcare utilisations, with women younger than 35 years constituting a substantially higher proportion of those with an identified pregnancy within 1 year (63.87%) than those aged 35 and older (36.13%, p<0.001). Healthcare coverage types also differed significantly (p<0.001), with NHI enrollees representing a greater share among those meeting the outcome definition than Medical Aid beneficiaries. ART, assisted reproductive technology. Regarding occupational industry type, women in public administration/social services and manufacturing/logistics had relatively higher proportions meeting the outcome definition, whereas those without formal employment had the lowest. Women who had a pregnancy identified without recorded loss had fewer medical visits after infertility diagnosis (mean: 7.29 vs 10.01 visits, p<0.001), which may reflect changes in healthcare needs or care-seeking behaviour following pregnancy identification, rather than reduced need for care once pregnancy is established. The mean number of financial assistance claims and ART procedures was also lower among women who met the outcome definition (p<0.001). There was no significant difference in prediagnosis healthcare visit frequency between the two groups (p=0.15), suggesting that baseline healthcare engagement alone did not distinguish subsequent reproductive healthcare utilisations. Logistic regression analysis identified several factors associated with pregnancy, without recorded loss, within 1 year after infertility diagnosis ( table 3 ). Receiving financial assistance for infertility treatment was associated with higher odds of meeting this outcome definition (OR=1.60, 95% CI 1.54 to 1.67) in model 1. ART utilisation remained positively associated with pregnancy identified without recorded loss (OR=1.66, 95% CI 1.60 to 1.72) in model 2, while other healthcare utilisation variables showed similar patterns across models. *p<0.05. ART, assisted reproductive technology. A greater number of medical visits after diagnosis was associated with lower odds of a pregnancy being identified without a recorded loss (OR=0.95, 95% CI 0.95 to 0.95 per additional visit). The number of distinct prediagnosis diagnoses was positively associated with the outcome (OR=1.04, 95% CI 1.04 to 1.05), whereas higher prediagnosis visit frequency was associated with lower odds (OR=0.99, 95% CI 0.99 to 0.99). Women diagnosed with infertility in 2020 had higher odds of meeting the outcome definition compared with those diagnosed in 2019 (OR=1.11, 95% CI 1.09 to 1.14). Older age and socioeconomic disadvantage were associated with lower odds of having a pregnancy identified without a recorded loss. Women aged 35 and older had significantly lower odds of having a pregnancy identified without recorded loss compared with those under 35 (OR=0.53, 95% CI 0.52 to 0.55). Medical Aid beneficiaries had lower odds than NHI enrollees (OR=0.63, 95% CI 0.51 to 0.79). Employment in different occupational sectors showed varied associations with pregnancy identified without recorded loss during follow-up. Compared with unemployed women, those employed in public administration or social services (OR=1.39, 95% CI 1.35 to 1.44), finance and technology (OR=1.34, 95% CI 1.29 to 1.40), art and leisure (OR=1.22, 95% CI 1.13 to 1.32), manufacturing and logistics (OR=1.30, 95% CI 1.26 to 1.34) and primary industry (OR=1.46, 95% CI 1.16 to 1.84) all had significantly higher odds of having a pregnancy identified without recorded loss, compared with unemployed women. A higher NHI premium quantile was positively associated with the outcome, with the third quantile (OR=1.21, 95% CI 1.17 to 1.24) exhibiting the highest odds. Women with disabilities had significantly lower odds of having a pregnancy identified without recorded loss (OR=0.70, 95% CI 0.60 to 0.83). Those residing outside the Seoul metropolitan area had higher odds of pregnancy than those living in the capital region (OR=1.04, 95% CI 1.01 to 1.06). Before the infertility diagnosis, many women had O-code diagnoses related to pregnancy, childbirth and the puerperium (O00–O99) as well as antenatal care and pregnancy testing, among the most common disease categories. Using O00–O99 codes to define prediagnostic pregnancy history, 24 122 women had a recorded pregnancy history, and 1 27 448 women had no records. We conducted subgroup analyses stratified by these two groups, and logistic regression was performed within each subgroup ( table 4 ). Among women without a recorded history of pregnancy, ART use was associated with higher odds of having a pregnancy identified without recorded loss within 1 year (OR=1.67, 95% CI 1.61 to 1.74), among those with a prior pregnancy history, ART use also increased the odds of meeting the same outcome definition (OR=1.60, 95% CI 1.46 to 1.75). Other utilisation indicators, including postdiagnosis visit frequency, prediagnosis disease burden and prediagnosis visit frequency, showed similar directions and statistical significance across both subgroups. These subgroup analyses suggest that the associations observed in the overall cohort were robust across heterogeneous reproductive histories. Unlike in the overall population, region did not reach statistical significance in subgroup analyses. *p<0.05. ART, assisted reproductive technology.

This nationwide study analysed prediagnosis health status and healthcare utilisation patterns among women with an infertility diagnosis in South Korea and examined how these were associated with subsequent reproductive healthcare utilisations, including pregnancy identified without recorded loss during the follow-up period. This study provides empirical evidence on how infertility diagnosis is embedded within women’s broader patterns of healthcare use in a universal coverage setting. First, the differences in the past history between women who did and did not have a pregnancy identified without recorded loss during follow-up were more pronounced in pregnancy-related diagnostic codes than in genitourinary disorders. Antenatal screening and supervision of pregnancy (Z34–Z36) and encounters for other reproductive services were more common among women who later met the outcome definition. In contrast, conditions typically classified as underlying causes of infertility showed minimal variation between groups. One possible interpretation of this pattern is that infertility diagnosis may occur at different points along women’s reproductive health histories, including during periods of active pregnancy planning or repeated engagement with reproductive healthcare, rather than marking a uniform transition from health to pathology. This interpretation is consistent with prior studies that have emphasised the temporal and episodic nature of infertility-related care, although alternative explanations—such as differences in care-seeking behaviour or diagnostic practices—cannot be ruled out. 14 17 35 These findings suggest that an infertility diagnosis does not necessarily imply the presence of distinct underlying pathologies among all diagnosed women. 14 19 Second, ART use and financial assistance were associated with higher odds of having a pregnancy identified without recorded loss within 1 year. However, these technological and financial inputs did not fully account for the observed variations in reproductive outcomes. Greater postdiagnostic healthcare utilisation was associated with lower odds of meeting the outcome definition, whereas a larger number of distinct diagnoses prior to diagnosis were associated with higher odds. These patterns may reflect heterogeneity in how women engage with reproductive healthcare, where prior diagnostic intensity could signal both proactive health-seeking and the accumulation of comorbid conditions. Alternatively, greater postdiagnostic utilisation may reflect the fact that treatment has become more complex or prolonged, rather than an improvement in reproductive prospects. Such interpretations are consistent with previous research suggesting that treatment continuity and intensity are shaped not only by clinical indications but also by emotional, financial, and temporal constraints, which may lead to intermittent engagement with care or discontinuation. 12 14 20 Third, gradients by age, socioeconomic status and occupational sector were observed despite universal NHI coverage. Women aged 35 and over and Medical Aid beneficiaries had lower odds of having a pregnancy identified without a recorded loss than their younger and better-resourced counterparts. Employment in more stable and formal sectors was associated with higher odds of meeting the outcome definition compared with non-employment. These findings align with existing evidence from Korea and other settings that working conditions, income and opportunity costs are associated with continuity of infertility-related care, even under conditions of formal financial coverage. 2539 41 They further suggest that formally universal policies may coexist with persistent differences in who is able to translate eligibility into sustained engagement with care and subsequent outcomes. 14 1622 Nevertheless, since our analytic objective remains strictly associative rather than causal, these gradients should be interpreted as system-level reflections of healthcare interaction rather than deterministic predictors of reproductive outcomes. Furthermore, we caution against overgeneralising these cross-sectional patterns into causal mechanisms. Our findings suggest that trends in infertility diagnosis and treatment intensity in South Korea reflect not only underlying reproductive health needs but also how clinical diagnosis is embedded within eligibility criteria for publicly supported care in a government-regulated healthcare system. When access to ART and related financial support is administratively linked to formal diagnosis, women with heterogeneous reproductive histories—including those with prior pregnancies—may enter diagnostic categories at different stages of their reproductive lives. Such diagnostic dynamics may be understood as adaptation to existing institutional arrangements rather than as evidence of inappropriate diagnosis at the individual level. 30 42 This perspective is consistent with the existing literature highlighting how diagnostic practices can evolve in response to policy design and service delivery structures. Several limitations should be considered when interpreting these findings. First, we used a 1-year wash-out period to identify first recorded infertility diagnoses, consistent with the clinical definition of infertility, which may not be enough to rule out recurrent or episodic infertility. The relatively high prevalence of pregnancy-related care prior to diagnosis is compatible with such non-linear reproductive patterns, but alternative explanations cannot be excluded. 14 18 Second, outcomes were assessed within a fixed 1-year observation window and did not capture long-term fertility histories, repeated treatment cycles or definitive live birth outcomes, which warrant further investigation. Third, from an epidemiological perspective, because the defined pregnancy outcome is relatively common within our study cohort (28.69%), the reported multivariable ORs may systematically overestimate the true relative risks. While this mathematical property is inherent to logistic regression profiles applied to common events, it should be taken into account when interpreting the absolute scale of the observed socioeconomic and clinical gradients. Nevertheless, because our analytical focus remains descriptive and associative rather than causal, this overestimation does not alter the statistical significance or the overall direction of the identified phenomenal trends. Fourth, the inclusion of postdiagnostic healthcare utilisation as an independent variable carries a risk of reverse causation, as pregnancy itself may alter the frequency and nature of subsequent healthcare visits. While the 1-year follow-up window was applied to minimise temporal ambiguity, we cannot fully exclude the possibility that postdiagnostic utilisation patterns partly reflect care received after pregnancy was established rather than care predating its identification. This limitation is consistent with our strictly associative analytic objective and underscores the need for future studies using time-to-event or instrumental variable designs. Fifth, the scope of infertility-related treatments in the data source captures those medicines and technologies approved by the Ministry of Food and Drug Safety and services covered by the NHIS, potentially underestimating other non-covered or alternative services. Finally, male-factor infertility (N46) was excluded from the analytical case definition. By limiting the cohort to a primary diagnostic code for female infertility (ICD-10: N97), we aimed to minimise the aetiological ambiguity associated with male-factor conditions; this approach directly aligns with the study’s objective to examine the overall status of women’s health and reproductive healthcare utilisation patterns.

This study identified substantial heterogeneity in reproductive healthcare utilisation among women diagnosed with infertility in South Korea. Many women had prior pregnancy-related care and diverse prediagnostic health profiles, and subsequent pregnancy identification within 1 year of diagnosis was associated with age, socioeconomic position, occupational conditions and patterns of healthcare utilisation, in addition to infertility-related interventions. ART use and financial assistance were positively associated with having a pregnancy identified during follow-up, but these associations operated within a broader context of social and institutional conditions and did not translate into uniform utilisation patterns or reproductive outcomes. Our findings highlight that interpreting an infertility diagnosis as the beginning of a single, linear treatment process has its limitations. In this context, infertility diagnosis appears to function less as a discrete clinical transition and more as an administrative classification system through which women may move back and forth between pregnancy-related and infertility care over time. Understanding infertility diagnosis from this perspective may help situate clinical and policy discussions within patient-centred care focused on women’s reproductive health, while calling for future research on how eligibility criteria, social conditions and individual circumstances shape engagement with infertility-related care.

This study uses the National Health Information Database provided by the NHIS. The National Health Insurance is a single-payer mandatory health insurance scheme that covers approximately 97% of the Korean population, with the remaining low-income population covered by the Medical Aid programme; together, the two schemes provide near-universal coverage of the resident population. Following the integration of ART into the national insurance benefit system in October 2017, the database robustly captures reproductive care. While entirely privately funded treatments are not eligible for insurance claims, the government’s financial subsidy programme that covers out-of-pocket payment ensures highly reliable case identification by requiring diagnosis records from a medical professional. The initial dataset comprised 6 362 141 claim records for women aged 20–49 who were diagnosed with infertility between 2018 and 2021. The unit of analysis was shifted from the claim-record level to the individual level, yielding a baseline pool of 392 665 unique women with female infertility (ICD-10: N97) as their primary diagnosis. To identify women with a first recorded infertility diagnosis and to analyse subsequent pregnancy-related events, we applied a 1-year wash-out period prior to the date of initial diagnosis and a 1-year observation period thereafter. Women with repeated diagnoses of male infertility (N46) were excluded to minimise misclassification of diagnostic attribution. We conceptualise the first infertility diagnosis observed after the wash-out period as an administrative entry point into infertility care, recognising that some women may have had previous reproductive difficulties that were not captured within this timeframe. 12 17 Figure 1 illustrates the selection process, identifying women aged 20–49 years newly diagnosed with infertility in 2019 or 2020. Of the initial eligible cohort of 156 781 women, 5211 individuals (3.32%) were excluded due to missing values or extreme outliers in administrative covariates, including age, health insurance premiums, occupational industry type and residential region. These data anomalies primarily result from unrecoverable administrative recording errors or database mismatches during the data linkage process. The final analytic sample thus comprised 151 570 women with complete information on all variables analysed. The dependent variable was defined as pregnancy identified without recorded loss during the 1-year follow-up period after the first infertility diagnosis. Specifically, this outcome was operationalised as receiving pregnancy-related medical care (ICD-10: O00–O99) within 1 year following the first infertility diagnosis with no subsequent record of miscarriage or stillbirth (O00–O08) during the same observation period. 34 Women with a recorded miscarriage or stillbirth code, as well as those with no pregnancy-related claims during the follow-up period, were classified as not having the defined pregnancy outcome. The outcome should be interpreted as a proxy for clinically recognised pregnancies documented within the healthcare system, rather than as a measure of all biological conceptions. Measurement error is possible because early pregnancies, early pregnancy losses without billable claims and pregnancies first documented after the observation window may not be captured. 35 However, the likelihood of such measurement errors or the under-reporting of reproductive events is substantially mitigated by exceptional institutional accessibility and utilisation trends in the context of South Korea’s healthcare system. The annual number of outpatient visits per capita in South Korea is more than double the OECD average, the frequency of prenatal care visits exceeds the WHO recommendations, and almost all deliveries occur within licensed medical facilities. 36 38 Furthermore, universal insurance coverage, paired with extensive state-led financial support packages for pregnancy and infertility care, fosters high medical adherence and continuous clinical monitoring within the healthcare system. Reflecting these contextual and institutional characteristics, our operationalisation of the pregnancy outcome was designed not to estimate the performance of ART. Instead, it aims to capture the inherent complexities of reproductive health, characterised by pregnancy, its repetition and recurrent failures, while minimising the empirical ambiguity stemming from the temporal proximity between the initial infertility diagnosis and subsequent healthcare utilisation. Independent variables include sociodemographic characteristics and categorical variables related to healthcare utilisation patterns. Age at diagnosis was dichotomised into under 35 years and 35 years or older. Healthcare coverage was categorised as National Health Insurance or Medical Aid. Occupational industry type, based on the Korean Standard Industrial Classification, was categorised into six groups: public administration or social services, finance and technology, art and leisure, manufacturing and logistics, primary industries (e.g., agriculture, forestry and fishing) and not applicable-including those not in formal employment. Household income was proxied by health insurance premiums and categorised into quartiles, with the lowest quartile serving as the reference group. Variables related to healthcare utilisation were constructed using information from claims, including ICD codes, billing codes for cost reduction and visit frequency before and after infertility diagnosis. These included the number of outpatient visits in the year before and after diagnosis, the number of distinct diagnoses recorded prior to diagnosis, indicators of pregnancy history before infertility diagnosis, whether financial assistance for infertility treatment was received, and the recorded ART use. The correlation coefficients among these variables ranged from 0.221 to 0.763, indicating minor-to-moderate positive correlations. These measures capture how women interacted with the healthcare system around the time of diagnosis, rather than direct measures of underlying biological risk. We first examined the 20 most prevalent disease groups recorded during the year prior to the first infertility diagnosis and assessed whether prevalence differed by pregnancy status during the 1-year follow-up period. We applied a 298-category disease classification system and constructed indicators drawn from literature review and national statistics to reflect both health status and healthcare utilisation. Logistic regression analyses were performed to identify empirical associations between selected covariates and pregnancy identified without recorded loss within 1 year after infertility diagnosis. The analytic objective was associative and descriptive rather than causal; coefficients were interpreted as patterns of association observed within the healthcare system, rather than as causal or deterministic effects on reproductive outcomes. The covariates entered into the regression models were constructed to represent conceptually distinct domains such as sociodemographic characteristics (e.g., age, employment industry, healthcare coverage type, premium quantile, region, disability), healthcare engagement intensity (prediagnostic visit count, prediagnostic distinct-diagnosis count, postdiagnostic visit count) and infertility policy variables (e.g., financial assistance and ART use) based on the literature on infertility care and outcomes. In particular, the status of ART use and financial assistance for infertility treatment were analysed in separate models because they were conceptually related but not empirically identical in the study population. The overlap between these two variables was not complete, suggesting that policy support and clinical treatment captured different aspects of postdiagnosis infertility treatment. As a result, model 1 included financial assistance for infertility treatment, whereas model 2 included ART use. This modelling strategy was used to examine whether the observed associations differed depending on whether postdiagnostic infertility care was represented by policy support or clinical treatment. To examine robustness and potential effect modification, subgroup analyses were conducted by women’s history of prior pregnancy before infertility diagnosis, with separate logistic regression models applied to women with and without such history. Adjusted ORs and 95% CIs were calculated. All analyses were conducted using SAS software (V.9.4; SAS Institute, Cary, North Carolina). This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.