Absolute and relative socioeconomic inequalities in cancer screening participation among adults aged 40–74 years in South Korea: a repeated cross-sectional study

Abstract

Background

Cancer is the leading cause of death in South Korea, underscoring the importance of screening. Despite the implementation of the National Cancer Screening Programme, persistent socioeconomic disparities remain a public health challenge. While previous studies have examined income-related inequalities in cancer screening, education-related inequalities have received comparatively less attention, and few studies have simultaneously quantified both dimensions of socioeconomic inequality over time. We aimed to quantify the absolute and relative socioeconomic inequalities in cancer screening among Korean adults aged 40–74 years.

Methods

Using repeated cross-sectional data from the Korea National Health and Nutrition Examination Survey (2016–2024), age- and sex-adjusted Slope Index of Inequality (SII) and Relative Index of Inequality (RII) were estimated annually, with the RII estimated using modified Poisson regression with robust variance to yield prevalence-ratio-based estimates. Joinpoint regression was used to assess temporal trends, and the Wagstaff Concentration Index (C-index) decomposition was used to identify contributing factors.

Results

Altogether, 31,272 participants were included. Overall screening rates increased from 69.0% in 2016 to 74.6% in 2024. Pro-rich inequalities persisted throughout the study period, with positive SII values and RII values greater than 1. Education-related inequality peaked in 2018 (SII 0.21; RII 1.34) and declined in 2024 (SII 0.10; RII 1.14), with no significant trend (SII: Annual Percent Change [APC] −4.25%, P = 0.142; RII: APC −0.98%, P = 0.225). Income-related inequality was consistently larger than education-related inequality across all years, peaked in 2016 (SII 0.27; RII 1.47) and declined significantly (SII: APC −3.26%, P = 0.011; RII: APC −1.36%, P = 0.010). The C-index decomposition indicated that income-related inequality was associated with several observed socioeconomic and healthcare coverage factors. Non-enrolment in private health insurance was a consistent positive contributor across all years, while unemployment or economic inactivity and Medical Aid beneficiary status showed larger positive contributions during the COVID-19 pandemic period.

Conclusion

Despite increases in overall screening uptake, socioeconomic inequalities favouring advantaged groups persisted in South Korea. The narrowing of income-based disparities is consistent with partial mitigation of financial barriers, whereas education-based disparities persisted. Achieving equitable screening will require multifaceted interventions, including sickness benefits, tailored health literacy support, and proactive community outreach.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28540-5.

Background

Cancer is one of the leading causes of death worldwide, accounting for 10 million deaths in 2020 [1]. This pattern has also been observed in South Korea, where cancer has consistently been reported as the leading cause of death [2]. According to the ‘2024 Causes of Death Statistics’ by the Ministry of Data and Statistics, malignant neoplasms accounted for 24.8% of all deaths, maintaining their position as the number one cause of death with a mortality rate of 174.3 per 100,000 population [3]. Notably, the cancer mortality rate has shown a continuous upward trend, increasing from 150.9 per 100,000 in 2014 to 166.7 in 2023 [3]. This persistent increase in cancer-related mortality underscores the importance of cancer prevention and early detection.

The World Health Organization emphasises the necessity of early diagnosis and screening, as early detection and treatment of cancer significantly reduce mortality rates [1]. In South Korea, the government has supported the National Cancer Screening Programme (NCSP) since 1999 to reduce cancer-related deaths by promoting early detection and subsequent treatment [4–6]. The major cancers with the highest incidence rates in South Korea are stomach, liver, colorectal, breast, cervical, and lung cancers [5]. These cancers can be detected early through screening and have been reported to show high survival rates when treated in their early stages [5]. Building on these benefits, the Korean government has implemented various policies to improve participation rates in cancer screening. In 2005, the target population for NCSP was expanded to include those in the lower 50% of the National Health Insurance (NHI) premium bracket and the copayment rate for cancer screening was reduced to 10% by 2010 [6]. Furthermore, in 2018, copayment for colorectal and cervical cancer screening was eliminated (adjusted to 0%) to further enhance accessibility [6]. Opportunistic private screening provided by medical institutions is also used as a screening tool; however, since individuals must bear the entire cost, it presents limitations in terms of socioeconomic accessibility [7, 8].

South Korea has achieved significant outcomes through the NCSP, including increased screening participation rates and improved cancer survival rates [6, 9]. However, in terms of screening participation, socioeconomic disparities persistently remain, posing a critical challenge [6–8]. The overall participation rate has risen since the introduction of NCSP; however, the gap in screening rates between NHI beneficiaries and Medical Aid recipients remained at 20 percentage points annually from 2016 to 2020 [6]. Furthermore, socioeconomic inequalities based on education and income levels exist in opportunistic private screening, where individuals bear the full cost out-of-pocket [8, 10–13]. This issue is not confined to South Korea but is a common phenomenon in various countries [14–17]. In Japan, inequalities based on education level have continuously been observed, and it has been reported that this gap further widened particularly after the COVID-19 pandemic [14]. Similar income- and education-related gradients have been reported in China and across 28 European countries, where lower income and education levels correspond with lower screening rates [15, 17].

Socioeconomic inequalities in cancer screening have been reported worldwide; however, studies conducted in South Korea are limited. Existing research on cancer screening disparities has largely been confined to cross-sectional analyses at a single point in time or focused solely on socioeconomic inequalities associated with specific types of cancer screening [8, 10, 18–20]. Furthermore, long-term trend studies that multidimensionally elucidate how evolving health policies or unexpected public health crises correlate with the trajectories and structural causes of these inequalities over time are lacking [13].

Although education and income are often used interchangeably as markers of socioeconomic position, they capture conceptually distinct dimensions and may influence cancer screening through different pathways [21]. Education reflects human capital acquired early in life and shapes health literacy, risk perception, and the ability to navigate health information and screening systems [22]. Income, in contrast, reflects material resources that determine the opportunity costs of screening, the affordability of opportunistic private screening, and exposure to competing financial constraints [23]. It is also closely tied to employment conditions and occupational security [24]. Because these mechanisms are not interchangeable, inequalities indexed by education and by income need not move in parallel, and examining them separately can reveal pathway-specific barriers that a single composite indicator would obscure [25].

Therefore, the purpose of this study was threefold. First, using a continuous 9-year dataset (2016–2024) from the Korea National Health and Nutrition Examination Survey (KNHANES), a nationally representative sample, we quantified trends in education- and income-based inequalities in cancer screening participation among Korean adults aged 40–74 years. For this, we used the Slope Index of Inequality (SII) and the Relative Index of Inequality (RII). Second, we sought to identify long-term trend inflection points of socioeconomic inequality using joinpoint regression analysis. Third, by applying the Concentration Index (C-index) decomposition method, we conducted an in-depth analysis of how the contributions of specific underlying factors that constitute inequality fluctuate over specific periods. Through these findings, we intend to identify vulnerable populations in cancer screening and provide policy evidence for the efficient allocation of healthcare resources, thereby contributing to the enhancement of socioeconomic equity and strengthening of management systems for cancer prevention.

Methods

Study design and data source

KNHANES is an annual nationwide cross-sectional survey conducted by the Korea Disease Control and Prevention Agency employing a stratified multistage probability sample design [26, 27].

In this cross-sectional study, we used data from the seventh (2016–2018), eighth (2019–2021), and ninth (2022–2024) phases of KNHANES. To prevent seasonal bias, we included the entire 2016–2024 data continuously collected throughout the year and used fully finalised and publicly released datasets for all years.

The number of targeted individuals for each phase was 31,689 (2016–2018), 30,490 (2019–2021), and 29,076 (2022–2024). The actual number of participants was 24,269 (2016–2018), 22,559 (2019–2021), and 20,191 (2022–2024), demonstrating participation rates of 76.6% (2016–2018), 74.0% (2019–2021), and 69.4% (2022–2024), respectively.

The study population was restricted to adults aged 40–74 years, a pragmatic analytic range that captures the main adult age groups targeted by multiple NCSP components rather than a strict cancer-specific eligibility criterion. NCSP eligibility varies by cancer type, sex, age, and risk status, with most non-cervical screening programmes beginning at age 40 or 50 and lung cancer screening limited to high-risk adults aged 54–74 years [5, 6]. Because the KNHANES item assessed any cancer screening within the past 2 years and did not consistently identify cancer type or screening programme, adults aged 20–39 years were excluded to reduce eligibility-related heterogeneity, particularly the mixing of women eligible for cervical cancer screening with adults outside most non-cervical NCSP indications.

First, individuals with missing values for the outcome variable, cancer screening participation (N = 3,033), were excluded from the study. Multiple imputation were not applied to missing outcome variables because artificially generating unobserved outcomes risks undermining methodological validity [28]. Furthermore, to address potential selection bias, we confirmed that there was no substantial difference in key socioeconomic indicators, specifically education and income levels, between these excluded individuals and the included study population (P > 0.05). Among the participants with observed outcome variables (N = 31,627), 355 (1.1%) with missing covariate values were excluded. This exclusion without applying multiple imputation was based on methodological guidelines stating that complete-case analysis produces negligible bias when the missing rate of covariates is < 5% [29, 30]. Ultimately, we included 31,272 individuals in the final analysis (Fig. 1).

Fig. 1.

Flow diagram of study participants. Note: KNHANES, Korea National Health and Nutrition Examination Survey

Outcome variable

The outcome variable was cancer screening participation. Respondents who answered “Yes” to the question “Have you been screened for cancer in the past 2 years?” in the KNHANES were considered participants. Participation in cancer screening in KNHANES was evaluated by simultaneously considering not only cancer screening provided by the NHI service but also those conducted at public and maternal and child health centres, as well as opportunistic private cancer screenings conducted at clinics, hospitals, or comprehensive health check-up centres [13]. Because the KNHANES item captures any cancer screening within the past 2 years, without consistently distinguishing cancer type or programme type, the outcome should be interpreted as overall cancer-screening participation rather than cancer-specific NCSP uptake [13]. Including both NCSP and opportunistic private screening may better reflect actual screening utilisation, as NCSP-only definitions could underestimate total screening uptake [31]; however, this approach cannot estimate the effect of specific NCSP components, eligibility criteria, or cost-sharing policies.

Socioeconomic variables

Education level was classified into four categories: elementary school graduate or lower, middle school graduate, high school graduate, and college degree or higher. Income level was divided into quintiles (Q1–Q5) based on the equivalised monthly household income.

To precisely calculate the socioeconomic inequality indices, including SII and RII, education and income levels were converted into ridit scores, which are relative ranking indicators based on empirical distributions [32]. Ridit scores control for changes in population size and heterogeneity of the categorical distribution within each survey year, representing an individual’s relative socioeconomic position as a continuous value ranging from 0 to 1 [32]. Specifically, within the total study population for each year, the ridit score of an individual belonging to a particular socioeconomic category was calculated as the overall proportion of the cumulative population in lower-ranking categories plus half the proportion of the population in the current category (the midpoint of the cumulative distribution) [32–34]. Through this transformation, the relative hierarchical position of socioeconomic status was standardised, thereby ensuring the validity of comparing inequality trends over time.

Covariates

The covariates included in the analysis were age (continuous variable), sex (male, female), occupation status (managers & professionals, office workers, service & sales workers, agriculture/forestry, craft/machine operators, elementary occupations, unemployed), marital status (married, unmarried), region (urban, rural), subjective health status (good, fair, poor), current smoking (yes, no), monthly drinking (yes, no), NHI type (workplace, regional, Medical Aid), and private health insurance enrolment (yes, no).

Statistical analysis

Descriptive statistics for all study variables were calculated by survey year, with categorical variables presented as weighted percentages and continuous variables as weighted means.

To quantify the socioeconomic inequality in cancer screening participation, SII and RII by education and income levels were calculated for each year. SII represents the absolute difference in cancer screening rates between the lowest and highest socioeconomic groups. SII was estimated through linear regression analysis, with the ridit score and cancer screening participation as the exposure and outcome variables, respectively. For the SII, we verified that no predicted probabilities fell outside the [0, 1] interval in any survey year (minimum and maximum fitted probabilities: 0.47 and 0.95, respectively), and the design-based (Taylor-linearized) variance estimator used in ‘PROC SURVEYREG’ yields standard errors that are robust to the heteroscedasticity inherent in a linear probability model. RII, an indicator reflecting relative inequality, was estimated using modified Poisson regression with a log link and robust variance estimation using KNHANES sampling weights (PROC GENMOD with a REPEATED statement clustering on the primary sampling unit within strata), with the ridit score as the exposure and cancer screening participation as the outcome. Because the overall screening rate was high (> 68%), this risk-based approach was preferred over logistic regression to avoid overestimating relative inequality that occurs when an odds ratio is interpreted as a prevalence ratio for a common outcome. The resulting RII is therefore expressed as a prevalence ratio [35, 36]. Both indices were adjusted for age and sex and calculated independently for each year from 2016 to 2024 [37, 38]. Additionally, to examine alterations in the magnitude of socioeconomic inequality owing to the COVID-19 pandemic, an interaction term between the pandemic period indicator and education/income levels was tested. The results were presented using F- and P-values. The pandemic periods were categorised as pre-pandemic (2016–2019), pandemic (2020–2021), transition (2022), and post-pandemic/recovery period (2023–2024).

To evaluate the temporal trends of SII and RII, joinpoint regression analysis was conducted using the Joinpoint Trend Analysis Software (Version 5.0.2; National Cancer Institute). The standard error (SE) of each annual estimate was used as the weight in the model. A logarithmic transformation was applied to the RII, a relative ratio measure, whereas the SII, an absolute measure, was modelled on the original scale. The optimal number of joinpoints was determined through permutation tests. Considering the characteristics of the 9-year dataset, a maximum of two joinpoints were allowed. In cases where a single linear trend with zero joinpoints emerged, a significant annual percent change (APC) was interpreted as a consistent directional change across the entire study period.

To decompose the structural causes underlying the inequality in cancer screening participation, the Wagstaff C-index decomposition method was applied annually [39]. Since the outcome variable in this study was binary, a non-linear decomposition method was employed to avoid the predicted-value problem associated with linear probability models [40]. C-index was calculated as a measure of income-related inequality in cancer screening and a positive value indicated inequality favouring the high-income group. C-index was decomposed into the contributions of each covariate, based on the following equation:

Here, denotes the marginal effect of the covariate  k derived from the logistic model, which was calculated by transforming the logistic regression coefficient () of each covariate as = × (1−). is the weighted mean of the covariate  k, is the mean screening rate, and refers to the C-index of the covariate  k with respect to income rank. The C-index for each covariate () was estimated by regressing the covariate against the fractional income rank (ridit score) using ordinary least squares regression. Here, the fractional income rank was constructed by standardising the weighted cumulative household income distribution into a range of 0–1. The total C-index derived from this model consisted of the sum of the contributions of the observed variables in the model and a residual. The residual represents the remaining portion of the concentration index not accounted for by the observed covariates included in the decomposition model. This residual may reflect unmeasured factors, model specification, or random measurement error; therefore, it was interpreted descriptively as the unexplained component of the decomposition rather than as a direct measure of omitted-variable bias.

In the primary decomposition model, household income ridit rank was included as a covariate, following previous applications of C-index decomposition in which the ranking variable is incorporated to estimate the direct income-rank component of inequality [13]. Because the concentration index itself is defined with respect to income rank, we additionally conducted a sensitivity analysis excluding household income ridit rank to examine how the remaining observed covariates accounted for income-related inequality when the direct income-rank component was not included [39, 41].

Moreover, owing to the binary nature of the outcome variable, the mathematical limits of the standard C-index are constrained by the mean screening rate, potentially leading to an underestimation of the true magnitude of inequality. Therefore, to verify that the observed trends in socioeconomic inequality were not merely mathematical artefacts driven by annual variations in mean screening rates, Wagstaff- and Erreygers-normalised C-indices were additionally calculated to assess the robustness of the trends [42, 43]. The contribution of each covariate was presented as a percentage relative to the overall C-index, with all decomposition analyses performed independently for each year.

Finally, to assess the robustness of the primary findings, supplementary analyses including sex-stratified SII/RII analyses, a robustness analysis based on directly age-standardised rates, and an analysis restricted solely to NCSP participants were conducted. To verify that the observed temporal trends were not driven by the changing age structure of the study population (mean age increased from 54.2 years in 2016 to 55.9 years in 2024), directly age-standardised screening rates were computed using a fixed reference population, and the annual SII and RII were re-derived from these standardised rates. The fixed reference was the pooled 2016–2024 weighted age distribution in five age bands (40–49, 50–54, 55–59, 60–64, and 65–74 years), so that every survey shared an identical age structure. Standardised rates were computed within strata of education and income, and the SII and RII were obtained by weighted regression of the standardised stratum-specific rates on the ridit-scaled socioeconomic rank. For this analysis, the education variable was collapsed into three categories (≤ middle school, high school, ≥college) because the lowest educational stratum among the younger age bands contained too few observations for stable age-specific estimation; income was retained in five quintiles.

All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) to account for the complex sampling design of the KNHANES (stratification, kstrata; cluster, psu; weight, wt_itvex). When calculating annual indicators, the annual weights of the corresponding years were independently used. However, exclusively for the integrated 9-year analysis, to prevent overestimation of population size, the individual annual weights were divided by nine to generate and apply a new integrated annual weight.

The variance inflation factor (VIF) was calculated to diagnose multicollinearity among variables. The maximum VIF was 3.13, confirming the absence of multicollinearity issues. In all analyses, the level of significance was set at a two-tailed P-value < 0.05. This study was exempted from ethical review by the Institutional Review Board (IRB) of the corresponding author’s affiliated institution (IRB No. 1041849-202604-SB-073-01) and conducted in accordance with the principles of the Declaration of Helsinki.

Results

Sociodemographic characteristics by survey years

The mean age of the participants significantly increased across the cross-sectional survey waves, rising from 54.2 years in 2016 to 55.9 years in 2024 (P < 0.001). In contrast, the sex distribution remained constant across all survey years, with females accounting for 50–51% of the population (P = 0.999) (Table 1; see Additional file 1).

Table 1.

Demographics and socioeconomic characteristics of the study population (2016–2024)

Variables	Years									P-value
	2016 (N = 3484)	2017 (N = 3589)	2018 (N = 3676)	2019 (N = 3678)	2020 (N = 3374)	2021 (N = 3363)	2022 (N = 3088)	2023 (N = 3536)	2024 (N = 3484)
Age (mean ± SE*)	54.20 ± 0.22	54.48 ± 0.27	54.63 ± 0.25	54.78 ± 0.27	54.87 ± 0.28	55.28 ± 0.27	55.56 ± 0.32	55.65 ± 0.25	55.94 ± 0.29	< 0.001
Sex (N, %†)
Male	1499, 49.3%	1573, 49.1%	1563, 49.0%	1560, 49.0%	1484, 49.6%	1456, 49.5%	1307, 49.0%	1485, 48.8%	1441, 49.1%	0.999
Female	1985, 50.7%	2016, 50.9%	2113, 51.0%	2118, 51.0%	1890, 50.4%	1907, 50.5%	1781, 51.0%	2051, 51.2%	2043, 50.9%
Education level (N, %)
≤ Elementary school	833, 18.3%	782, 16.7%	714, 15.6%	666, 13.5%	588, 12.5%	585, 12.6%	535, 11.9%	494, 10.6%	482, 10.3%	< 0.001
Middle school	498, 13.6%	502, 13.5%	493, 12.0%	462, 11.2%	439, 10.9%	422, 10.4%	334, 9.3%	430, 10.2%	453, 11.1%
High school	1117, 35.1%	1132, 33.1%	1300, 37.5%	1293, 37.0%	1192, 37.9%	1189, 38.2%	1055, 35.8%	1265, 36.9%	1185, 34.1%
≥ College graduate	1036, 33.0%	1173, 36.7%	1169, 34.9%	1257, 38.3%	1155, 38.7%	1167, 38.8%	1164, 43.0%	1347, 42.3%	1364, 44.5%
Household income (N, %)
Q1	744, 17.8%	467, 10.7%	422, 9.5%	455, 9.8%	384, 9.1%	388, 8.8%	381, 9.5%	402, 9.0%	356, 8.1%	< 0.001
Q2	686, 17.8%	693, 17.2%	729, 18.9%	709, 17.1%	617, 15.9%	626, 16.8%	622, 16.8%	663, 17.1%	601, 15.4%
Q3	637, 19.4%	751, 21.1%	751, 20.8%	742, 20.8%	712, 21.6%	762, 23.2%	620, 20.9%	751, 21.3%	798, 23.0%
Q4	644, 20.5%	772, 23.3%	902, 27.0%	872, 25.9%	824, 26.0%	754, 24.1%	692, 24.6%	824, 25.0%	823, 25.0%
Q5	773, 24.5%	906, 27.7%	872, 23.8%	900, 26.4%	837, 27.4%	833, 27.1%	773, 28.2%	896, 27.6%	906, 28.5%

All values are weighted to represent the general South Korean population

Statistically significant results (P < 0.05) are indicated in bold

SE Standard error

* Values are presented as weighted mean ± standard error

† Values are presented as weighted percentage

Like age, significant differences were observed in education and income levels according to survey year. Regarding the education level, the proportion of participants with an elementary school graduate or lower decreased by 8.0 percentage points, from 18.3% in 2016 to 10.3% in 2024. Conversely, the proportion of those with a college degree or higher, which was 33.0% in 2016, steadily maintained a level in the low-to-mid 40% range from 2022 onwards (P < 0.001). The proportion of participants in the lowest income quintile (Q1) decreased from 17.8% in 2016 to 8.1% in 2024, whereas the proportion in the highest income quintile (Q5) increased from 24.5% in 2016 to 28.5% in 2024 (P < 0.001).

Cancer screening rates by survey years

Cancer screening rates gradually increased over the study period (Fig. 2). The overall screening rate increased from 69.0% (SE 1.0) in 2016 to 74.6% (SE 1.0) in 2024. When examining the screening rates according to sex, females consistently exhibited higher cancer screening rates than did males throughout the study period. Specifically, the rate among females increased from 72.7% (SE 1.2) in 2016 to 76.6% (SE 1.1) in 2024. During the same period, the rate among males increased from 65.0% (SE 1.4) in 2016 to 72.0% (SE 1.3) in 2024. However, the absolute sex gap in screening participation narrowed from 7.7 percentage points in 2016 to 4.6 percentage points in 2024.

Fig. 2.

Cancer screening participation rates by sex and overall (2016–2024). Notes: Error bars represent standard error. Estimates were calculated using survey design-adjusted methods accounting for stratification, clustering, and sampling weights

Magnitude of socioeconomic inequalities in cancer screening (2016–2024)

The age- and sex-adjusted SII and RII by education and income levels from 2016 to 2024 are presented in Fig. 3. Throughout the study period, significant pro-rich inequalities were persistently observed on both absolute and relative scales. The positive SII values and RII values greater than 1 indicated that higher socioeconomic status was associated with higher predicted probabilities and a higher prevalence ratio of cancer screening participation, respectively. Regarding education level, SII and RII showed annual fluctuations but followed a similar trajectory. Education-based inequality peaked in 2018 at an SII of 0.21 (95% confidence interval [CI] 0.14 to 0.29; P < 0.001) and an RII of 1.34 (95% CI 1.21 to 1.48; P < 0.001), and was observed to be lowest in 2024 with an SII of 0.10 (95% CI 0.02 to 0.18; P < 0.05) and an RII of 1.14 (95% CI 1.03 to 1.26; P < 0.05). Inequality driven by income level was more pronounced compared to that driven by education level. Both the income-based SII and RII were highest in 2016 (SII 0.27, 95% CI 0.20 to 0.33; RII 1.47, 95% CI 1.34 to 1.62); both reached their lowest points in 2022 (SII 0.16, 95% CI 0.09 to 0.23; RII 1.24, 95% CI 1.13 to 1.37). The income-based RII rebounded in 2023 (1.36, 95% CI 1.24 to 1.48) but decreased again in 2024 (1.32, 95% CI 1.20 to 1.45).

Fig. 3.

Socioeconomic inequalities in cancer screening participation from 2016 to 2024. Notes: (A) The slope index of inequality (SII) represents the absolute difference in the predicted probability of screening between the highest and lowest socioeconomic ranks (B) The relative index of inequality (RII) was estimated using modified Poisson regression with robust variance and represents the adjusted prevalence ratio of screening participation comparing the theoretical highest socioeconomic rank with the lowest rank

Additionally, the interaction term tested between the COVID-19 pandemic period and education/income levels was not significant (Education × Pandemic: F = 0.62, P = 0.601; Income × Pandemic: F = 1.76, P = 0.152).

The sex-stratified analysis was broadly consistent with the main findings (see Additional file 2). Among males, income-based inequality was consistently larger than education-based inequality, with income-related SII ranging from 0.38 to 0.26 and income-related RII ranging from 1.79 to 1.36. Among females, pro-rich inequalities were also observed across most survey years, although the estimates were smaller and statistical significance was attenuated in some years, particularly in 2020. These findings indicate that socioeconomic inequality in cancer screening participation persisted in both sexes but was more pronounced among males.

Robustness to age standardisation

The age-standardised indices were highly concordant with the model-based estimates. The income-based RII declined from 1.48 in 2016 to 1.30 in 2024, closely tracking the main analysis (Pearson correlation ≈ 0.90 for SII and ≈ 0.94 for RII). The education-based indices preserved the same overall magnitude and consistently pro-rich direction, although their year-to-year pattern differed somewhat owing to the three-category specifications used for standardisation. These results indicate that the reported inequalities, particularly the income-based trends, are robust to the changing age structure and are not artefacts of conditional age adjustment (see Additional file 3).

Trend analysis of inequality indices using joinpoint regression

No joinpoints were detected across the inequality indices; therefore, the data were summarised as a single linear segment from 2016 to 2024 (Table 2). Given the limited number of annual estimates, this result should be interpreted as a descriptive summary of temporal patterns rather than as definitive evidence of the absence of structural breakpoints. Within this period, the patterns of inequality in cancer screening participation differed between education and income levels. Specifically, education-based inequality showed a numerical decrease (SII: APC −4.25%, 95% CI −11.20 to 2.08, P = 0.142; RII: APC −0.98%, 95% CI −2.59 to 0.70, P = 0.225); however, this declining trend was not significant. In contrast, APC for the income-based SII and RII were 3.26% (95% CI −5.99 to −0.78; P = 0.011) and −1.36% (95% CI −2.34 to −0.36; P = 0.010), respectively, indicating a significant narrowing of income-based inequality.

Table 2.

Trend in socioeconomic inequalities in cancer screening participation: Joinpoint regression results (2016–2024)

Variables	Education level		Household income
	SII	RII	SII	RII
Joinpoints (n)	0	0	0	0
APC (95% CI)	−4.25 (11.20, 2.08)	−0.98 (−2.59, 0.70)	−3.26 (−5.99, −0.78)	−1.36 (−2.34, −0.36)
P value	0.142	0.225	0.011	0.010

Joinpoint regression was performed using the joinpoint trend analysis software (Version 5.0.2, National Cancer Institute, USA). The number of joinpoints was determined by permutation test with a maximum of two joinpoints permitted. Zero joinpoints indicate a single linear trend over the study period, in which case APC equals AAPC (average annual percent change)

APC Annual percent change. CI Confidence interval, SII Slope index of inequality, RII Relative index of inequality

C-index decomposition of income-related inequality in cancer screening participation

Throughout the study period, the overall C-index maintained significant positive values ranging from 0.03 to 0.05, reaffirming that cancer screening participation was consistently distributed in favour of high-income groups (pro-rich inequality) (Table 3).

Table 3.

Decomposition of the concentration index for income-related inequality in cancer screening participation, 2016–2024

Variables		Years
		2016 (%)	2017 (%)	2018 (%)	2019 (%)	2020 (%)	2021 (%)	2022 (%)	2023 (%)	2024 (%)
Household income (ridit rank)		83.9%	88.3%	78.0%	62.5%	57.5%	51.3%	40.8%	67.8%	92.7%
Education level (ref. ≥ college graduate)
	≤ Elementary school	8.7%	−9.1%	14.1%	8.6%	7.8%	−0.4%	4.9%	2.5%	−3.3%
	Middle school	3.2%	−0.1%	3.0%	−0.1%	−0.4%	−1.0%	−3.1%	1.5%	1.1%
	High school	−1.5%	0.4%	−0.4%	0.1%	−0.0%	−0.0%	0.0%	−0.1%	−0.1%
Age (years)		−31.9%	−18.5%	−28.3%	−19.4%	−17.8%	−17.6%	−15.4%	−17.6%	−35.7%
Sex (ref. male)	Female	−2.8%	−2.7%	−2.4%	−3.2%	−2.7%	−1.6%	−3.0%	−5.7%	−4.4%
Occupation status (ref. managers & professionals)
	Office workers	4.1%	−3.6%	3.2%	2.6%	8.5%	−1.9%	−11.9%	10.7%	4.1%
	Service & sales workers	−0.0%	−0.5%	−0.4%	−1.5%	−0.0%	−1.4%	0.1%	−0.2%	0.2%
	Agriculture/forestry	−0.7%	−0.6%	−1.0%	−1.3%	−0.1%	0.6%	1.1%	0.0%	−0.2%
	Craft/machine operators	0.7%	−0.1%	−1.0%	−0.4%	0.0%	−4.0%	−2.8%	−1.5%	0.8%
	Elementary occupations	−1.6%	3.1%	1.5%	3.5%	−3.0%	−1.1%	4.6%	1.7%	1.2%
	Unemployed	1.1%	9.2%	4.7%	9.7%	0.5%	15.1%	11.2%	5.8%	3.8%
Marital status (ref. married)	Unmarried	2.1%	1.9%	2.4%	4.1%	2.8%	3.2%	7.9%	5.2%	3.4%
Region (ref. urban)	Rural	0.7%	0.4%	0.4%	0.1%	−0.1%	0.3%	7.0%	1.9%	−1.4%
Subjective health status (ref. good)	Fair	−0.1%	−0.1%	−0.1%	0.1%	0.4%	0.2%	−1.0%	−0.1%	0.1%
	Poor	2.4%	−0.7%	1.5%	2.1%	−2.7%	−1.6%	−5.5%	−3.3%	1.4%
Current smoking (ref. no)	Yes	0.9%	0.7%	1.3%	2.3%	1.8%	5.0%	5.2%	2.7%	−0.7%
Monthly drinking (ref. no)	Yes	1.1%	1.7%	−1.1%	1.3%	0.4%	4.6%	4.9%	3.2%	2.3%
National health insurance type (ref. workplace)
	Regional	4.0%	6.4%	5.4%	5.3%	8.1%	7.7%	3.6%	4.5%	10.5%
	Medical Aid	1.0%	0.7%	1.8%	4.1%	9.0%	10.2%	13.6%	9.5%	5.2%
Private health insurance enrolment (ref. yes)
	No	29.7%	25.0%	21.1%	15.5%	31.3%	27.4%	28.1%	15.9%	15.8%
	Total concentration index*	0.05	0.05	0.04	0.05	0.04	0.04	0.03	0.04	0.03
	Σ Explained (%)†	105.2%	101.8%	102.8%	96.1%	101.3%	94.9%	90.4%	104.4%	96.8%
	Residual (%)‡	−5.2%	−1.8%	−2.8%	3.9%	−1.3%	5.1%	9.6%	−4.4%	3.2%

Table values are presented as the percentage contribution of each covariate to overall inequality. A positive contribution indicates a factor associated with a widening contribution to pro-rich inequality within this decomposition framework, whereas a negative contribution indicates a contribution in the opposite direction. These values should be interpreted descriptively and not as causal effects

* Total concentration index represents the overall standard concentration index prior to decomposition

† Σ Explained (%) represents the total explained fraction, calculated as the sum of the signed contributions of all observed covariates; it may exceed 100% owing to the offsetting effects of positive and negative contributions

‡ Residual (%) represents the portion of the concentration index not accounted for by the observed covariates included in the decomposition model.

In this decomposition, each covariate’s percentage contribution reflects the joint product of its unequal distribution across income ranks and its association with screening participation within the fitted model. A positive contribution indicates a contribution in the direction of greater pro-rich inequality, whereas a negative contribution indicates a contribution in the opposite direction. The larger the absolute percentage, the greater that factor’s share of the overall concentration index. These percentage contributions should not be interpreted as causal effects.

In the primary decomposition model including household income ridit rank, the contribution of the income-rank term to the overall income-related concentration index reached 92.7% in 2024. Given that the C-index is defined with respect to income rank, this finding is best interpreted as the decomposition component attributable to the income-rank variable, rather than as a causal explanation of the observed inequality. Non-enrolment in private health insurance, a supplementary form of coverage, consistently showed a high positive contribution ranging from 15.5 to 31.3% annually, identifying it as a covariate consistently associated with income-related inequality. Conversely, age exhibited a strong negative contribution consistently ranging from −15.4 to −35.7% throughout the study period, acting as a significant offsetting factor that mitigated income inequality. Sex also showed a marginal but negative contribution, partially offsetting the overall inequality.

Notably, prior interaction analysis revealed that the pandemic did not significantly alter the overall magnitude of inequality; however, there were notable fluctuations in the relative contributions of its internal components. During the peak of the public health crisis in 2021, the contribution of unemployment reached 15.1%, and the contribution of Medical Aid beneficiaries surged to 13.6% in 2022, emerging as major components of inequality during those periods.

The total explained fraction (Σ Explained) of the factors included in this decomposition ranged from 90.4 to 105.2%, and the contribution of the residual term was small, ranging from −5.2 to 9.6% (mean absolute value: 4.1%). This suggests that the measured covariates accounted for most of the decomposed C-index within this model specification. However, this should not be interpreted as excluding the possibility of unmeasured mechanisms.

Furthermore, in the sensitivity analysis excluding the income variable, the residual increased to an average of 51.2% (see Additional file 4), suggesting that household income rank accounted for a large share of the decomposed C-index in the primary model specification. Conversely, in the model excluding income, the contributions of non-enrolment in private insurance (average 25.7%), educational attainment of elementary school or lower (average 9.6%), and unemployment (average 9.8%) increased. This pattern suggests that these factors may partly capture socioeconomic patterning related to income.

Finally, to adjust for the boundary limits of the C-index caused by the binary nature of the outcome variable (mean screening rate > 68%), normalisation sensitivity analysis was conducted. The results showed that the absolute magnitudes of the Wagstaff and Erreygers normalised C-indices were larger than those of the standard C-index (see Additional file 5). However, Pearson correlation coefficients among the three indices were high (> 0.93), suggesting that the observed trends and directions, including the temporary alleviation of the gap in 2022 and the rebound of inequality in 2023–2024, were not solely attributable to mathematical boundary effects driven by fluctuations in mean screening rates.

Inequalities restricted to the NCSP (2016 and 2019)

Using data from 2016 and 2019, which allowed for the classification of screening types, analysis resulted in the outcome variable being solely participation in NCSP (see Additional file 6). The results revealed a reduction in the magnitude of socioeconomic inequality compared with the main analysis, which included all types of cancer screening. In 2016, neither education-based inequality (SII 0.03, P = 0.50; RII 1.05, P = 0.51) nor income-based inequality (SII 0.06, P = 0.07; RII 1.11, P = 0.08) was significant. Conversely, in 2019, education-based inequality (SII 0.09, P < 0.05; RII 1.14, P < 0.05) and income-based inequality (SII 0.08, P < 0.05; RII 1.13, P < 0.05) persisted. Overall, the magnitude of inequality decreased significantly when analysing NCSP participation only; however, these findings should be interpreted with caution. Owing to the limitations of the survey questionnaire, this analysis was fragmentarily conducted on data from only 2 specific years out of the entire study period.

Discussion

Our findings showed that despite an increase in the overall screening rate from 69.0 to 74.6% during the study period, inequalities favouring higher socioeconomic groups consistently persisted across all survey years (SII > 0, RII > 1). These findings are consistent with those of previous studies in Korea that focused on screening for specific single cancers such as stomach, colorectal, breast, and cervical cancers [10, 20, 23, 44], and demonstrated a continuous trend aligned with studies that analysed overall cancer screening beyond individual cancer types [8, 13, 45]. The persistence of inequality, even in the context of increasing overall screening rates, suggests that improvements in cancer screening accessibility are not equally distributed across socioeconomic strata [13].

Furthermore, a major finding of this study is that income-based inequality consistently exhibited a larger magnitude than did education-based inequality throughout the study period. Similar findings have been reported in previous studies conducted in Korea. For instance, a study focusing on breast and cervical cancer screening reported that income-based inequality was persistently more pronounced compared to education-based inequality [44]. Additionally, although it did not convert income levels into ridit scores, a study utilising the 2010–2012 KNHANES data found a significant positive association between higher household income and cancer screening participation, consistent with our results [46]. This pattern also corresponds with findings from studies conducted in neighbouring Asian countries. In Japan, a recent study on cervical cancer screening during the COVID-19 pandemic showed that lower income had a stronger negative association with screening participation than did lower educational attainment [47]. Another study involving a nationally representative sample in Japan, while not explicitly analysing screening rates by income level, elucidated that the underlying causes of low screening rates among less-educated groups were fundamentally economic factors [48]. These include exclusion from workplace-based screenings due to non-regular employment and the out-of-pocket cost burden of municipal screening [48]. Similarly, a study from China reported that income level consistently exhibited a stronger independent association with cervical cancer screening participation than did education level [15]. Beyond East Asia, comparable socioeconomic gradients have been documented in countries with long-established organised screening programmes. In England, where the National Health Service offers cancer screening free of charge, colorectal cancer screening uptake nonetheless declined steadily across area-level deprivation, from roughly one-half of invitees in the least deprived areas to under one-third in the most deprived [49]. Similar education- and income-related inequalities have been reported in colorectal cancer screening in Canada [16] and across European countries with organised programmes [17]. The persistence of these gradients even under organised, low-cost or free screening suggests that the Korean experience reflects a broader international pattern rather than a feature unique to its health system; however, the larger contribution of income in Korea may be amplified by the substantial role of out-of-pocket opportunistic private screening [8].

A common implication of these domestic and international findings is that, despite the existence of universal health coverage and NCSP, access remains constrained by a combination of economic vulnerability and disparities in health literacy. This economic barrier extends beyond the direct out-of-pocket medical costs associated with screening. Previous studies have reported that multidimensional opportunity costs incurred to undergo screening, such as lost wages, travel time, and limited access to information, can act as significant barriers to participation [50–53]. Therefore, policies aimed solely at reducing direct out-of-pocket costs of cancer screening may have limitations in alleviating inequality. Instead, it is crucial to address the multidimensional economic barriers faced by low-income populations [20]. These opportunity costs act as substantial constraints, particularly for non-regular and daily wage workers [47]. According to a study conducted in the United States, mandatory paid sick leave was significantly associated with an increased uptake of breast and colorectal cancer screenings, suggesting that such policies can serve as an effective mechanism for mitigating structural barriers among low-income workers [51]. In South Korea, policies such as the ‘Seoul-type Paid Sick Leave Support’, which provides cash compensation for wages lost on the days of screening for non-regular workers and vulnerable populations, have recently been introduced and their potential to enhance preventive health-screening rates has been exploratorily evaluated [54]. However, systems that compensate for economic opportunity costs have yet to be established as universal nationwide programmes. Therefore, to substantively alleviate socioeconomic inequalities in cancer screening participation, it is necessary to move beyond merely reducing copayments for vulnerable groups. Direct economic support systems at an inter-ministerial level, linked with labour and welfare policies, such as guaranteeing official leave for screening purposes, providing paid leave support, and accelerating the introduction of a nationwide sickness benefit system, must be considered.

Income level has emerged as a factor closely associated with socioeconomic inequality in cancer screening. Furthermore, the joinpoint trend analysis revealed that income-based inequality demonstrated a significant decreasing trend during the study period (SII: APC −3.26%; P < 0.05, RII: APC −1.36%; P = 0.010). Conversely, education-based inequality exhibited a numerical tendency to decrease; however, this trend was not significant and presented a pattern different from that of income-based inequality. Given the limited number of annual observations, the joinpoint analysis should be interpreted as an exploratory assessment of overall trend direction rather than as definitive evidence for the absence of structural changes. This limitation is particularly relevant for the education-based indices: with only nine annual data points, the joinpoint procedure has limited statistical power to detect inflection points or modest monotonic trends, so the non-significant education-based APC should be regarded as statistically inconclusive rather than as evidence that educational inequality remained stable over the study period. However, the observed single trend without statistically significant inflection points across all inequality indices may still be examined from multiple perspectives. First, the expansion of financial support for the NCSP, such as the reduction of copayments for cancer screening in 2010 and the complete exemption of copayments for colorectal and cervical cancer screenings in 2018, appears to have manifested as a long-term, gradual alleviation of economic barriers for vulnerable populations over the 9-year period, rather than causing an abrupt structural shift at a specific point in time [6].

Second, the observation that the inequality trend did not bend despite the massive shock of the COVID-19 pandemic indicates that the NCSP remained relatively stable, while also reflecting the structural characteristics of the two indicators. Income-based inequality demonstrated a gradual but continuous downward trend; however, its magnitude remained persistently higher than that of education-based inequality, underscoring the need to strengthen policy measures beyond existing economic support. Notably, income-based inequality experienced short-term fluctuations, rebounding temporarily in 2023 after reaching a trough in 2022, before decreasing again in 2024. The specific causes of these short-term fluctuations are difficult to definitively ascertain within the scope of this study; however, they may be related to changes in healthcare utilisation behaviours surrounding the transition to the post-pandemic phase. Nevertheless, the findings suggest an overarching long-term declining trend, highlighting the necessity for continuous monitoring of inequality trajectories in the post-pandemic period. In contrast, education-based inequality did not exhibit a clear and significant trend. This can be interpreted as a structural barrier, such as a lack of health literacy and insufficient awareness of the need for screening, which cannot be easily resolved through economic support alone [13, 52, 53]. To overcome this, multifaceted cognitive interventions must be implemented concurrently, such as introducing customised screening invitations tailored for less-educated populations (e.g. utilising visual aids and plain language) and enhancing recommendations from primary care physicians [55, 56]. Furthermore, the absence of an identifiable inflection point in the joinpoint analysis aligns with the non-significant interaction term between the pandemic period and socioeconomic indicators, reinforcing the robustness of the findings of this study.

According to the results of the C-index decomposition, in addition to income level, non-enrolment in private health insurance was identified as a consistent positive contributing factor, accounting for 15.5–31.3% of the total C-index throughout the study period. Furthermore, in the decomposition model excluding income level, this proportion increased to 17.2–32.3%. This highlights a dual-structured reality: individuals enrolled in private insurance can easily access more comprehensive, cost-intensive opportunistic screening through supplementary coverage, whereas low-income populations must rely solely on the NCSP [8]. This pattern suggests that a supplementary healthcare coverage structure centred on private insurance may be associated with greater polarisation in cancer screening participation, rather than mitigating socioeconomic disparities. Therefore, to enhance equity in cancer screening, multifaceted intervention policies targeting key vulnerable groups, such as those without private insurance, must be discussed along with strengthening the coverage of NCSP. Individuals without private insurance are likely to live in environments with limited access to healthcare resources. Consequently, expanding nationwide mobile screening programmes that directly reach vulnerable populations can be considered a complementary strategy to lower spatial and temporal barriers, going beyond merely subsidising costs at healthcare facilities [57]. Furthermore, targeted approaches that provide close-contact support from screening appointments to follow-up care through community-based outreach and navigation services are expected to contribute to bridging this gap [58–60].

The analysis of the pandemic and transition periods from 2020 to 2022 presents an intriguing contrast between the macroscopic and microscopic compositions of cancer screening inequality. As observed earlier, Korea’s universal health coverage system likely served as a buffer, limiting the drastic expansion of inequality even during a public health crisis. Indeed, according to the results of this study, the overall magnitude of the inequality indices did not significantly change during the pandemic [61, 62]. However, when examining the internal factors through the C-index decomposition results, periodic fluctuations were observed in the components contributing to inequality. During this period, the contribution of unemployment reached its highest level at 15.1% in 2021, and the contribution of Medical Aid beneficiaries increased to 13.6% in 2022. These findings support the necessity of policies aimed at compensating opportunity costs [8, 63, 64]. Although the elevated contributions of these employment- and coverage-related factors subsided after 2022, suggesting a largely transient shock at the aggregate level, they index labour-market and coverage vulnerabilities that can re-emerge during any future economic disruption. These patterns are better interpreted as indicators of durable structural vulnerability in preventive care for economically insecure populations rather than as a one-off pandemic effect [64]. This distinction has direct policy relevance: it implies that the opportunity-cost compensation mechanisms discussed above should operate as standing entitlements rather than emergency-only measures, so that vulnerable groups remain protected when the next economic shock occurs.

This study has some limitations. First, because KNHANES is a cross-sectional survey, causal relationships between socioeconomic factors and participation in cancer screening cannot be inferred; therefore, all findings should be interpreted in terms of association. In addition, given the nature of the repeated cross-sectional data, it was not possible to track the dynamic causal effects of long-term health trajectories or changes in socioeconomic status such as unemployment on cancer screening participation at the individual level. Therefore, longitudinal studies using panel data are required. Although annual ridit scoring accounts for changes in the relative size of socioeconomic groups, it cannot fully account for unobserved compositional changes within categories over time, particularly among older adults with low educational attainment. As the proportion of adults with elementary-school education or less declined, this group may increasingly represent a more socially, economically, or health-relatedly marginalised subgroup. Therefore, temporal changes in education-based SII and RII should be interpreted as reflecting both changes in screening inequality and possible within-category compositional shifts.

Second, the KNHANES cancer-screening item did not allow us to identify cancer type-specific screening participation across the entire study period, because cancer type-specific screening items were removed from KNHANES in 2013. In addition, the outcome combined participation in the NCSP and opportunistic private screening. Therefore, our estimates should be interpreted as socioeconomic inequalities in overall cancer-screening participation and broad screening access, rather than as cancer-specific NCSP uptake or the effect of a particular NCSP policy. This distinction is important because NCSP eligibility differs by cancer type, sex, age, and risk status, but the survey item did not allow cancer-specific or eligibility-specific analyses. Accordingly, the main analysis may reflect both inequalities in organised public screening and inequalities arising from access to opportunistic private screening. In the sensitivity analysis restricted to NCSP participation using the 2016 and 2019 data, when screening type could be distinguished, the magnitude of socioeconomic inequality was reduced compared with the main analysis. However, income-based inequality was not fully eliminated, suggesting that socioeconomic gradients may persist even under the more restrictive NCSP-only definition. These findings are consistent with the main conclusion but should be interpreted cautiously because the NCSP-only analysis was available for only two survey years.

Third, in addition to education and income, multidimensional socioeconomic factors that may explain health inequality, such as accumulated personal assets and housing stability, could not be sufficiently incorporated into the analytical model owing to data source limitations. Future studies should conduct more comprehensive decomposition analyses of inequality indices by incorporating additional variables.

Fourth, the C-index decomposition should be interpreted as a descriptive analysis of observed contributors to income-related inequality rather than as a causal decomposition. In particular, because household income ridit rank was used both to rank individuals for the C-index and as a covariate in the primary decomposition model, the contribution of the income-rank term may partly reflect the structure of the index itself. We therefore interpreted this component as the direct income-rank component and conducted a sensitivity analysis excluding household income ridit rank to examine the contributions of other observed covariates. In addition, because the decomposition is based on repeated cross-sections, the year-to-year change in a given covariate’s percentage contribution may arise from a change in its population prevalence, a change in its socioeconomic distribution, or a change in its marginal association with screening, and these components cannot be disentangled within the present design. The temporal fluctuations in individual contributions—such as the elevated contributions of unemployment and Medical Aid beneficiary status during the pandemic period—should therefore be interpreted as descriptive shifts in the composition of measured inequality rather than as evidence of a changing causal influence of any single factor over time.

Finally, KNHANES does not capture psychosocial determinants of screening such as cancer-related beliefs, perceived susceptibility, trust in healthcare institutions, and social support networks. These unmeasured factors may contribute to the persistence of education-based inequality even in a setting where financial barriers have been substantially reduced, and their omission should be considered when interpreting the residual educational gradient.

Despite these limitations, this study has several strengths. Using data from nine consecutive years (2016–2024) of a nationally representative sample from KNHANES, we conducted a robust trend analysis of socioeconomic inequality. Furthermore, by calculating the SII and RII based on annual ridit scores, we comprehensively captured the magnitude and direction of inequalities that were difficult to detect through simple comparisons of screening rates. Additionally, the simultaneous analysis of two distinct socioeconomic indicators—education and income levels—allowed us to identify the differentiated patterns of inequality that would remain obscured if we relied on a single indicator. However, future research should expand beyond the conventional socioeconomic indicators of education and income that were the focus of this study. Specifically, there is a need to analyse multidimensional interactions affecting cancer screening participation by incorporating various factors such as residential area, occupational status, and individual health literacy levels. Furthermore, conducting stratified analyses to examine the heterogeneity of subgroups by age group and region will enable a clearer identification of the factors associated with missed cancer screenings. Based on these insights, more precise and targeted intervention strategies should be developed.

Conclusions

Our findings showed that despite an overall improvement in cancer screening rates among Korean adults aged 40–74 years, socioeconomic inequality in cancer screening participation persisted. Notably, income-based inequality was more pronounced compared to education-based inequality. Furthermore, non-enrolment in private health insurance and employment-related vulnerability during the COVID-19 pandemic were associated with the observed pattern of inequality in screening participation.

Although income-based inequality in cancer screening participation narrowed over time, education-based inequality remained persistent. These findings suggest that expanded financial support may reduce some economic barriers, but additional strategies are needed to address non-financial barriers related to educational attainment. Therefore, to achieve substantive equity in cancer screening, a multidimensional and interministerial policy approach is required beyond the mere reduction of out-of-pocket costs. This approach should include the introduction of paid leave policies to compensate for the opportunity costs of screening, tailored health literacy interventions for individuals with lower educational attainment, and outreach and mobile screening programmes directly targeting marginalised populations.

Abbreviations

APC

Annual Percent Change

C-index

Concentration Index

CI

Confidence Interval

KNHANES

Korea National Health and Nutrition Examination Survey

NCSP

National Cancer Screening Programme

NHI

National Health Insurance

RII

Relative Index of Inequality

SII

Slope Index of Inequality

VIF

Variance Inflation Factor

Authors’ contributions

TW conceptualised the study, conducted the investigation, curated the data, performed the formal analysis, created the visualisations, and drafted the original manuscript. EK contributed to the interpretation of the results, critically revised the manuscript for important intellectual content, and assisted in drafting the policy implications. YW supervised the study, developed the methodology, and contributed to reviewing and editing the manuscript. All authors have read and approved the final version of the manuscript.

Funding

The authors did not receive any funding for this study.

Data availability

The datasets generated and/or analysed in the current study are available on the KNHANES website (https://knhanes.kdca.go.kr/knhanes/main.do).

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. Because this study used secondary data that did not contain any personally identifiable information, the requirement for written informed consent was waived. Ethical approval was obtained from the Institutional Review Board of the Yonsei University Mirae Campus (IRB No. 1041849-202604-SB-073-01).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.