The association between health literacy, private health insurance, and medical expenditure in South Korea

Health literacy (HL) refers to the ability to understand and process health information [1, 2]. Individuals with high HL are expected to effectively utilize health information, adopt healthy lifestyles, actively engage in disease prevention, and make appropriate use of healthcare services [3‐5]. The World Health Organization considers HL a crucial determinant of health [4, 6]. Individuals with high HL, equipped with a better understanding of the healthcare system, are more likely to follow medical advice and navigate complex healthcare environments to access necessary services [7‐9]. Consequently, these individuals are expected to have appropriate healthcare utilization, resulting in fewer hospitalizations due to ambulatory care sensitive conditions, fewer unnecessary emergency room visits, and lower medical expenses [10‐15].

Research on HL in Korea has been limited, particularly within the general population. While the Korea Health Panel Survey (KHPS) used in this study represents the first nationwide and systematic investigation of HL, only a few studies have explored this topic. Notably, there is limited understanding of the relationship between HL and personal healthcare expenditures, including insurance-related decisions. Among the previously conducted studies, one study that assessed HL using the Newest Vital Sign questionnaire found that factors associated with low HL included advanced age, low educational attainment, lower income level, limited access to health fairs, and low digital literacy [16]. When measured with this tool, although the survey was conducted in specific population groups, the HL level was lower compared to the United States [16]. In patients with mild cognitive impairment, HL showed a positive relationship with health-related quality of life [17]. Using a brief HL questionnaire in the Community Health Survey, it was also found that higher HL is associated with a lower likelihood of developing depression [18].

South Korea's universal healthcare system covers 98.5% of the population through National Health Insurance, with the remaining 1.5% receiving benefits via the Medical Aid Program [19]. Despite the introduction of universal health insurance in 1988, out-of-pocket expenses remain high, accounting for 20–30% of total medical costs [20]. Additionally, many medical services, including essential medical care, are not covered by insurance benefits [21, 22]. To address these gaps, indemnity health insurance was introduced in the 2000s to cover out-of-pocket and non-covered services [23].

Over the years, efforts have been made to enhance coverage, including the Reduced Co-payment Program for Severe Diseases, an out-of-pocket payment cap, and policies to include medically necessary non-covered services [24‐28]. Despite policies to strengthen public health insurance coverage, enrollment in private health insurance (PHI), particularly indemnity insurance, has continued to increase in Korea. Korean PHI is divided into fixed-payment insurance, which takes the form of sickness benefits, and indemnity insurance. While varying by insurance product and generation, indemnity insurance covers out-of-pocket expenses including non-covered services.

Studies have consistently reported moral hazard in healthcare utilization among indemnity insurance policy holders, leading to excessive use of medical services [29, 30]. This overutilization driven by indemnity insurance consequently results in increased medical expenditure. Furthermore, as insurance companies determine enrollment eligibility and premiums based on health status assessments and projected medical costs, individuals expected to have high medical utilization may be unable to enroll [31]. This potentially exacerbates inequalities in health and healthcare access.

Nevertheless, PHI, including indemnity insurance, serves a necessary complementary or supplementary role with its specific coverage benefits [32]. Therefore, it is crucial for citizens to understand the role and necessity of PHI appropriate to their circumstances. Unlike public health insurance, PHI requires population-level understanding of its specific functions and roles. However, research on this aspect remains limited in Korea.

The authors of this study suggested that while an interest in health may lead to the purchase of PHI, individuals with higher HL might be less likely to enroll in PHI due to a better understanding of the current government's ongoing efforts to strengthen insurance coverage. Thus, we emphasized the need to study the relationship between HL and PHI enrollment. Additionally, although higher HL is generally associated with lower healthcare spending, PHI enrollment could increase discretionary healthcare utilization [33], leading to higher costs. Therefore, the study aims to examine the relationships between HL, PHI enrollment, and OOP medical expenditure. To the best of the authors' knowledge, no studies have elucidated the relationship between enrollment in this type of insurance and HL in the Korean context.

Table 1 shows the distribution of study variables and the mean values of age and healthcare costs stratified by participants' PHI status. Indemnity insurance coverage was 70.05% of participants (p < 0.0001). The distribution of PHI policies was as follows: 11.98% was uninsured, 33.03% had one policy, 25.04% had two policies, and 29.96% had three or more policies (p < 0.0001). Within the inadequate HL group, 63.17% had indemnity insurance, while 71.47% of the sufficient HL group had indemnity insurance (p < 0.001). The uninsured rate was 16.31% in the inadequate HL group and 11.07% in the sufficient HL group. The distribution of PHI policies varied by HL level: those with sufficient health literacy had the highest proportion of single policy ownership (34.69%), while problematic HL was associated with the highest proportion of 3 or more policies (32.57%). The indemnity insurance rates were higher among women (73.91%) compared to men (65.65%), and among those with the highest income group (Q4) (75.16%), college education (72.86%), good SRH (71.26%), and no chronic diseases (72.02%). Annual medical expenditure (USD/year) was 357.68 (SE: 26.50) for those without indemnity insurance and 507.33 (SE: 21.07) for those with indemnity insurance. The annual medical expenditure increased as the number of policies increased, from 214.64 (0 policies) to 363.29 (1 policy), 496.97 (2 policies), and 642.20 (3 and more policies).

Table 2 presents the unadjusted odds ratios (OR) according to PHI status. The OR for indemnity insurance enrollment was 1.46 (95% CI: 1.19–1.80) in the sufficient HL group versus the inadequate group. The OR was 1.48 (95% CI: 1.29–1.71) for women versus men. The OR for Q4 household income was 2.05 (95% CI: 1.68–2.52) versus Q1. The OR for college graduates or higher was 2.14 (95% CI: 1.75–2.63) versus middle school education or lower. The OR for three or more chronic diseases was 0.47 (95% CI: 0.36–0.62). For number of PHI policies, the sufficient HL group had ORs of 1.76 (95% CI: 1.29–2.40), 1.54 (1.12–2.13), and 1.41 (95% CI: 1.03–1.91) for one, two, and three or more policies, respectively. The Q4 income group had ORs of 3.44 (95% CI: 2.45–4.84), 3.29 (95% CI: 2.33–4.65), and 4.03 (95% CI: 2.86–5.68) for one, two, and three or more policies.

Table 3 presents the adjusted odds ratios (aOR) according to indemnity insurance and number of PHI. In the adjusted models, HL levels showed no significant association with both indemnity insurance enrollment and number of PHI policies. College education or higher was significantly associated with both higher indemnity insurance enrollment (aOR 1.45, 95% CI 1.11–1.88) and ownership of 3 + PHI policies (aOR 1.62, 95% CI 1.06–2.47). The aOR for indemnity insurance enrollment was 0.74 (95% CI: 0.55–1.01) in those with 3 or more chronic diseases. For number of PHI policies, the aOR for having one PHI policy was 1.01 (95% CI: 0.62–1.65) in those with 3 or more chronic diseases.

Table 4 presents a generalized linear model analysis of OOP medical expenditure of indemnity health insurance status, HL level and other factors. In Model 1, the coefficient for indemnity insurance was 0.23 (p < 0.01, 25.9% increase). In Model 2, the coefficient for sufficient HL level was −0.21 (p < 0.01, 18.9% decrease). In Model 3, the coefficient was −0.22 (p < 0.01, 19.7% decrease,) for sufficient HL and 0.24 (p < 0.01, 27.1% increase) for indemnity health insurance. In Model 4, the coefficient for sufficient HL was −0.03 (not statistically significant), and for indemnity health insurance was 0.31 (p < 0.01, 36.3% increase). Additionally in Model 4, the coefficients were: 0.17 (p < 0.01, 18.5% increase) for women, 0.43 (p < 0.01, 53.7% increase) for Q4 household income, −0.21 (p < 0.01, 18.9% decrease) for both high school and college or higher education, 0.47 (p < 0.01, 60.0% increase) for poor self-rated health, and 0.65 (p < 0.01, 91.6% increase) for three or more chronic diseases.

Table 5 presents a generalized linear model analysis of OOP medical expenditure of number of PHI policies, HL level and other factors. In Model 1, the coefficients for number of PHI policies were 0.35 (p < 0.01, 41.9% increase) for one policy, 0.57 (p < 0.01, 76.8% increase) for two policies, and 0.79 (p < 0.01, 120.3% increase) for three or more policies. In Model 3, the coefficients for number of PHI were 0.37 (p < 0.01, 44.8% increase), 0.59 (p < 0.01, 80.4% increase), and 0.81 (p < 0.01, 124.8% increase) for one, two, and three or more policies respectively, while sufficient HL level showed a coefficient of −0.24 (p < 0.01, 21.3% decrease). In the fully adjusted Model 4, the coefficients for number of PHI were 0.33 (p < 0.01, 39.1% increase) for one policy, 0.54 (p < 0.01, 71.6% increase) for two policies, and 0.73 (p < 0.01, 107.5% increase) for three or more policies. The coefficient for sufficient HL level in Model 4 was −0.05 (not statistically significant). Additionally in Model 4, the coefficients were: 0.35 (p < 0.01, 41.9% increase) for Q4 household income, −0.24 (p < 0.01, 21.3% decrease) for high school education, −0.23 (p < 0.05, 20.5% decrease) for college or higher education, 0.46 (p < 0.01, 58.4% increase) for poor SRH, and 0.60 (p < 0.01, 82.2% increase) for three or more chronic diseases.

Multicollinearity diagnostics revealed a moderate correlation between number of PHI and indemnity health insurance (Cramer's V = 0.57), while all other pairs of categorical variables showed weak correlations (V < 0.28). Age showed significant correlations with most categorical variables (p < 0.001), particularly with educational level (r = −0.53) and number of chronic diseases (r = 0.46). The VIF values ranged from 1.05 to 3.77 (Supplementary Tables 3, 4, and 5).

This study examined the relationship between HL, PHI, and annual OOP medical expenditure using KHPS data. Results indicated that women, higher income levels, and higher educational attainment were associated with increased enrollment in indemnity health insurance and a greater number of PHI policies. Conversely, there was a tendency for individuals with a higher number of chronic diseases and poorer SRH to be less likely to enroll in PHI.

Our findings demonstrated that individuals with sufficient HL showed ORs greater than 1 in the unadjusted model for both indemnity insurance enrollment and number of PHI subscriptions, but after adjusting for other variables, the statistical significance of the ORs disappeared. Regarding annual OOP medical expenditure, the Gamma GLM with log link function showed that models including HL level alone or both HL level and PHI status demonstrated significant cost reductions (18.9% and 19.7% decrease in the indemnity insurance model, and 18.9% and 21.3% decrease in the number of PHI policies model) when HL level was sufficient. However, when health-related and socioeconomic variables were included in the models, the coefficient for sufficient HL level lost its statistical significance.

Studies examining the relationship between HL and PHI enrollment remain limited in the literature. Research conducted in Germany demonstrated an association between low HL and high PHI enrollment [36]. In the United States, numerous studies have investigated the relationship between health insurance literacy (HIL) and PHI enrollment, finding that higher HIL correlates with increased Medicare Advantage enrollment rates [37]. However, these findings must be interpreted within their respective healthcare system contexts. In Germany's insurance system, which is primarily based on Social Health Insurance (SHI), individuals with higher income levels must choose between SHI and PHI. However, the situation regarding PHI in South Korea differs from that in Germany. While PHI in Germany is primarily substitutive [36], in South Korea it is characterized as complementary and supplementary. The concept of HIL in the United States aligns closely with informed choice-making [37]. The US healthcare system is primarily based PHI. Furthermore, unlike the United States, South Korea has lower healthcare accessibility barriers and generally more affordable medical costs. In essence, HL interacts with various healthcare systems in distinct ways, influenced by the specific characteristics of each system [38].

Previous studies indicate that HL is associated with lower medical expenditure. These studies analyzed the effect of HL on the reduction of medical expenditure because individuals with high HL tend to focus on health promotion, maintain healthy lifestyles, and use preventive medical services [11, 12]. Consequently, they are less likely to visit medical institutions due to health deterioration. In our study, while models including only HL and PHI showed that higher HL levels were associated with lower annual OOP medical expenditure, coefficient for sufficient HL level lost its statistical significance in the fully adjusted model. This suggests that in Korea, HL may not independently influence OOP medical expenditure, but rather represents a complex factor that should be considered in conjunction with other socioeconomic determinants.

The OOP medical expenditure models revealed that higher income levels were associated with increased OOP medical expenditure, while higher educational levels were associated with decreased expenditure. However, the OR for education level and household income remained statistically significant across both crude and adjusted models for PHI enrollment. Given these results and the significant inverse relationship between educational level and OOP medical expenditure, this suggests that educational attainment may substantially explain the observed HL effects which are use of preventive medicine, health promotion and effective healthcare utilization.

In models examining the relationship between the number of PHI policies and OOP medical expenditure, a positive association was observed; a higher number of policies corresponded with increased OOP medical expenditure. Similarly, individuals enrolled in indemnity health insurance demonstrated significantly higher OOP medical expenditure compared to those without such coverage. PHI enrollment has been shown to increase discretionary medical utilization [33]. The positive relationship between PHI and OOP medical expenditure found in this study is likely to be in line with these previous findings [21, 39, 40].

The over-utilization of medical care induced by indemnity insurance will lead to a depletion of public insurance finances. Exacerbating this situation is the supply-induced care provided by some healthcare providers [41, 42]. Under the fee-for-service system, indemnity insurance effectively reduces patients' out-of-pocket expenses for non-covered services to zero, creating an environment where healthcare providers are incentivized to establish private clinics where they can generate higher revenues through increased provision of non-covered services [43‐45]. This has led to a "brain drain," which is a concerning trend where healthcare providers are increasingly leaving essential medical fields to open private clinics that offer more profitable non-covered services alongside covered ones [43‐45]. This has resulted in a reduction of medical professionals responsible for treating serious conditions in essential care sectors. While PHI poses difficulties for Korea's healthcare system, France shows a more harmonious approach where PHI serves as complementary insurance for minor illnesses and SHI covers serious conditions [46]. This approach is particularly significant because while Koreans primarily purchase PHI to protect against catastrophic healthcare costs [47], insurance companies engage in risk skimming by excluding high-risk individuals from PHI coverage [48].

However, this study was conducted during the COVID-19 pandemic, when decreased discretionary healthcare utilization and elective procedures led to reduced medical costs [49]. Additionally, the COVID-19 pandemic may have become an opportunity for people to become interested in health-related information and proactively manage the health-related information [50]. It is possible those with higher HL reduced their discretionary health care use. While HL levels are unlikely to have decreased post-pandemic, discretionary care utilization may have increased again. Therefore, additional research is needed in the post-COVID-19 period to examine these relationships.

This study, however, has several limitations. First, it did not classify medical utilization into preventive care, appropriate care, and excessive care. As such, the mechanisms through which PHI enrollment and HL affect healthcare expenditures remain unclear. Second, there was insufficient adjustment for health status. Previous studies using the KHPS have adjusted for health status using the Charlson Comorbidity Index (CCI). However, as the KHPS transitioned to its second panel and restructured its data, it was not possible to calculate the CCI. Third, while indemnity insurance has enrollment restrictions for those aged 65–70, excluding those over 65 may introduce bias to this study, particularly in models examining the number of insurance policies. Forth, HL was measured using HLS-EU-Q16. While this tool has not been validated for measuring HL in Korea and may be less accurate due to its self-reported nature, it offers the advantage of enabling international comparisons of health literacy levels as a widely studied instrument. Fifth, there is an issue regarding the treatment of respondents who answered "don't know" to HL questions. Given the nature of these questions and responses, those who answered "don't know" are likely to have lower HL. However, for precision, this study excluded such respondents. This may have led to an underrepresentation of the low HL group, suggesting that our study might have overestimated overall HL levels. However, this sampling bias may have resulted in a conservative estimation of the relationship between HL and OOP medical expenditure.

Despite these limitations, this study holds significant value as the first to examine the complex relationships among HL, PHI enrollment, and OOP medical expenditure. Despite its universal coverage, these issues include PHI-associated increases in OOP medical expenditure with high enrollment rates of PHI, and lower PHI enrollment among individuals with poor health status. The disappearance of the relationship between HL and OOP medical expenditure reduction when controlling for other variables, particularly the observed reduction in OOP medical expenditure with educational level, suggests that HL is associated with educational level and indicates the need for public intervention in HL. Further research is required to explore these findings in greater depth.