Supplementary Private Health Insurance and Household Debt, Health Care Utilization, and Medical Spending Following A Health Shock

Abstract

This article aims to evaluate the effect of enrolling in supplementary private health insurance on household debt, medical spending, and medical service use among South Koreans experiencing a health shock. Using data from the Korean Welfare Panel Study from 2009 through 2017, we compared household debt and health service use for those with and without private supplemental health insurance after experiencing a health shock. We found no significant differences in household debt or the financial burden of a health shock between those with and without supplemental health insurance coverage following a shock. Households with supplemental coverage used more medical services compared to households without supplementary coverage and incurred additional medical expenses.

Adverse health shocks are the main cause of financial burden for households in South Korea, leading to increased mortgage and household debt and, in extreme cases, bankruptcy.^1–3 The primary rationale of universal health insurance is to provide protection against such risk; however, because compulsory social insurance often fails to satisfy all medical needs, supplementary health insurance has become an increasingly popular option.^4–7

Despite achieving universal coverage in 1989, the National Health Insurance program has never provided comprehensive coverage with cost sharing ranging from 20 percent to 60 percent, depending on the type of care and place of service. ⁸ Further, the Medical Aid program ¹ , targeted at low-income groups, only covers 3 percent of the total population. ⁹ In 2017, efforts were made to expand National Health Insurance coverage. Key elements of the expansion included integrating all uncovered medical items into the National Health Insurance program, reducing medical expenses for vulnerable populations, and strengthening the medical safety net. ⁸ The expansion aimed to increase the public funding share of health spending by up to 70 percent, specifically incorporating non-covered services required for exams and treatments, such as ultrasonography and magnetic resonance imaging. ¹⁰ Additionally, the expansion aimed to eliminate three major non-covered items: optional medical costs ² , advanced hospital room fees, and nursing fees, which together accounted for 60 percent of non-covered spending. ¹¹ Following the expansion, the government's share of total health expenditures increased from 59.0 percent in 2016 to 62.6 percent in 2020, but continues to lag behind the 2019 OECD average of 74 percent. ¹²

These coverage gaps in the National Health Insurance program have led to the introduction and growth of supplementary private health insurance in South Korea. Private health insurance in South Korea can be purchased as a standalone product or as a supplement to life or accident insurance, and typically covers inpatient services for high-cost conditions like cancer. Benefits are often paid out in the form of lump sum cash payments for surgeries or per diem payments for hospital admission. ¹³ Prior to 2002, only non-life insurance companies could provide private supplementary health insurance coverage and uptake was low due to limited plan variety and insufficient managerial capabilities for the companies to administer such plans effectively. ¹⁴ Following a revision to the Insurance Business Act in 2002, life insurance companies were allowed to offer supplementary private health insurance coverage, leading to major life insurance providers introducing comprehensive packages that combined life insurance coverage with critical illness and medical expense benefits. ¹⁵ As of 2020, the subscription rate for private insurance was 78.61 percent, and the share of private insurance financing of health expenditures had risen from 3.9 percent in 2008 to 9.7 percent in 2020. ¹⁶

The recent surge in private coverage in South Korea has raised concerns over the role of private health insurance in the country's health care system. Some have argued that an enhanced role for private health insurance could alleviate the financial burden facing the National Health Insurance system and enhance efficiency through competition. ¹⁷ However, concerns have also been raised about the possible negative effects of private coverage, including the risk of moral hazard, issues of adverse selection, and consequences for equity.^18,19 Within the South Korean government, divergent views on private health insurance exist, as the Ministry of Health and Welfare has expressed concern regarding potential adverse impacts, while the Ministry of Economy and Finance considers private health insurance an opportunity to broaden the scope of the financial sector. ²⁰ A central controversy concerning private health insurance in South Korea is its impact on health care use. Private health insurance coverage reduces out-of-pocket expenses associated with cost-sharing under the National Health Insurance program thereby encouraging patients to utilize more health care services. ²¹ This increased utilization can lead to fiscal spillover effects on the National Health Insurance system and contribute to disparities in health care usage between those who hold private insurance and those who do not. ²⁰ Furthermore, private health insurance offers the greatest potential welfare benefits for the elderly and low-income groups due to their high financial vulnerability in times of illness, though these groups tend to have the lowest enrollment rates in private health insurance, which raises concerns over equity.^18,22

Despite the popularity of supplementary private insurance and efforts to increase National Health Insurance coverage, out-of-pocket payments comprise nearly 30 percent of health expenditures in South Korea, in some cases causing catastrophic medical debt and seriously disrupting household spending patterns.^8,23 Concerns have been raised regarding the ability of private health insurance to effectively protect its members from financial burden and the risk of impoverishment from medical debt. ²⁴

Previous studies of the benefits of supplementary private health insurance in South Korea have mainly focused on medical expenditures. Criticism has been levied that prior research has underestimated the total burden of disease, not accounting for nonmedical expenses such as transportation, patient time, informal or unpaid caregiving, and lost productivity.^3,25,26 To address this issue, O’Donnell ²⁶ raised the need for restructuring the concept of health insurance coverage, suggesting that it should not be limited to medical expenses but also consider how those expenses impact nonmedical spending to ensure that households are able to manage medical costs in a sustainable manner over time. ²⁶ Evaluating household debt as a component of the total burden of an adverse medical event can provide insight into the total burden of disease and how households respond to ill health.

This study seeks to address the need to evaluate the effectiveness of private health insurance in mitigating the financial burden of health shocks, which prior research has not fully captured. To do so, we focused on the impact of private health insurance coverage on household debt following a health shock, when the financial burden of medical care and lost wages is expected to increase dramatically.

We compared changes in several measures of household debt for those with and without private supplemental health insurance in addition to standard National Health Insurance coverage following the onset of a health shock. We hypothesized that private health insurance coverage would offer greater protection against financial burdens and leave those with private coverage less vulnerable to the financial risk of a health shock. In addition, we considered whether reduced out-of-pocket cost sharing associated with private insurance coverage would alter medical spending and health care utilization following a health shock.

This study provides an important empirical assessment of the effect of enrolling in supplementary private health insurance in a setting in which a public program exists but is criticized for insufficient coverage. Using detailed measures of household debt, we provide insight into the total burden of disease and how households respond to adverse health events. Results provide important evidence concerning the role of supplementary private health insurance, particularly in settings where compulsory social insurance often fails to satisfy all medical needs. Our study is timely given the growing enrollment in in private supplementary insurance plans in countries with national health insurance programs and aims to improve our limited understanding of the impact of supplementary coverage on health care access and financial security.

Methods

Data and Population

This study used the Korean Welfare Panel Survey from 2009 to 2017. The Korean Welfare Panel Survey provides yearly aggregated data at the household level on a representative sample of the Korean population. The survey panel was constructed using a stratified double sampling model that selected 24,711 households from 517 enumeration districts in the 2005 census and followed these same households over time beginning in 2006. Trained interviewers conducted face-to-face home visits with the selected panel of households. The survey collected general and socioeconomic information, including data on income, living expenses and conditions, economic activity, social insurance status, and the use of medical and welfare services by households and their members. The survey period covers January 1 to December 31 each year for flow variables, and as of December 31 each year for stock variables.

We defined a health shock as the onset of one of the following diseases: cancer, cardiovascular disease, cerebrovascular disease (stroke), and rare and incurable disease. Rare and incurable disease, as defined by the Rare Disease Management Act, is a condition with a prevalence of fewer than 20,000 patients or one that is difficult to diagnose. ²⁷ As of 2023, 1,248 diseases have been classified as rare and incurable. In selecting these conditions, we followed Smith's (2005) definition of health shock to include only severe diseases that typically result in a sudden decline in an individual's health. ²⁸ These four disease categories are recognized as the top four high-burden disease groups in South Korea due to their high prevalence rates and associated costs. ²⁹ These four disease categories constituted 3.4 percent of all physician visits in 2012, yet they represented 28 percent of the overall expenditures of the National Health Insurance program. ³⁰ We limited the initial sample to households that experienced a health shock from 2011 through 2015 to ensure that there was no additional health shock two years before the onset of the index shock and to allow follow up on the changes in outcomes two years after the occurrence of the health shock. We excluded those with multiple shocks to eliminate the possibility that a preceding health shock could affect the index shock and to remove potential reverse causality in which the worsened household financial circumstances due to preceding health shock caused the subsequent shock. Additionally, households who enrolled in private health insurance coverage after the onset of the health shock were excluded from the analysis. After these exclusions, our sample included 842 households that experienced a health shock during our sample period, 321 of which had private supplemental health insurance coverage and 521 of which did not. Households were grouped into treatment or control status based on whether they reported supplementary private health insurance prior to the onset of a health shock. Figure 1 shows the sample extraction process.

Figure 1.

Flow chart showing data extraction from the Korean welfare panel survey data.

We included both Medical Aid beneficiaries and National Health Insurance members since the medical costs of non-covered services impact the household financial circumstances of both groups. Both groups share the same service coverage, and the main difference between the two lies in the copayment and coinsurance required to utilize health care services. Consequently, each group requires out-of-pocket spending to utilize non-covered services. The estimated annual out-of-pocket spending per individual was 299,010 KRW for those covered by the National Health Insurance and 80,040 KRW for Medical Aid beneficiaries. ³¹

Outcomes

The study assessed the impact of a health shock on several outcome measures, including household debt, household medical costs, and medical service utilization. Household debt was assessed using five measures: debt borrowed from financial institutions, debt borrowed from nonfinancial institutions and private lenders, credit card debt, any other debt, and the total sum of all debts. Medical costs were measured as any cost incurred by utilizing inpatient and outpatient services, dental services, traditional Korean medicine services, and medications. Medical service utilization was measured using outpatient visits, inpatient visits, and inpatient days.

Statistical Analysis

Our analysis method compared the financial and medical care implications of a health shock for those with and without private supplemental health insurance. However, as private health insurance is not randomly assigned, we were concerned with potential confounding related to both the propensity to enroll in private coverage and changes in our outcomes. To mitigate this concern, we applied a matching technique known as entropy balancing that assigned unit weights to our control observations based on the distribution of the means, variance, and skewness of the covariates for our treatment units. ³² Entropy balancing is a robust alternative to the weighting estimators generated by propensity score matching, since entropy balancing directly generates weights based on the distributions of covariates, bypassing the process of estimating propensity scores and thus minimizing potential estimation errors.^32,33

Characteristics used in the entropy balancing included family type (single, non-single), number of household members, household head sex, household head employment status (employed, unemployed, or economically inactive), the share of household members with bad self-reported health status, the share of household members with chronic disease, and the share of household members with disability. To ensure that the included covariates were balanced between the groups with private insurance coverage and those without, we employed standardized mean differences as a verification measure. After calculating weights using the entropy balancing technique, we estimated difference-in-differences models to examine changes in household debt and medical costs among those with and without private health insurance over the pre-and post-health shock periods. The health shock was centered at t = 0, with two years before and after as t = -2, t = -1, and t = 1, and t = 2, respectively. We centered each health shock at t = 0 to a minimum of two years follow-up time and to remove any households experiencing a shock in the two years leading up to the index event. This approach allowed us to compare changes over the same time horizon between households with and without private health insurance after the onset of a health shock.

$${\mathrm{Y}}_{\mathrm{it}}=\alpha +\beta Privat{e}_i\times Shoc{k}_t+{\delta }_i+{\gamma }_t+{\epsilon }_{it}$$

Where $Privat{e}_i$ is an indicator for whether individual i has private insurance; $Shoc{k}_t$ is an indicator for whether time t is pre- or post-shock. δ represents a vector of household fixed effects and γ represents time fixed effects. The coefficient $\beta$ represents the difference-in-difference estimate of changes to our outcomes of interest for those with private health insurance following a health shock compared to those with no private insurance. Note that a standalone indicator for $Privat{e}_t$ would be subsumed by our household fixed effects and a standalone indicator for $Shoc{k}_t$ by our time fixed effects. All models were weighted using the weights generated through the entropy balancing technique.

We used a two-part model to address the issue of excessive zero data structures in outcomes. The first part of the model employed a linear probability model to estimate changes in the extensive margin for each outcome (e.g., any debt). The second part used the same approach but estimated the outcomes conditional on households having nonzero outcomes incurred after the shock.

Robustness Checks

The validity of our difference-in-differences models depend on whether, in the absence of a health shock, trends in our outcome measures for those with and without private health insurance coverage would have followed a stable path through the post-period. While the parallel trends assumption is untestable, we provide evidence that it holds by estimating an event-study design that compared trend changes between the two groups over time.

In 2017, the South Korea expanded the National Health Insurance scheme to included coverage for additional services and reduce copayments for members. ⁸ To ensure that the expansion did not affect the outcomes we study, we performed an additional robustness check where we excluded those households that experienced a health shock in 2015. Since we follow individuals in our sample over a two-year period, the t + 2 period for those experiencing a health shock in 2015 would coincide with the National Health Insurance coverage expansion. By dropping this group from our sample, we ensure that our estimates are not confounded by this coverage change.

We also performed further subgroup analyses to test the robustness of our results. We excluded Medical Aid beneficiaries who have reduced or no cost sharing, potentially limiting the financial impacts of a health shock for National Health Insurance members.

Results

Descriptive Statistics

Table 1 displays the descriptive characteristics of the study population separately for those with and without private health insurance. The results indicate that those with private health insurance had a better health and socioeconomic status compared to those without. On average, those with private supplemental coverage subscribed to 1.73 health insurance plans with an average monthly premium of 31,170 KRW (approximately US$23).

Table 1.

Descriptive Statistics of the Study Population.

	Nonsubscriber of PHI (N = 521)		Subscriber of PHI (N = 321)
	Mean/%	SD	Mean/%	SD
Household demographics
family type, single	35.51%	0.479	7.17%	0.258
number of Household members	1.889	0.962	3.123	1.357
household annual income, 10,000 KRW	1778.676	1750.884	4742.979	3393.495
monthly household total living expenditure, 10,000 KRW	145.868	125.956	362.812	243.412
health coverage, Medical Aid	16.12%	0.368	4.05%	0.197
Head of household demographics
gender, male	74.66%	0.435	86.92%	0.338
age	71.621	11.158	56.71	12.308
employment type, unemployed or economically inactive	65.83%	0.475	30.84%	0.463
Household health status
share of household members with self-reported health status below bad†	34.80%	0.46	13.60%	0.247
share of household members with chronic disease	63.30%	0.482	41.20%	0.373
share of household members with a disability	25.00%	0.349	11.20%	0.206
Private health insurance
number of private health insurance subscribed			1.731	3.117
monthly private health insurance premium, 10,000 KRW			3.117	18.732
Health shock type
cancer	24.95%	0.433	37.69%	0.485
coronary heart disease	31.67%	0.466	23.36%	0.424
cerebrovascular disease (stroke)	31.48%	0.465	22.74%	0.420
rare and incurable disease	11.90%	0.324	16.20%	0.369
Debt
all debt	815.432	3129.169	2502.710	8083.748
financial institution debt	665.192	2937.171	2154.944	7706.794
card debt	41.747	466.605	92.031	1400.839
non-financial institution and private debt	7.303	100.692	12.835	124.591
other debt	63.628	464.238	122.249	819.288
Medical cost and utilization
medical cost	26.034	37.203	31.663	40.914
outpatient visit	49.679	48.649	47.137	49.787
inpatient visit	0.919	1.890	0.907	1.957
hospitalized days	19.021	49.368	13.078	28.327

SD: standard deviation; PHI: supplementary private health insurance; 10,000 KRW is approximately US$7.5.

† Health status was measured on a scale of excellent, very good, good, bad, and very bad. The percentage represents the proportion of household members who reported their health status as either bad or very bad.

Figure 2 illustrates the standardized mean differences in covariates between those with and those without private health insurance before and after entropy balancing. Supplementary Table 1 shows the mean, variance, and standardized mean differences in covariates before and after applying entropy balancing. Prior to applying entropy balancing, all covariates had significant differences in standardized means exceeding 0.1. However, after applying entropy balancing, the standardized mean differences in all covariates were balanced to nearly zero.

Figure 2.

Standardized mean difference in covariates included in the entropy balancing. The red dot represents the standardized mean difference before adjustment, while the black dot represents the standardized mean difference after adjustment. A hollow dot indicates that the differences were not statistically significant at p < 0.05, whereas a solid dot indicates that the differences were statistically significant at p < 0.05. For example, a hollow black dot represents the standardized mean difference after adjustment, where the difference was not statistically significant at p < 0.05.

Impact of Supplementary Health Insurance Subscription in Case of Health Shock

Household Debt

The results of the analyses of the impact of private health insurance coverage on household debt following a health shock are presented columns 1–10 in Table 2. We found no statistically significant effect of private health insurance coverage on the extensive margin of total debt accumulation, financial institution debt, credit card debt, non-financial institution/private debt, or other debt. The results were generally constant across the subgroup analyses. Event study results were consistent with the difference-in-difference estimates, showing no significant difference in the outcomes in the post-health shock period and no evidence of diverging pre-period trends.

Table 2.

Effect of Private Health Insurance Subscriptions on Household Debt and Household Asset in Health Shock.

		All debt		Financial institution debt		Card debt		Non-financial institutions and private debt		Other debt
		LPM	OLS	LPM	OLS	LPM	OLS	LPM	OLS	LPM	OLS
		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Full sample	DD estimate	−0.003	394.336	−0.045	−299.144	−0.004	−29.603	−0.022	528.060	0.055*	914.174*
	SE	0.044	1,312.967	0.041	1,100.711	0.018	82.415	0.015	646.084	0.030	534.274
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
NHI only	DD estimate	−0.014	349.335	−0.043	−505.397	−0.017	−47.787	−0.010	940.679	0.037	1,135.836
	SE	0.048	1,551.461	0.045	1,293.010	0.017	48.695	0.012	776.976	0.026	712.022
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Exclude 2015	DD estimate	−0.002	228.387	−0.055	−448.819	−0.003	−31.647	−0.028*	369.519	0.071*	925.514*
	SE	0.050	1,404.702	0.047	1,202.170	0.021	88.663	0.017	640.595	0.034	549.392
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
NHI only and Exclude 2015	DD estimate	−0.017	153.322	−0.053	−713.858	−0.019	−60.322	−0.014	827.111	0.049	1,162.073
	SE	0.056	1,671.968	0.052	1,428.240	0.021	53.866	0.014	777.068	0.030	745.117
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Household fixed effect		YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Year fixed effect		YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Baseline mean		0.177	699.568	0.113	523.670	0.007	7.270	0.018	52.898	0.052	81.197
SD		0.382	3182.499	0.317	2776.019	0.086	117.075	0.136	696.233	0.222	1009.947

Unit of debt for the Ordinary Least Squares model: 10,000 KRW (approximately US$7.5); LPM: linear probability model; DD estimate: difference-in-difference estimate; SE: standard error; NHI: National health insurance; Exclude 2015: excluded households who faced health shock at 2015; Baseline mean indicates the mean of the outcomes on the pre-shock period among the nonsubscribers of the supplementary private health insurance; SD: standard deviation; *p < 0.1 **p < 0.05 ***p < 0.01.

Number in the LPM column indicates difference in the probability of incurring any amount of debt for each debt category relative to nonsubscribers (e.g., LPM column 1 for the full sample row indicates a 0.3% decrease in the probability of incurring any amount of debt for the subscribers compared to nonsubscribers). Number in the OLS column indicates the estimated difference in the amount of incurred debt for subscribers compared to nonsubscribers, excluding those who did not incur any debt in each of the debt category (e.g., OLS column 2 for the full sample row indicates an increase of 394,336 KRW in all debt among subscribers compared to nonsubscribers).

Household Medical Spending

We found no evidence of differences in household medical spending between those with and without private health insurance coverage following a health shock (Table 3, columns 1 and 2). However, the event study reveals that those with private health insurance coverage experienced a significant increase in medical costs of 122,000 KRW (approximately US$94.38; standard error [SE]: 2.630) at the onset of a health shock (t = 0). This increase in medical costs was consistent across all subgroups analyzed.

Table 3.

Effect of Private Health Insurance Subscriptions on Medical Spending and Health Care Utilization.

		Medical cost		Outpatient visit		Inpatient visit		Hospitalized days
		LPM	OLS	LPM	OLS	LPM	OLS	LPM	OLS
		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Full sample	DD estimate	0.012	−1.057	0.005	14.036**	−0.074	−0.076	−0.070	−3.981
	SE	0.035	3.477	0.010	7.013	0.059	0.136	0.059	3.867
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES
NHI only	DD estimate	−0.001	−1.971	0.004	13.971*	−0.081	−0.078	−0.076	−4.395
	SE	0.039	3.948	0.009	7.796	0.065	0.146	0.065	4.183
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES
Exclude 2015	DD estimate	0.018	−0.935	0.004	14.852*	−0.047	−0.038	−0.044	−2.807
	SE	0.039	3.819	0.010	8.153	0.064	0.148	0.064	4.333
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES
NHI only and Exclude 2015	DD estimate	0.004	−1.906	0.003	14.760	−0.051	−0.038	−0.047	−3.002
	SE	0.045	4.417	0.010	9.199	0.073	0.160	0.072	4.722
Parallel trend assumption		YES	YES	YES	YES	YES	YES	YES	YES
Household fixed effect		YES	YES	YES	YES	YES	YES	YES	YES
Year fixed effect		YES	YES	YES	YES	YES	YES	YES	YES
Baseline mean		0.715	11.531	0.754	42.203	0.231	0.340	0.229	7.086
SD		0.451	1009.947	0.430	56.800	0.421	0.735	0.420	26.017

Unit of medical cost for the Ordinary Least Squares model: 10,000 KRW (approximately 7.5 USD); LPM: linear probability model; DD estimate: difference-in-difference estimate; SE: standard error; NHI: National health insurance; Exclude 2015: excluded households who faced health shock at 2015; Baseline mean indicates the mean of the outcomes on the pre-shock period among the nonsubscribers of the supplementary private health insurance; SD: standard deviation; *p < 0.1 **p < 0.05 ***p < 0.01.

Number in the LPM column indicates the difference in the probability of incurring any amount of medical cost, any number of outpatient/inpatient visits or hospitalized days (e.g., LPM column 1 for the full sample row indicates a 1.2% increase in the probability of incurring any amount of medical cost for the subscribers compared to nonsubscribers). Number in the OLS column indicates the estimated difference in the amount of incurred medical cost, the number of outpatient/inpatient visits or hospitalized days for subscribers compared to nonsubscribers, excluding those who did not incur any medical cost or had no outpatient and inpatient visit or hospitalized days (e.g., OLS column 2 for the full sample row indicates a decrease of 1,057 KRW in medical cost among subscribers compared to nonsubscribers).

Household Medical Service Utilization

Columns 3–8 of Table 3 present the impact of private health insurance coverage on household inpatient and outpatient service utilization. Although having private health insurance did not increase the likelihood of incurring an outpatient visit, our analysis revealed a significant increase in number of household outpatient visits among those with private health insurance conditional on any visit, with second part of the two-part model showing 14.036 more visits per household (SE: 7.013) in the post-shock period compared to those without private coverage. However, the statistical significance becomes weaker in the subgroup analyses. Our difference-in-differences estimate indicated no statistically significant difference in inpatient service use, though the event study showed a significant increase in household inpatient service use at the onset of the health shock (t = 0) among those with private coverage. The event study findings suggest that having private supplementary coverage was associated with an increase in inpatient visits. In the first part of the model, at t = 0, households with supplementary private health insurance coverage were 17.38 percent (SE: 0.043) more likely to experience an inpatient visit. On the intensive margin, private health insurance was associated with an additional 0.572 (SE:0.158) inpatient visits. Similarly, at t = 0, private supplementary coverage led to an increase of 7.423 (SE: 2.815) more hospitalized days. The results from the event study were consistent across subgroup analyses.

Robustness Checks

Parallel Trends Assumption

Figure 3 displays the event study estimates of changes over time for total household debts. Supplementary Figures 1 through 11 show the event study estimates of changes for outcomes in other outcomes and subgroup analyses. We found no indication of diverging pre-period trends for either the base model or the subgroup analyses.

Figure 3.

Event study: all debt. Figure shows the estimated differences in the Linear Probability Model (first) and Ordinary Least Squares Model (second) comparing subscribers to nonsubscribers. The first indicates the first part of a two-part model, with linear probability model to estimate the changes in the extensive margins of the outcomes. The second indicates the second part of a two-part model, with linear regression to estimate the outcomes conditional on households having nonzero outcomes incurred after the shock. The middle lie of the figure on the left (“first”) indicates the estimated difference in the probability of incurring nonzero outcome for the LPM model. The middle line of the figure on the right (“second”) indicates estimated difference of incurred outcome, condition to households who had nonzero outcome for the OLS model. Upper and lower dashed lines indicate the upper and lower boundaries of the 95% confidence interval for the estimated difference.

Robustness Checks on Excluding Households Affected by the 2017 National Health Insurance Coverage Expansion

The “Exclude 2015” column in Table 2 and Table 3 represents the subgroup analyses that exclude households affected by the 2017 National Health Insurance coverage expansion. The findings from these subgroup analyses are consistent with the results of the base model.

The column titled “NHI only” in Table 2 and Table 3 indicates the subgroup analyses that exclude households under the Medical Aid program. The outcomes from these subgroup analyses align consistently with the base model results. The same consistency was observed when we excluded both the households affected by the National Health Insurance expansion in 2017 and those under the Medical Aid program.

Discussion

Our study found no evidence that those covered through South Korea's National Health Insurance program with additional supplementary private health insurance coverage were better protected from the financial burden of a health shock than those with no supplementary coverage. This could be attributed to the limited effectiveness of supplementary private health insurance in reducing financial risk, as has been indicated in previous studies.^24,34 Another explanation could involve the implementation of a National Health Insurance coverage expansion directed specifically toward the four disease groups we studied. Recognizing these four conditions as the top contributors to disease burden in South Korea, the Korean government implemented an extension of the National Health Insurance coverage during the third welfare plan (2013–2017) that minimized out-of-pocket payments and bridged service gaps for individuals suffering from these four disease groups. ³⁵ Since this study defined a health shock as the occurrence of one of these four diseases, it is possible that the coverage expansion effectively protected all households from the financial implications of these health shocks. Consequently, households dealing with these diseases might not have needed the extra financial security provided by supplementary private health insurance. However, prior studies assessing the impact of this coverage expansion on household medical expenditures related to these same diseases found that the expansion failed to significantly relieve the financial burden on households or address the unmet needs associated with these four diseases, thus leaving room for private supplementary coverage to offer improved financial protection.^29,36,37 Either explanation is consistent with our finding that supplementary private insurance had no effect on providing additional financial protection to its subscribers experiencing one of the health shocks we study. Thus, our findings contribute to the existing empirical body of evidence on the role of private health insurance in protecting its subscribers from financial risk. Even with limiting the population to households that experienced a health shock and tend to face a dramatic increase in financial burden from higher medical spending and lost wages, we found no evidence to suggest that private health insurance protected its members from financial burden.

Health shocks affect households’ financial circumstances by generating unplanned medical expenses and reducing wages.^26,38 Households choose how much to spend on health care based on the share of medical expenses and lost wages covered by their health insurance. If the expenses exceed savings, households may adopt cost-prevention or cost-management strategies. Cost-prevention strategies may include ignoring illnesses or delaying seeking health care services, while cost-management strategies could include using health care services and managing time and financial costs by adjusting labor supply, mobilizing resources, or adjusting household spending. An increase in household debt indicates that households are mobilizing resources to cover not only medical costs but also nonmedical expenses and productivity losses associated with ill health. Private health insurance offers additional coverage that could shield subscribers from the costs of and unanticipated health shock, making private health insurance subscribers less vulnerable to the financial risk of illness. This would allow subscribers to accrue less household debt compared to nonsubscribers in the case of health shock. However, our results show that the additional coverage provided by private health insurance may have failed to mitigate the financial implications of a health shock.

Private health insurance coverage was associated with an increase in outpatient visits and utilization of inpatient care at the onset of a health shock. This finding is consistent with previous research showing an increase in inpatient and outpatient use among those with private health insurance.^39,40 These findings are similar in countries that have both a public health insurance program and a supplementary private health insurance system and may indicate that the additional coverage provided through supplementary insurance likely affects individual decision-making and health service utilization.^41,42

To the best of our knowledge, this is the first paper to analyze the effect of supplementary private health insurance enrollment for those experiencing a severe health shock. Health shocks can cause long-term financial strain on households and increase their vulnerability to additional risks, such as bankruptcy and impoverishment with delayed repayment or delinquency of debt. ⁴³ The role of supplementary private health insurance is not only to fulfill the diverse medical demands that public schemes fail to cover but also to supplement the insufficient level of coverage provided through the National Health Insurance scheme in South Korea. However, results from this study show that its ability to mitigate financial risks for subscribers, particularly in catastrophic cases like health shocks, is limited.

This study has several limitations. First, our data were only available annually, which may not accurately reflect the timing of health shocks and their impact on households. As the onset of health shocks may differ by up to one year, the period before the health shock and after the health shock can differ up to a year. Second, the data do not provide information on disease severity, which could lead to differences in the effects of health shocks on household debt. Furthermore, the health shock events examined in this study are typically preceded by deteriorating health conditions that may influence patient behavior, health care utilization, and financial stress prior to diagnosis. To address the issue of endogeneity caused by changes in patient behavior prior to a health shock, we centered the onset of each health shock at t = 0, ensuring that the treatment and control groups had a similar expectation of the likelihood of the health shock occurring in the future. We also used a difference-in-differences estimation strategy with entropy balancing to control for possible omitted confounders and found no evidence of diverging pre-period trends in the outcomes we studied. Finally, near the end of our study period, South Korea expanded the National Health Insurance program to include a catastrophic health expenditure support program, which could have attenuated estimates of the financial protection offered by private health insurance. However, we conducted various robustness checks to evaluate the impact of this policy change and did not find any evidence of policy effects on our study sample.

Conclusion

Supplementary private health insurance had no significant effect on changes in household debt measures among a population of survey respondents in South Korea experiencing a health shock. Our findings imply that supplementary private health insurance may have a limited role in addressing financial risk during such events. It is important for future research to consider defining health shocks in different ways and to estimate the direct reimbursement from private health insurance to obtain a more accurate assessment of its effectiveness.

Author Biographies

Sooyeol Park is a doctoral student at the University of North Carolina Chapel Hill in health policy and management with a minor in economics. He received his Master of Health Science at Johns Hopkins University in health economics and outcome research. His PhD research focuses on the effect of telehealth integration on care accessibility and quality, its potential to substitute in-person care, and the cost-effectiveness of using telehealth approaches in managing chronic conditions.

Kanghee Kim is a deputy director at the Korea International Cooperation Agency's El Salvador office. She received her Master of Public Health in health policy and management from Seoul National University. Her primary research interest lies in implementation research, emphasizing strategies to mitigate social disparities in both health and development domains.

Kevin Callison is an associate professor in the Department of Health Policy and Management at the Tulane University School of Public Health and Tropical Medicine. He also has appointments in the Department of Economics and the Murphy Institute for Political Economy at Tulane. His research primarily focuses on issues related to the fields of health economics, labor economics, and applied econometrics, and is particularly interested in evaluating policy interventions that aim to improve population health, including issues related to health insurance and health service use. In addition, He conducts research on health behaviors, health determinants, and substance use.