Effect of national antimicrobial resistance policies on antibiotic prescribing rates in South Korea: an interrupted time series analysis

Abstract

Background

Despite the antimicrobial resistance (AMR) policies implemented in Korea, outpatient antibiotic consumption remains high. Evidence on the effects of these policies is limited. We assessed the effects of Korea’s AMR control policies on antibiotic prescription rates by examining short- and long-term impacts across healthcare institution types and policy intervention periods.

Methods

We conducted an interrupted time series analysis using an ARIMAX (1,0,0) model with nationwide quarterly data from tertiary hospitals, general hospitals, hospitals, and clinics from the first quarter of 2000 to the fourth quarter of 2023. Four policy interventions were modeled as level and slope changes to estimate short-term effects immediately after each intervention and long-term effects eight quarters later.

Results

The 2001 policy was associated with short-term reductions in antibiotic prescription rates across all institution types, decreasing by 2.21% points (pp) in tertiary hospitals, 3.65 pp in general hospitals, 1.60 pp in hospitals, and 3.21 pp in clinics. Long term, reductions were observed only in general hospitals (− 2.95 pp), while increases occurred in other settings. Following the 2006 policy, short-term prescription rates increased across all institutions but declined in the long term in tertiary (− 2.94 pp) and general hospitals (− 2.61 pp), with increases persisting in hospitals and clinics. After the 2013 policy, short-and long-term prescription rates increased across all institution types. The 2017 policy was associated with short-term increases across all settings, long-term decline was observed only in clinics (− 0.94 pp).

Conclusion

The effects of Korea’s AMR policies varied according to policy type, implementation period, and healthcare institution type, with substantial differences between short-term and long-term outcomes. Although several policies were associated with reductions in antibiotic prescribing, these effects were often not sustained over time. These findings suggest that regulation, public reporting, and financial incentives alone are insufficient to achieve lasting changes in prescribing behavior. Sustainable AMR control will require institution-specific and multifaceted strategies that integrate antimicrobial stewardship, continuous monitoring and feedback, patient engagement, and One Health approaches.

Background

Antimicrobial resistance (AMR) is a growing public health burden worldwide [1–5]. The World Health Organization (WHO) has identified AMR as a major threat to humanity, prompting governments across the globe to implement strategies to address the problem [6, 7]. In Korea, the severity of AMR increased in the 1990s. High resistance rates for major bacterial pathogens were reported, including methicillin-resistant Staphylococcus aureus (MRSA), penicillin-non-susceptible Streptococcus pneumoniae, and ampicillin-resistant Haemophilus influenzae, with resistance exceeding 50% in some strains [8, 9]. The results of the 1998 Korea Nationwide Surveillance of the Antimicrobial Resistance program confirmed the widespread dissemination of resistant organisms and underscored the need for national antimicrobial stewardship [10]. Simultaneously, antibiotic prescription rates for outpatients—especially those with upper respiratory tract infections—reached 50–80%, highlighting excessive use and variability across healthcare institutions [8]. In response, the Korean government launched the Evaluation Project on Appropriate Prescribing (EPAP) in 2001 and subsequently introduced a series of additional policies to reduce the rate of antibiotic prescribing. The EPAP aimed to compare prescribing trends for key pharmaceuticals, such as antibiotics and injectable drugs across healthcare institutions and provide feedback to prescribers, reducing inappropriate use and promoting rational prescribing [10–14]. Despite these efforts, reports from the Organization for Economic Co-operation and Development (OECD) indicate that Korea remains one of the countries with the highest antibiotic consumption. Numerous studies have documented excessive and variable antibiotic use across medical institutions [11, 15–18]. According to reports from the Health Insurance Review and Assessment Service (HIRA), antibiotic prescription rates have gradually declined following policy implementation, and some studies have evaluated the positive effects of these antimicrobial resistance policies [19]. However, most previous studies have been limited by short timeframes, specific populations, or narrow regional or institutional scopes. Empirical analyses using the full 23-year dataset since the policy’s inception, to assess short- and long-term effects by intervention period, remain scarce [20–25]. Moreover, few studies have examined how antimicrobial resistance control policies have affected antibiotic prescribing across different types of healthcare institutions. Therefore, we aimed to evaluate the impact of Korea’s antimicrobial resistance control policies on antibiotic prescription rates by analyzing the short- and long-term effects across intervention periods and institution types.

Methods

Study design

To evaluate the causal effects of policy interventions, a quarterly interrupted time series (ITS) analysis was employed. The ITS design is widely used to assess and infer the effects of large-scale public health interventions, such as healthcare policy changes [26–29]. The analytical model was the auto regressive integrated moving average with exogenous inputs (ARIMAX) (1,0,0) within the Seasonal ARIMAX (SARIMAX) framework, a state-space estimation procedure. The seasonal component (seasonal order = [0,0,0,0]) was excluded, rendering the model equivalent to a standard ARIMAX specification.

Data source

We used data from the National Health Insurance Statistical Yearbooks published annually by the National Health Insurance Service (NHIS) and HIRA [30]. Korea introduced a universal National Health Insurance (NHI) system in 1989, under which the NHIS, as the single public insurance institution, provides health coverage to the entire population. NHIS and HIRA maintain nationwide administrative healthcare databases and make data available for research purposes. The unit of analysis was the national quarterly antibiotic prescription rate. The observation period spanned 24 years, from the first quarter of 2000 to the fourth quarter of 2023 (96 quarters in total). The primary data were available on a quarterly basis; however, antibiotic prescription rates have been reported annually since 2018. To generate quarterly estimates for this recent period, seasonal interpolation was applied. The number of physicians, which showed a gradual annual increase with minimal fluctuation, was converted into quarterly data using linear interpolation for periods with missing observations.

Variables

The outcome variables were antibiotic prescription rates by type of healthcare institution—tertiary hospitals, general hospitals, hospitals, and clinics.

Four major national policy interventions targeting antibiotic prescribing were examined. First, the EPAP was introduced in the third quarter (Q3) of 2001. This program was implemented to systematically assess the appropriateness of antibiotic prescribing at the national level, monitoring antibiotic use across healthcare institutions and establishing standardized evaluation criteria. Second, in the second quarter (Q2) of 2006, EPAP evaluation results were publicly disclosed, representing a policy shift from internal monitoring to external transparency. This measure enabled comparisons of performance across healthcare institutions and introduced reputational accountability as a mechanism to influence prescribing behavior. Third, in the third quarter (Q3) of 2013, an incentive–penalty scheme for clinics was introduced. Antibiotic prescribing performance was linked to financial rewards or sanctions, allowing evaluation results to be directly reflected in institutional reimbursement. Finally, in the fourth quarter (Q4) of 2017, the evaluation framework transitioned from a relative to an absolute assessment system. This change replaced peer-based comparisons with fixed performance standards.

Control variables included the number of outpatient visits for acute upper respiratory infections, total number of physicians, quarterly seasonal dummy variables, influenza epidemic intensity (measured by the number of outpatient visits with International Classification of Diseases, Tenth edition codes J10–J11), and a coronavirus disease (COVID-19) dummy variable. The COVID-19 dummy was defined based on surveillance data from the Korea Disease Control and Prevention Agency: it was assigned a value of 1 if more than half of a given quarter overlapped with a nationwide epidemic period or if the national peak month occurred within that quarter and 0 otherwise [31].

The total number of physicians was used as a proxy for healthcare supply rather than the number of physicians per facility (PPF). In Korea, the number of physicians changes gradually because medical school admissions and the supply of specialists are regulated by the government and the medical community. Contrarily, the number of healthcare institutions fluctuates in the short term due to frequent openings and closures [32]. Therefore, using PPF may exaggerate denominator effects arising from institutional volatility rather than reflecting actual changes in service capacity. Furthermore, as antibiotic prescribing decisions are made at the level of physicians rather than institutions, the total number of physicians provides a more direct and causally appropriate measure of prescribing behavior.

Sensitivity analyses were conducted by substituting PPF for the physician count. While the overall estimates of policy effects were broadly similar, the long-term effect of the 2017 policy among clinics differed in direction. For these reasons, the number of physicians was considered a more appropriate supply-side.

Statistical analysis

Estimation model

The ARIMAX (1,0,0) model was used to estimate level and slope changes associated with each policy intervention. The short-term effect was defined as the immediate level change following the intervention, while the long-term effect was defined as the slope change observed eight quarters after the intervention.

The choice of an eight-quarter lag was based on prior methodological recommendations acknowledging that the effects of policy interventions are not instantaneous but may stabilize gradually over time due to lagged responses. Several simulation-based ITS studies have incorporated scenarios assuming a 2-year delayed effect to reflect this progressive stabilization [25–28].

The long-term effect for each intervention k was calculated using the formula β₂_k + 8 × β₃_k.

Model specification

antibiotic prescription rate at time t. quarterly time variable within the study period. dummy variables for four policy interventions. Immediate level change atfer the policy intervention. number of quarters since each intervention. slope change in trend after the policy intervention. set of control variables. .

All statistical analyses were conducted using Python (version 3.10; Python Software Foundation, Wilmington, DE, USA). The time series analysis employed the SARIMAX model implemented in the statsmodels package, following an ITS framework based on an ARIMAX (1,0,0) specification.

Results

The results of the interrupted time-series analyses are presented in Table 1 and Figs. 1 and 2.

Table 1.

Immediate level and slope changes in antibiotic prescription rates by facility type following policy interventions

Healthcare facilities type	Policy intervention	Immediate level change β (95% CI)	Slope change β (8Q, 95% CI)
Tertiary hospitals	3Q 2001	− 2.21*** (− 2.24 to − 2.18)	+ 4.78*** (4.61–4.95)
	2Q 2006	+ 0.08*** (0.05–0.11)	− 2.94*** (− 3.23 to − 2.65)
	3Q 2013	+ 0.80*** (0.78–0.82)	+ 2.52*** (2.36–2.68)
	4Q 2017	+ 0.60*** (0.58–0.62)	+ 0.10 (− 0.21 to 0.41)
General hospitals	3Q 2001	− 3.65*** (− 3.72 to − 3.58)	− 2.95*** (− 3.11 to − 2.79)
	2Q 2006	+ 0.51*** (0.48–0.54)	− 2.61*** (− 2.83 to − 2.39)
	3Q 2013	+ 0.55*** (0.53–0.57)	+ 1.29*** (1.08–1.50)
	4Q 2017	+ 0.88*** (0.86–0.90)	+ 0.23 (− 0.07 to 0.53)
Hospitals	3Q 2001	− 1.60*** (− 1.64 to − 1.56)	+ 1.35*** (1.15–1.55)
	2Q 2006	+ 1.72*** (1.69–1.75)	+ 1.51*** (1.31–1.71)
	3Q 2013	+ 0.95*** (0.92–0.98)	+ 1.08*** (0.89–1.27)
	4Q 2017	+ 1.01*** (0.98–1.04)	+ 0.42*** (0.16–0.68)
Clinics	3Q 2001	− 3.21*** (− 3.23 to − 3.19)	+ 10.51*** (9.89–11.14)
	2Q 2006	+ 5.28*** (5.26–5.30)	+ 11.24*** (10.61–11.87)
	3Q 2013	+ 1.90*** (1.88–1.92)	+ 0.79*** (0.07–1.51)
	4Q 2017	+ 1.53*** (1.51–1.55)	− 0.94*** (− 1.48 to − 0.40)

Significance levels: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***). CI, confidence interval

Fig. 1.

Immediate level changes in antibiotic prescription rates by facility type following national antimicrobial resistance policy interventions in South Korea, Q1 2000–Q4 2023. Estimates were obtained using a quarterly interrupted time series analysis with an ARIMAX (1,0,0) model. Facility types include tertiary hospitals, general hospitals, hospitals, and clinics. Immediate effects represent level changes observed in the quarter following each policy intervention

Fig. 2.

Changes in slopes of antibiotic prescription rates by facility type following national antimicrobial resistance policy interventions in South Korea, Q1 2000–Q4 2023. Estimates were obtained using a quarterly interrupted time series analysis with an ARIMAX (1,0,0) model. Facility types include tertiary hospitals, general hospitals, hospitals, and clinics. Slope changes represent differences in the trend of antibiotic prescription rates observed over the eight quarters following each policy intervention

Antibiotic prescription rate in tertiary hospitals

Immediately after the 2001 policy intervention, the antibiotic prescription rate in tertiary hospitals decreased by 2.21 pp (95% confidence interval [CI] − 2.24 to − 2.18; p < 0.001); however, it increased by 4.78 pp after eight quarters (95% CI 4.61–4.95; p < 0.001). Following the 2006 intervention, there was an immediate increase of 0.08 pp (95% CI 0.05–0.11; p < 0.001) in antibiotic prescription rate, followed by a decrease of 2.94 pp (95% CI − 3.23 to − 2.65; p < 0.001) eight quarters after the intervention.

After the 2013 policy intervention, the antibiotic prescription rate increased by 0.80 pp (95% CI 0.78–0.82; p < 0.001) immediately and continued to rise by 2.52 pp after eight quarters (95% CI 2.36–2.68; p < 0.001). In 2017, the rate increased by 0.60 pp immediately after the intervention (95% CI 0.58–0.62; p < 0.001) and by 0.10 pp after eight quarters (95% CI − 0.21 to 0.41; p > 0.05); however, the latter change was not significant.

Antibiotic prescription rate in general hospitals

Immediately after the 2001 policy intervention, the antibiotic prescription rate in general hospitals decreased by 3.65% points (95% CI − 3.72 to − 3.58; p < 0.001), and after eight quarters, it further decreased by 2.95% points (95% CI − 3.11 to − 2.79; p < 0.001). Following the 2006 intervention, there was an immediate increase of 0.51 pp (95% CI 0.48–0.54; p < 0.001) in antibiotic prescription rate, followed by a decrease of 2.61 pp (95% CI − 2.83 to − 2.39; p < 0.001) eight quarters after the intervention.

Contrarily, after the 2013 intervention, the antibiotic prescription rate increased by 0.55 pp (95% CI 0.53–0.57; p < 0.001) immediately and by 1.29 pp after eight quarters (95% CI 1.08–1.50; p < 0.001). The 2017 intervention produced a similar pattern, with an immediate increase of 0.88 percentage points (95% CI 0.86–0.90; p < 0.001). After eight quarters, the antibiotic prescription rate increased by 0.23 pp (95% CI − 0.07 to 0.53; p > 0.05), but this change was not statistically significant.

Antibiotic prescription rate in hospitals

Immediately after the 2001 policy intervention, the antibiotic prescription rate in hospitals decreased by 1.60 percentage points (95% CI − 1.64 to − 1.56) but increased by 1.35 pp after eight quarters (95% CI 1.15–1.55). Following the 2006 intervention, the rate increased by 1.72 pp immediately (95% CI 1.69–1.75) and by 1.51 pp after eight quarters (95% CI 1.31–1.71).

The 2013 and 2017 interventions also resulted in increases in antibiotic prescription rates: 0.95 pp (95% CI 0.92–0.98) and 1.01 percentage points (95% CI 0.98–1.04), respectively, immediately after the interventions, with further increases of 1.08 (95% CI 0.89–1.27) and 0.42 pp (95% CI 0.16–0.68), respectively, after eight quarters.

Antibiotic prescription rate in clinics

The policy effects in clinics differed from those observed in other types of healthcare institutions. Immediately after the 2001 intervention, the antibiotic prescription rate in clinics decreased by 3.21 pp (95% CI − 3.23 to − 3.19); however, it subsequently increased markedly by 10.51 pp after eight quarters (95% CI 9.89–11.14). Following the 2006 intervention, the rate increased by 5.28 pp immediately (95% CI 5.26–5.30) and by 11.24 pp after eight quarters (95% CI 10.61–11.87).

After the 2013 policy intervention, the antibiotic prescription rate increased by 1.90 pp immediately (95% CI 1.88–1.92) and by 0.79 pp after eight quarters (95% CI: 0.07–1.51). Contrarily, the 2017 intervention led to an immediate increase of 1.53 percentage points (95% CI 1.51–1.55) but a subsequent decrease of 0.94 pp after eight quarters (95% CI − 1.48 to − 0.40), showing an opposite long-term trend.

Discussion

In this study, we analyzed the impact of Korea’s AMR control policies on antibiotic prescription rates by distinguishing short- and long-term effects according to policy intervention periods and types of healthcare institutions. The results indicate that the effects of AMR policies were not uniform but exhibited substantial heterogeneity depending on the timing of policy implementation, policy duration, and the institutional characteristics of healthcare settings.

The 2001 policy intervention was associated with immediate reductions in antibiotic prescription rates across all types of healthcare institutions, demonstrating a strong short-term behavioral response following the introduction of the EPAP. However, long-term, sustained reductions in antibiotic prescription rates were observed only in general hospitals, whereas rebound increases were identified in tertiary hospitals, hospitals, and clinics. The discrepancy between the short and long-term effects may be related to the policy adaptation process among healthcare providers. During the initial implementation period, healthcare institutions may have reduced antibiotic prescribing in response to regulatory scrutiny, evaluation requirements, and administrative pressure. Nam and Kim (2003) reported that the antibiotic prescribing rate in primary care clinics declined from 59.2% in 1999 to 48.9% in 2002 following the introduction of EPAP, representing a reduction of more than 10 percentage points. They attributed this decline to the implementation of EPAP [33].

In addition, the pharmaceutical separation reform implemented in July 2000, immediately prior to the introduction of EPAP, may also have contributed to the reduction in antibiotic prescribing. By separating physicians’ prescribing authority from pharmacists' dispensing authority, non-prescription channels for antibiotic distribution were eliminated. This change may have structurally restricted access to antibiotics and strengthened control of antibiotic use, thereby reducing antibiotic consumption and promoting appropriate use. Therefore, the short-term effects observed in 2001 may reflect the direct impact of EPAP, influence of strengthened controls on antibiotic access and more appropriate prescribing environment established. Kim et al. (2016) suggested that the separation of prescribing and dispensing functions under the pharmaceutical reform policy may have contributed to reductions in antibiotic prescribing by restricting access to antibiotics without a physician’s prescription and strengthening antibiotic control [8].

However, over time, healthcare institutions and physicians may have adapted to policy requirements or adjusted their prescribing behavior in ways that circumvented the intended effects of the policy. Primary care clinics may have faced difficulties sustaining reductions in antibiotic prescribing because of concerns regarding patient retention and competition within the healthcare market. Patient expectations and demands may have substantially influenced prescribing decisions in primary care settings. Expert consultations conducted for this study indicated that physicians working in clinics perceived withholding antibiotics as potentially leading to reduced patient satisfaction and the loss of patients to competing providers.

Experts working in tertiary hospitals indicated that government policies promoting appropriate antibiotic prescribing were often perceived as regulations that constrained physicians’ clinical judgment and professional autonomy. Although prescribing behavior changed immediately following policy implementation, healthcare providers gradually adapted to the evaluation system over time. In some cases, providers may have focused on responding to or avoiding evaluations rather than fundamentally changing their prescribing practices. Such policy adaptations may influence the long-term attenuation of the initial policy effects and subsequent rebound.

The 2006 policy intervention exhibited a different pattern. Short-term, antibiotic prescription rates increased across all types of healthcare institutions, while long-term effects varied by institution type. Prescription rates declined in tertiary and general hospitals but continued to increase in hospitals and clinics. One possible explanation for the long-term effect is the temporal gap between the implementation of the public reporting policy and the subsequent collection and public disclosure of evaluation results. Antibiotic prescription data following the 2006 policy were released approximately one year after implementation, and thus the changes observed eight quarters after the intervention are more appropriately understood as behavioral responses following the disclosure of evaluation results.

Similarly, Kim et al. reported a 6.1 pp reduction in antibiotic prescription rates following the 2001 policy and an additional 3.9 pp decrease after the public disclosure of evaluation results in 2006 [34]. Jung et al. likewise confirmed a 5.79% reduction in antibiotic prescription rates following the 2006 policy intervention, while Kim et al. reported a 12.3 pp decrease after public disclosure of evaluation outcomes [20, 35].

Further, the public disclosure of evaluation results may have influenced reputational competition differently among institutions. According to expert consultations conducted for this study, tertiary and general hospitals were relatively more sensitive to the reputational consequences and public image implications associated with the disclosure of evaluation results, whereas physicians in primary care clinics perceived that such information had limited influence on patients’ choices of healthcare providers. This difference may partially explain why the same policy intervention produced heterogeneous effects across different types of healthcare institutions.

Tertiary and general hospitals may have been more responsive to publicly reported evaluation results because maintaining public accountability and institutional reputation is important for them. Contrarily, primary care clinics tend to rely more on local patient populations, and information such as antibiotic prescribing rates may not translate into actual healthcare use decisions among patients. Thus, the disclosed information may not have generated sufficient reputational pressure to influence prescribing.

Further, tertiary hospitals and general hospitals possess dedicated infection control personnel and data management systems, enabling them to continuously monitor evaluation outcomes and implement quality improvement activities at the organizational level. Contrarily, primary care clinics have limited personnel and resources, making it difficult to sustain prescribing improvement initiatives. Consequently, while policy interventions may have produced immediate behavioral changes across healthcare settings, differences in organizational capacity may have contributed to variations in the sustainability of policy effects over time.

Furthermore, compared with primary care clinics, tertiary and general hospitals are continuously exposed to multiple quality assurance and performance evaluation systems, including hospital accreditation programs, infection control assessments, and healthcare quality evaluations. Such institutional environments may strengthen organizational commitment to appropriate antibiotic use, enhance infection control capacity, and facilitate the implementation and maintenance of antimicrobial stewardship programs (ASPs). Therefore, the relatively sustained effects observed in larger hospitals following certain policy interventions may reflect not only the direct impact of EPAP but also the influence of these broader institutional and organizational infrastructures.

The 2001 and 2006 policies functioned as regulatory interventions characterized by legal authority and enforceability, aligning with policy theories that predict immediate effects when new regulations or institutional mechanisms are introduced. According to policy change theories, regulatory or institutional innovations can trigger abrupt and punctuated changes at specific points in time. Similarly, the policy shock theory suggests that major events or shocks can capture the attention of policymakers and the public, leading to immediate behavioral change following new regulatory actions. Furthermore, policy implementation and compliance theories propose that policies with coercive or legally legitimate authority are likely to induce rapid compliance behaviors once implemented [36–41]. EPAP was the first policy directly aimed at addressing the issue of antimicrobial resistance in Korea. Since key stakeholders were already aware of the severity of the AMR problem and the government’s determination to address it, the policy’s introduction produced an immediate impact. Particularly, the EPAP was introduced in the wake of the 2000 separation of prescription and dispensation functions—a major healthcare reform that brought antibiotic misuse into sharp public focus. Given this context, the immediate effects observed following the 2001 policy are understandable [42]. For the 2006 policy, the delayed emergence of its effects can be explained by the timeline of public disclosure. Although the policy was implemented in 2006, the antibiotic prescription data were compiled and released roughly one year later. Thus, the significant effects observed after eight quarters can still be interpreted as immediate post-implementation effects, like those following the 2001 policy.

The outcomes of the 2013 and 2017 policy interventions differed from those of earlier regulatory policies. Following the 2013 intervention, short- and long-term antibiotic prescription rates increased across all types of healthcare institutions, indicating that the policy effects were not aligned with the intended goal of reducing antibiotic use. This divergence can be largely attributed to the scope of the 2013 policy, which targeted only clinics. In 2013, the government modified the EPAP to address concerns that the program was relatively disadvantageous to clinics. The revised policy introduced a financial incentive system that rewarded clinics to achieve better performance. However, the lack of observable effects, even among clinics, may be explained by two factors: the small magnitude of financial incentives, which failed to provide meaningful economic motivation, and the limited policy influence compared with that of the 2006 public disclosure initiative. Kim (2018) similarly reported that, although the 2013 policy led to a slight reduction in antibiotic prescription rates, the effect was minimal and did not induce behavioral changes in prescribing practices among healthcare providers [43]. The author attributed this to the weak financial incentives and the comparatively limited policy impact. Likewise, Kim et al. (2018) found no significant policy effect of the 2013 intervention, citing the insufficient incentive size to drive behavioral change among prescribers [44].

In primary care settings, concerns regarding patient dissatisfaction, patient attrition, and potential medico-legal disputes may have exerted a greater influence on prescribing behavior than financial incentives. In Korea, patient expectations and satisfaction play an important role in outpatient prescribing decisions, particularly among patients with mild respiratory tract infections, where requests for antibiotics remain common. This phenomenon may be related to patients’ expectations for a “visible therapeutic intervention”.

According to expert consultations, patients frequently request antibiotics even for viral illnesses such as the common cold. Physicians reported that refusing such requests may lead patients to perceive that the physician is providing inadequate treatment or prescribing “weaker” medications, potentially influencing patients’ future healthcare-seeking behavior and provider choice. Given the highly competitive nature of the primary care market, physicians in clinic settings may be particularly sensitive to these patient responses.

Furthermore, experts noted that primary care clinics often face challenges in utilizing point-of-care testing (POCT) to rapidly distinguish between viral and bacterial infections. Under conditions of diagnostic uncertainty, some physicians expressed concerns that failure to prescribe antibiotics could expose them to potential medico-legal disputes if a patient’s condition subsequently deteriorates or an unexpected adverse clinical outcome occurs. Experts further emphasized that, within the Korean healthcare environment, physicians may face substantial legal and financial burdens when medical disputes arise. Compared with large hospitals, primary care clinics generally have fewer organizational and financial resources available to manage such risks.

The potential consequences of not prescribing antibiotics may be perceived as more immediate and tangible than the long-term and less visible risks associated with antimicrobial resistance. This perception may encourage more defensive prescribing behavior, even in situations where the clinical need for antibiotics is uncertain.

Taken together, patient expectations and demands, competition among healthcare providers, diagnostic uncertainty, and concerns regarding medico-legal liability may collectively limit the effectiveness of relatively modest financial incentive programs in changing established prescribing behaviors in primary care settings.

The 2017 policy intervention demonstrated heterogeneous effects across healthcare institution types. While short-term increases in antibiotic prescription rates were observed across all types of institutions, long-term reductions were identified only among clinics.

Short-term increases in antibiotic prescribing rates across all types of healthcare institutions were observed, suggesting that the transition to an absolute evaluation system did not immediately lead to improvements in prescribing behavior. This may indicate that the change in evaluation methodology was not perceived as a sufficiently strong policy signal by healthcare providers or that institutions experienced a period of adjustment as they adapted from the previous relative evaluation system to the new evaluation framework.

However, in the long term, a reduction in antibiotic prescribing rates was observed only among primary care clinics. According to expert consultations conducted for this study, the absolute evaluation system may have provided primary care physicians with a clearer target threshold for antibiotic prescribing than the previous relative evaluation system. Under the relative evaluation framework, performance was assessed based on rankings within peer groups, whereas under the absolute evaluation system, maintaining prescribing rates below a predefined threshold became the primary objective. This change may have facilitated gradual adjustments in prescribing behavior among primary care providers over time.

In addition, primary care clinics generally operate within a relatively simple outpatient-oriented practice structure, which may have enabled them to adapt more flexibly to changes in evaluation criteria. Contrarily, tertiary and general hospitals treat a higher proportion of severely ill patients, immunocompromised individuals, and patients with complex infections, resulting in greater clinical necessity for antibiotic use. Expert consultations also suggested that physicians working in larger hospitals often perceived that the absolute evaluation criteria did not adequately reflect differences in patient severity and clinical complexity. Consequently, the absolute evaluation system may have had limited influence on prescribing decisions in these settings.

Furthermore, tertiary and general hospitals had already established infection control systems and ASPs before the implementation of the absolute evaluation system. Thus, the change in evaluation methodology itself may not have constituted a sufficiently strong incentive to induce additional behavioral changes. Contrarily, for primary care clinics, the introduction of explicit prescribing targets may have functioned as a more tangible management tool, contributing to the gradual decline in antibiotic prescribing observed over the long term.

Taken together, the 2017 policy intervention did not appear to produce immediate improvements in antibiotic prescribing across healthcare institutions. Nevertheless, the long-term reduction observed exclusively among primary care clinics suggests that the absolute evaluation system may have operated differently across institution types. Particularly, the provision of clear performance thresholds may have been more effective in influencing prescribing behavior in primary care settings than in larger hospitals.

The 2017 transition to an absolute evaluation system has been criticized by scholars for imposing uniform evaluation criteria that may not adequately reflect differences in patient severity, clinical complexity, and real-world practice environments across healthcare institutions [45, 46].

The reasons underlying the heterogeneous policy effects across different policy intervention periods and healthcare institution types can be summarized as follows.

First, differences in responsiveness to regulatory and public reporting policies across healthcare institution types may explain the observed heterogeneity in policy effects. The introduction of the EPAP in 2001 was associated with a short-term reduction in antibiotic prescribing rates across all healthcare institution types; however, this effect was sustained in the long term only among general hospitals. The differential effects observed across healthcare institution types are consistent with prior evidence indicating that responses to regulatory policies may vary according to hospital size and organizational characteristics. For example, Qiu et al. found that the impact of the Hospital Readmissions Reduction Program (HRRP) was heterogeneous across hospitals, with larger hospitals demonstrating greater reductions in readmission rates than smaller hospitals. This suggests that the effectiveness and long-term sustainability of policy interventions may depend, at least in part, on the organizational capacity of healthcare providers [47]. Similarly, the public disclosure policy introduced in 2006 was associated with an initial increase in antibiotic prescribing rates across all institution types, whereas long-term reductions were observed only among tertiary and general hospitals. These findings suggest that regulatory measures and public reporting mechanisms did not operate uniformly across healthcare institutions but exerted differential effects depending on institutional size, organizational capacity, and sensitivity to reputational pressures.

Findings from the expert consultations revealed that tertiary and general hospitals generally perceived the public disclosure of evaluation results as being directly linked to institutional reputation, public image, and organizational performance management. Contrarily, physicians in primary care clinics often believed that publicly reported evaluation information had only a limited influence on patients’ healthcare provider choices. Consequently, tertiary hospitals and general hospitals may have responded more effectively to regulatory policies because of their greater adherence to administrative guidelines and stronger incentives to maintain institutional reputation in response to public reporting. Primary care clinics may have been less sensitive to administrative regulations and reputational pressures, resulting in less sustained changes in prescribing behavior.

These findings are consistent with those reported by Yun et al. (2015), who found that antibiotic prescribing rates decreased across all healthcare institution types following the implementation of public reporting; however, the magnitude of reduction was greatest among tertiary hospitals [21]. The authors suggested that the public perceived lower antibiotic prescribing rates as an indicator of higher-quality healthcare services. Similar patterns have also been reported in high-income countries such as the United Kingdom and Australia, where hospital performance evaluation and public reporting policies were effective in improving prescribing behaviors in large hospitals but demonstrated more limited effects in primary care clinics [48, 49].

Second, differences in patient demand and practice patterns may explain the observed variation in policy effects. The finding that short-term antibiotic prescribing rates in primary care clinics declined following the 2001 policy intervention but subsequently exhibited a substantial rebound in the long term suggests that regulatory policies alone may be insufficient to sustain long-term changes in prescribing behavior in outpatient settings. Primary care clinics treat a large proportion of patients with mild infections, and patient expectations for antibiotics may directly influence prescribing decisions. Nam and Kim (2003) also suggested that patients’ expectations and requests for antibiotics were important factors influencing physicians’ prescribing behavior [33]. Lucas et al. reported that although clinicians commonly prescribe antibiotics to children when a clear clinical indication exists, prescriptions are also frequently issued as a precautionary measure ("just in case") or in response to parental pressure [50]. Expert consultations conducted for this study consistently indicated that refusing patients’ requests for antibiotics could lead to reduced patient satisfaction, patient attrition, and movement to competing nearby healthcare providers. Therefore, the limited long-term sustainability of policy effects in primary care clinics may reflect not a lack of physician compliance with policy objectives, but the persistent influence of real-world clinical pressures, including patient satisfaction, the need to retain patients, and the competitive nature of outpatient practice environments. Contrarily, in tertiary hospitals, patient demands are less likely to directly influence prescribing decisions. Differences in patient characteristics and clinical practice environments across healthcare institution types may therefore contribute to variations in policy effectiveness [51–53]. Taken together, these findings suggest that regulatory policies alone may be insufficient to achieve sustained reductions in antibiotic prescribing rates in primary care settings.

Third, variation in the sustainability of policy effects may be explained by differences in monitoring and feedback systems. In this study, several policy interventions were associated with short-term reductions in antibiotic prescribing rates; however, these effects often weakened or rebounded over time. This finding suggests that the long-term sustainability of policy effects may depend less on the introduction of a policy itself and more on the existence of systems capable of continuously monitoring prescribing behaviors and providing feedback to healthcare institutions after implementation. In large hospitals, ASPs have been voluntarily implemented, primarily under the leadership of infectious disease specialists, since the mid-2000s. Between 2011 and 2013, hospital-level antibiotic prescription review and preauthorization systems became widespread among major tertiary hospitals. Furthermore, following the establishment of the National Action Plan on Antimicrobial Resistance in 2016, ASPs became mandatory in tertiary hospitals beginning in 2017 [54]. In addition, tertiary and general hospitals have been continuously exposed to various quality management and performance evaluation systems, including hospital accreditation, infection control assessments, and healthcare quality evaluations. These institutional frameworks may have enabled evaluation results to be translated into organization-wide quality improvement activities. In contrast, primary care clinics often face limitations in personnel and resources, making it relatively difficult to implement ASPs and establish continuous monitoring and feedback systems. Such differences in organizational infrastructure may have contributed to variations in the sustainability of policy effects across healthcare institution types. Similarly, Langford and Schwartz emphasized that continuous monitoring and feedback are critical components for improving antibiotic prescribing practices in primary care settings [55]. Similarly, Park et al. (2023) identified the absence of continuous monitoring and feedback systems as a key factor contributing to the short-lived and transient effects of policy interventions [56].

Fourth, the structural limitations of financial incentive policies may explain the observed policy effects. Despite the introduction of financial incentives targeting primary care clinics in 2013, the policy did not lead to reductions in antibiotic prescribing rates in either the short term or the long term. Furthermore, following the transition to an absolute evaluation system in 2017, antibiotic prescribing rates increased in the short term across all healthcare institution types, indicating that the policy did not produce an immediate effect on prescribing behavior.

These findings suggest that changes in financial incentives or evaluation methods alone may be insufficient to alter prescribing practices unless they adequately consider not only the magnitude of incentives but also the clinical environment, patient case-mix, and acceptability of the evaluation criteria among healthcare providers.

Contrarily, tertiary and general hospitals often manage a higher proportion of severely ill patients and patients with complex infections, therefore leading to the perception that uniform absolute performance thresholds are insufficient in reflecting the realities of clinical practice. In addition, because many large hospitals had already established ASPs and infection control systems prior to the policy change, the transition to an absolute evaluation system may not have served as a sufficiently strong policy signal to induce additional behavioral changes.

Consequently, the absolute evaluation system may have generated limited long-term behavioral changes among primary care clinics, while failing to produce further reductions in antibiotic prescribing among larger hospitals. These findings are in line with previous evidence indicating that financial incentive–based policies alone may be insufficient to reduce antimicrobial use effectively. Instead, sustained improvements are more likely to be achieved through a multifaceted strategy that integrates regulatory interventions, educational initiatives, continuous monitoring, and feedback mechanisms [57–59].

Fifth, the influence of defensive prescribing associated with increasing medico-legal risks and medical disputes may have contributed to the observed policy effects. This factor may represent an important socio-cultural characteristic of the Korean healthcare environment that has the potential to affect the effectiveness of all policy interventions. According to expert consultations conducted for this study, physicians in primary care clinics perceived that they bear substantial legal and financial responsibility when patient complaints or medical disputes arise. In addition, compared with large hospitals, primary care clinics generally have more limited organizational and financial support systems to mitigate such risks. Furthermore, primary care clinics often have limited access to diagnostic resources that can rapidly distinguish between viral and bacterial infections. These interpretations are supported by previous studies reporting that such phenomena are commonly observed in real-world healthcare settings [60–62]. Thus, in situations characterized by diagnostic uncertainty, physicians may perceive the potential risks associated with not prescribing antibiotics as being particularly significant. Consequently, healthcare providers may place greater emphasis on immediate and tangible risks—such as patient deterioration, medical disputes, and legal liability—than on longer-term and less visible risks, such as antimicrobial resistance. This perception may encourage defensive prescribing even in clinical situations where the necessity of antibiotic use is not clearly established. Therefore, limited financial incentives or regulatory policies alone may be insufficient to overcome these underlying structural and contextual factors that influence prescribing behavior.

The heterogeneous policy effects observed across different policy intervention periods and healthcare institution types suggest that Korean antimicrobial stewardship policies have been partially effective in influencing prescribing behaviors through regulation, evaluation, and public reporting in the short term. However, the sustainability of these effects appears to have been limited by differences in clinical practice environments, policy acceptability, patient demand, organizational capacity, financial incentives, and medico-legal risks across healthcare institution types. The divergence between short-term and long-term effects observed for most policy interventions may not necessarily indicate that the policies themselves lost effectiveness over time. Rather, it may reflect a process of policy adaptation, whereby healthcare institutions and providers gradually adjusted their prescribing behaviors in response to changing policy environments and real-world clinical pressures. While regulations, evaluations, and public reporting initially functioned as strong policy signals, the long-term sustainability of policy effects was likely shaped by practical considerations such as patient demand, organizational capacity, reputational pressures, financial incentives, and concerns regarding medical disputes.

Our findings are broadly consistent with those of previous studies evaluating antibiotic stewardship policies in Korea, but also provide several important extensions. Kim et al. reported that the public disclosure of antibiotic prescribing rates was associated with reductions in antibiotic prescribing for upper respiratory tract infections, suggesting that transparency and reputational incentives can influence prescribing behavior [20]. Similarly, we found that public reporting policies were associated with reductions in antibiotic prescribing in some healthcare settings. However, whereas previous studies primarily focused on average policy effects immediately following policy implementation, our results demonstrate that the effectiveness of public reporting differed substantially across healthcare institution types and was not always sustained over time.

Our findings are also consistent with those of Yun et al., stating that the reduction in antibiotic prescribing following public disclosure was greater in tertiary hospitals than in other healthcare institutions. The authors suggested that hospitals may have responded to public reporting because lower antibiotic prescribing rates were perceived as indicators of higher-quality care [21]. Prang et al. suggested that public reporting of physician and hospital performance information may influence patients’ healthcare provider choices and generate reputational pressures, which may explain why large hospitals tend to respond more strongly to publicly disclosed evaluation results [53]. Likewise, our study found that tertiary hospitals and general hospitals generally exhibited more favorable long-term responses to public reporting and regulatory policies than primary care clinics. However, by examining long-term trends, we further demonstrate that these differences persisted over time and may be explained by variations in organizational capacity, reputational pressures, and stewardship infrastructure.

In addition, our findings regarding the relatively sustained effects observed in larger hospitals are consistent with recent studies emphasizing the importance of ASPs. Previous research has shown that ASP implementation can improve the appropriateness of antibiotic use and strengthen institutional responses to antimicrobial resistance. The expansion of ASPs and infection control systems in Korean tertiary hospitals may therefore have contributed to the more sustained policy effects observed in these settings [22, 24, 25, 49, 63, 64].

Nevertheless, our findings differ from those of many previous studies in one important respect. While earlier studies generally reported favorable policy effects, most focused on short-term outcomes following policy implementation. By examining both short-term and long-term effects over an extended period, our study demonstrates that several policy interventions were associated with attenuation or reversal of their initial effects. These findings suggest that policy adaptation by healthcare providers and institutions may play an important role in determining the long-term effectiveness of antimicrobial stewardship policies. Consequently, future policy evaluations should consider not only whether a policy is effective immediately after implementation but also whether its effects can be sustained over time.

Therefore, reducing antibiotic prescribing and effectively addressing AMR will require a multidimensional policy strategy that considers the mechanisms underlying individual policy interventions and the distinct characteristics of different healthcare institution types. As demonstrated in this study, most policy interventions produced measurable short-term effects but were followed by attenuation or rebound effects in the long term. These findings suggest that regulatory approaches alone may be insufficient to sustain long-term changes in prescribing behavior and that future policies should integrate policy acceptability, patient demand management, continuous monitoring and feedback systems, institution-specific incentive structures, and mechanisms that address medico-legal concerns.

Although the policy recommendations presented in this study are intended to improve antibiotic stewardship across all healthcare settings, particular emphasis is placed on primary care clinics. This focus is justified by the fact that primary care clinics account for a substantial proportion of antibiotic prescribing in Korea and exhibited the most limited long-term responses to several policy interventions examined in this study. Furthermore, expert consultations consistently identified patient demand, competitive outpatient practice environments, diagnostic uncertainty, and concerns regarding medico-legal liability as key drivers of antibiotic prescribing in primary care settings. Therefore, while national antimicrobial stewardship policies should encompass all healthcare institutions, primary care clinics should be considered a priority target for future interventions aimed at achieving sustainable reductions in antibiotic use.

At the national level, several policy actions should be considered. First, a comprehensive and sustainable governance framework for AMR management should be established. Previous policies have largely relied on individual policy instruments such as regulation, evaluation, public reporting, and financial incentives; however, the findings of this study indicate that no single intervention is likely to achieve sustained changes in prescribing behavior. Future policies should therefore evolve toward multifaceted strategies that integrate regulation, education, surveillance, feedback mechanisms, financial incentives, public awareness initiatives, and investments in infection control infrastructure.

Second, antimicrobial stewardship education should be systematically incorporated throughout the continuum of medical education. This recommendation is supported by expert consultations, which suggested that established prescribing habits often persist for many years and that current clinical practice does not always adequately reflect updated antibiotic prescribing guidelines.

Third, strategies aimed at improving public awareness and patient understanding of appropriate antibiotic use are needed. Public education should emphasize that antibiotics are ineffective against viral illnesses such as the common cold and that inappropriate antibiotic use contributes not only to antimicrobial resistance but also to adverse health consequences at the individual level. Communication campaigns may be more effective if they emphasize personally relevant risks, such as disruption of the normal microbiome and increased susceptibility to future infections, rather than focusing solely on abstract public health concerns.

Fourth, broader social and workplace environments should be considered as part of AMR management strategies. Experts noted that Korean workplace culture often encourages rapid recovery and early return to work, which may increase patient expectations for antibiotics. Consequently, social policies that support adequate rest and sick leave during infectious illnesses may indirectly contribute to reducing unnecessary antibiotic use.

Fifth, AMR management should be strengthened through a One Health approach. Effective stewardship requires coordinated management of antibiotic use not only in human healthcare but also in animal husbandry and agriculture, as well as integrated surveillance of antimicrobial resistance in environmental settings.

For large hospitals, policy approaches should focus less on reducing antibiotic prescribing rates per se and more on promoting appropriate antibiotic use and preventing the transmission of antimicrobial-resistant organisms. Since these institutions care for a large proportion of severely ill, immunocompromised, and clinically complex patients, antibiotic use is often clinically necessary. Accordingly, evaluation systems should move beyond prescribing rates alone and incorporate measures of prescribing appropriateness, microbiological testing practices, de-escalation of therapy, and infection control performance indicators.

In addition, ASPs should be more closely linked to healthcare quality assessments and reimbursement systems to support the recruitment and retention of infectious disease specialists, clinical pharmacists, and infection control personnel.

Particular attention should also be given to infection prevention and control. Korea’s healthcare system is characterized by multibed hospital rooms, caregiver-centered care, and frequent patient transfers between institutions, all of which may facilitate the spread of antimicrobial-resistant organisms. Therefore, efforts should be focused on strengthening multidrug-resistant organism surveillance systems, improving information-sharing during patient transfers, and establishing inter-hospital infection control networks.

Furthermore, ASP support programs and antimicrobial resistance surveillance systems should be expanded to smaller and long-term care hospitals, which often have more limited infection control capacity than tertiary hospitals. Expert consultations also identified long-term care hospitals as a potential blind spot in national antimicrobial resistance management efforts.

More practical and clinic-centered interventions are needed for primary care settings. The recommendations presented below were derived directly from issues repeatedly highlighted during expert consultations.

First, broader reimbursement and implementation of POCT, including C-reactive protein (CRP) testing, should be considered. Experts identified diagnostic uncertainty—particularly the inability to completely rule out bacterial infections—as one of the primary drivers of antibiotic prescribing in primary care. POCT may help clinicians distinguish between viral and bacterial infections and support evidence-based prescribing decisions.

Second, delayed prescribing should be formally incorporated into clinical practice guidelines. Delayed prescribing has been shown to reduce antibiotic use while addressing patient concerns and uncertainty. Reimbursement mechanisms should also be developed to compensate physicians for the additional time required to educate patients regarding delayed prescribing and the natural course of illness.

Third, behavioral economics-based peer-comparison feedback systems should be implemented. Although current appropriateness assessments are primarily retrospective, real-time feedback comparing individual physicians’ prescribing rates with regional and specialty-specific benchmarks may help reduce prescribing inertia and promote voluntary behavior change.

Fourth, demand-side interventions targeting patient expectations are needed. Experts consistently reported that many patients continue to believe that antibiotics are necessary for faster recovery. Therefore, stewardship interventions in primary care should target both physician behavior and patient expectations.

Fifth, institutional mechanisms to reduce medico-legal risks should be established. According to expert consultations, concerns regarding potential medical disputes following non-prescribing decisions strongly influence prescribing behavior in primary care settings. Accordingly, physicians who adhere to evidence-based guidelines, use POCT appropriately, and follow recommended delayed-prescribing strategies should be protected from excessive legal liability.

Such a comprehensive and institution-specific approach will be essential for translating short-term policy gains into sustainable long-term reductions in antibiotic use.

This study has several important strengths. First, it employed a quasi-experimental design, specifically an ITS approach—which is a robust analytical method for evaluating policy effects when random assignment of policy exposure is not feasible. By distinguishing between short- and long-term effects, this study provides a more nuanced assessment of policy impact. Second, using 23 years of nationwide data, the study analyzed policy effects across different types of healthcare institutions according to the timing of antimicrobial resistance interventions. This enabled a detailed evaluation of how national antibiotic stewardship policies operated differently depending on healthcare institution type and policy period. Third, expert consultations were incorporated to support the interpretation of quantitative findings and to provide contextual explanations for heterogeneous policy effects. These findings may offer useful implications for other healthcare systems in which outpatient antibiotic prescribing is strongly influenced by patient expectations, competitive practice environments, and limited stewardship infrastructure.

Despite these strengths, this study has several limitations. First, as with all interrupted time-series designs, the influence of unmeasured confounders, such as concurrent healthcare policies, media coverage, changes in public awareness, or other secular trends, cannot be completely excluded. Second, some annual variables were converted into quarterly data using seasonal and linear interpolation, which may have introduced minor measurement error. Third, due to data constraints, the analysis relied on aggregated quarterly data rather than physician- or patient-level observations; therefore, micro-level variations in prescribing behavior could not be examined. Fourth, although expert consultations provided valuable contextual insights, the number and composition of participants were limited, and the findings should not be interpreted as representative of all healthcare providers. Finally, because Korea has a distinctive healthcare delivery system in which primary care clinics account for a large share of outpatient antibiotic prescribing, the generalizability of the findings to other countries may be limited. Future research should use individual-level patient and physician data, consider non-linear trends, and incorporate additional contextual factors occurring alongside policy implementation.

Conclusions

This study demonstrated that the effects of major AMR policies in Korea varied according to policy type, implementation period, and healthcare institution type, with important differences between short-term and long-term outcomes. These findings suggest that the introduction of policy instruments alone is insufficient to achieve sustained reductions in antibiotic prescribing.

The long-term effectiveness of AMR policies depends on whether healthcare institutions and providers can accept, implement, and sustain policy objectives within routine clinical practice. Future AMR strategies should therefore move beyond single-policy approaches and adopt institution-specific, multifaceted strategies that integrate regulation, antimicrobial stewardship, continuous monitoring and feedback, patient education, and One Health principles.

By using long-term time-series data and institution-specific analyses, this study provides evidence on both the strengths and limitations of existing AMR policies in Korea. Furthermore, antimicrobial resistance is a global One Health challenge that extends beyond national borders. Sustainable responses to AMR will require not only national efforts but also international collaboration and shared learning from global experiences. The findings of this study may contribute to the development of more effective and sustainable AMR policies in Korea and other healthcare systems facing similar challenges.

Abbreviations

AMR

Antimicrobial resistance

ARIMAX

Auto regressive integrated moving average

MRSA

Methicillin-resistant Staphylococcus aureus

EPAP

Evaluation project on appropriate prescribing

HIRA

Health Insurance Review and Assessment Service

ITS

Interrupted time series

SARIMAX

Seasonal auto regressive integrated moving average

NHIS

National Health Insurance Service

NHI

National Health Insurance

Q2

Second quarter

Q3

Third quarter

Q4

Fourth quarter

PPF

Physicians per facility

ASPs

Antimicrobial stewardship programs

POCT

Point-of-care testing

CRP

C-reactive protein

Jin Suk Kim

holds a PhD in Public Administration, with a specialization in public policy management. Her research focuses on evaluating the effects of health policy and institutional interventions, particularly in the areas of antibiotic use, antimicrobial resistance control policies, and the appropriateness of prescribing and reimbursement systems. She is currently affiliated with the Research Institute for Healthcare Policy, where she conducts policy evaluation studies using national administrative data, with methodological expertise in time series analysis and quasi-experimental research designs, including interrupted time series analysis.

Author contributions

JSK (Jin Suk Kim) developed the study concept and design, conducted the data curation and statistical analysis, and drafted the original manuscript. SKM (Seog-Kyun Mun), as the corresponding author, contributed to writing—review and editing, project administration, and supervision of the study. All authors have read and approved the final version of the manuscript.

Funding

The authors have not declared a specific grant for this research from any funding agency in the public, commercial or non-profit sectors.

Data availability

The datasets generated and analyzed during the current study are available in a public, open-access repository. The data used in this study are publicly available from the Health Insurance Review and Assessment Service (HIRA) of Korea at https://opendata.hira.or.kr/.

Declarations

Ethics approval and consent to participate

This study was conducted using complete quarterly national data covering 23 years (2000–2023), and no patients or members of the general community were involved. Therefore, this study is of minimal risk and does not fall under the definition of research involving human subjects. The study was granted exemption from ethical review by an institutional research ethics committee (Korean Medical Association Institutional Review Board, exemption number, 2025-05-01). Informed consent was not required.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.