Abstract
Background
The number of injured workers using Traditional Korean Medicine (KM) soared from 752 in 2013 to 10,273 in 2023, yet this trend remains underexplored. This cross-sectional study examined the frequency and factors linked to KM and Western Medicine (WM) use among injured workers.
Methods
Data from 6,985 participants in the 2018 and 2023 waves of the Panel Study of Workers’ Compensation Insurance (PSWCI) were analyzed. A hurdle model assessed outpatient visits: logistic regression for the probability of use and negative binomial regression for visit intensity. Demographic, socioeconomic, and clinical factors were included as covariates.
Results
Of 6,985 participants, 182 used KM, 5,970 used WM, and 833 reported no outpatient visits. Having private medical insurance and lower limb injuries showed higher Total Marginal Effects (TME) of both KM and WM utilization. KM use showed higher TMEs among females and those with occupational diseases but lower TMEs for injuries to other body regions. Higher WM utilization was linked to workers with disabilities, hypertension, chronic conditions, and extended rehabilitation, while the 2023 cohort, higher education and income, alcohol consumption, and re-employment were linked to lower WM use.
Conclusions
Between 2018 and 2023, adjusted KM utilization remained unchanged, whereas WM visits declined. KM use was positively linked to being female and having occupational diseases. WM visits tended to rise with greater medical severity yet decreased as socioeconomic conditions improved. Still, the rise in total KM claims under WCI indicates a growing demand for conservative, pain-relieving musculoskeletal care.
Keywords: Traditional Korean Medicine, Workers’ Compensation Insurance, Occupational injuries, Hurdle model, Total Marginal Effects
1. Introduction
The World Health Organization (WHO) underscores the importance of traditional medicine in achieving universal health coverage (UHC).1 In alignment with this goal, the Republic of Korea has integrated Traditional Korean Medicine (KM) into its health system to support UHC.2 Since 1987, KM therapies—including acupuncture, moxibustion, and cupping—have been reimbursed by the National Health Insurance Service (NHIS).3 Additionally, physical therapies offered at KM clinics, such as infrared, hot pack, and ice pack treatments, became reimbursable under NHIS in 2009.4
This combination has increased KM utilization for managing musculoskeletal disorders (MSDs). National claims data from 2013 to 2023 show that nine of the ten most frequently treated conditions in KM outpatient services were MSDs (Supplement 1).5 In addition, national survey results suggest that individuals perceived KM as effective for managing MSD symptoms, influenced by patient satisfaction and pain relief.6,7 This context motivated the government to designate KM as a covered treatment under Workers’ Compensation Insurance (WCI) in 1999.8
Since then, WCI has encouraged injured workers to access KM therapies alongside Western Medicine (WM). In 2009, WCI began to support injured workers' expenditure on herbal decoctions for up to 30 days, as they are not reimbursed under NHIS.9 KCOMWEL streamlined the approval process for six major MSDs as occupational diseases among manual workers in 2019.10,11 This led to a rapid increase in MSD patients covered by WCI (Fig. 1A).15, 16 In the same year, NHIS began covering Chuna manual therapy (up to 20 sessions per person), and KCOMWEL extended this coverage to WCI patients.12,13
Fig. 1.
National WCI trends, 2013–2023: (A) patient volume and expenditures for three main diseases; data from Ministry of Employment and Labor annual reports; (B) Distribution of Clinics – number of designated KM and WM clinics; data from KCOMWEL registry; (C) patient volume and expenditures for KM and WM; same source as (A).
CCVD: Cerebrocardiovascular Diseases; KCOMWEL: Korea Workers’ Compensation & Welfare Service; KM: Korean Medicine; WCI: Workers’ Compensation Insurance; KCOMWEL: Korea Workers’ Compensation & Welfare Service; WM: Western Medicine.
Two policies in 2019 have contributed to increased KM participation in the rehabilitation of MSDs among injured workers. Based on the WCI-designated medical institution registry, the number of KM institutions designated under WCI more than doubled (Fig. 1B), compared to a slight decrease in WM institutions.14 Improved institutional participation under the WCI corresponded with a more than tenfold increase in KM clinic visitors and expenditure, particularly after 2019 (Fig. 1C).15,16
Despite the upward trend, research on KM utilization for occupational injuries remains limited, largely due to insufficient real-world data. Thus, we utilized “the Panel Study of Workers’ Compensation Insurance (PSWCI)”, allowing for a more detailed analysis related to medical utilization. Guided by the hurdle model,17 we conducted a statistical analysis of medical visits involving both KM and WM. Based on data from the second and third PSWCI cohorts, this study aims to identify the sociodemographic and clinical factors related to outpatient utilization of KM and WM among workers covered by WCI.
2. Methods
2.1. Data source
The PSWCI has been conducted by KCOMWEL since 2013 under the Industrial Accident Compensation Insurance Act.18 The panel survey collects data on the personal characteristics, injury circumstances, and socioeconomic status of injured workers. Data are obtained through structured face-to-face interviews and KCOMWEL’s administrative records. The second PSWCI study targeted 75,392 workers who completed medical care in 2017 following an occupational injury; the analytical sample (n = 3,294) was drawn using stratified sampling based on gender, age group, and disability grade.19 Similarly, the third PSWCI study targeted 103,022 workers who completed medical care in 2022, and the analytical sample (n = 3,691) was stratified by age group, disability grade, and rehabilitation service use.17
2.2. Conceptual framework
This cross-sectional study utilized the hurdle model, also known as two-part model, to examine factors influencing medical utilization.20,21 The analysis was performed separately for KM and WM services. Incorporating demographic, socioeconomic, and clinical variables from PSWCI participants, we calculated both the probability and intensity of healthcare utilization for each variable. To assess the probability of utilization, we employed a logistic regression model across the entire study population. Intensity, defined as the frequency of prolonged medical visits, was analyzed using a negative binomial regression model among users of the same type of medical service.22 Lastly, we calculated total marginal effects (TME) by combining the probability and intensity estimates, interpreting the results as the expected change in the number of outpatient-visits relevant to each variable.23
2.3. Measures
2.3.1. PSWCI: 2018 versus 2023
We analyzed data from the second (2018, n = 3,294) and third (2023, n = 3,691) cohorts, totaling 6,985 participants. The first PSWCI cohort was excluded due to the absence of information on injury type and affected body parts. To minimize confounding from unrelated health events over time, only first-wave data from each cohort were used for statistical analysis. By comparing these data, we indirectly assessed KM and WM utilization patterns in relation to the 2019 policy changes.
2.3.2. Covariates
2.3.2.1. Demographic factors
Demographic characteristics included age, education level, marital status, and lifestyle factors such as smoking and drinking. Age group was categorized into <40, 40–49, 50–59, and ≥60 years, while education level was classified into middle school or below, high school, and college or above. Marital status was classified as unmarried or married. Lifestyle factors, such as smoking and drinking status, were included as binary variables.
2.3.2.2. Socioeconomic factors
Work-related and financial accessibility variables were considered as socioeconomic backgrounds. Work type was divided into binary variables, manual or non-manual, and annual earned income (KRW) was grouped into four categories: non-active, <15 million, 15–24 million, and ≥25 million. Private medical insurance status, a measure of healthcare financial accessibility, was binary (insured or uninsured). Employment status was divided into three categories: returned to original work, re-employed, or unemployed. These variables were used to assess socioeconomic influences on healthcare utilization.
2.3.2.3. Clinical factors
Clinical characteristics included disability, pre-existing chronic diseases, accident type, injury type, injury location, and rehabilitation duration. Disability registration, a proxy for accident severity, was analyzed as a binary variable for healthcare access. Pre-existing chronic diseases before injury were assessed with three variables: diabetes, hypertension, and other chronic conditions. The accident type was arranged as occupational disease, occupational accident, or commuting accident. Injury type was grouped into four categories: musculoskeletal conditions (e.g. sprain, lower back pain), fractures, external injuries (e.g. amputation, cut, abrasion), and internal/systemic conditions (e.g. infection, organ injury). Injury location was classified as head/neck, upper limb, lower limb, torso, or other. Rehabilitation duration was classified as <3 months, 3–6 months, 6 months–1 year, or >1 year, reflecting injury severity and recovery.
2.3.3. Medical utilization (KM/WM)
Medical utilization was assessed using two variables: the type of medical institution visited and the number of outpatient visits in the past year. Participants first reported the number of outpatient visits. Those with no visits were classified as non-clinic users. Respondents who used healthcare services could select from seven provider types: KCOMWEL hospitals, tertiary hospitals, local hospitals, public health centers, WM clinics, dental clinics, and KM clinics or hospitals.
For statistical analysis, outpatient service use was categorized into two outcomes: KM and WM. KM utilization was defined as visiting any KM clinic or hospital, and WM utilization as visiting any other provider type (e.g., KCOMWEL hospitals, WM clinics, tertiary hospitals, etc.). For both KM and WM, medical utilization was measured by the number of outpatient visits.
2.4. Statistical analysis
Due to the predominance of WM utilization (85.5 %), the KM data exhibited a high proportion of zero values, resulting in zero inflation. To address this and to assess both the likelihood and intensity of healthcare use, we employed the hurdle model.19,20 In the first part of the model, logistic regression was used to estimate the probability of utilizing KM or WM services, measured by the odds ratio (OR). In the second part, among those who used outpatient services, negative binomial regression was applied to analyze utilization intensity, measured by the incidence rate ratio (IRR). Although the outcome variable, counting outpatient visits, is nonnegative count data, Poisson regression was excluded due to overdispersion in both KM (variance 42) and WM (variance 38) data.21,24
To compare the hurdle and zero-inflated negative binomial (ZINB) models, the Vuong test for non-nested models and AIC were calculated. For KM utilization, the Vuong test favored the hurdle model (z = 3.31, p < .001), indicating a significantly better fit relative to the ZINB specification. In contrast, for WM utilization, the result showed no significant difference (z = 1.13, p = 0.136). Regarding AIC for KM utilization, the hurdle model had a lower AIC than the ZINB model (hurdle: 3309.72 vs. ZINB: 3395.46). The AIC for WM utilization revealed a similar pattern (hurdle: 52843.51 vs. ZINB: 52872.37). Based on these results, the hurdle model provided a better fit in both settings, compared to ZINB.
Based on the hurdle model results, we additionally calculated total marginal effects (TME), which estimate the expected change in the number of KM or WM outpatient visits associated with a one-unit change in each predictor. We employed TME to provide a more interpretable measure of effect size—expressing changes in visit counts rather than odds ratios or incidence rate ratios alone. TME integrates the average marginal effects (AME) from both the logistic and negative binomial components of the model, reflecting the combined changes in use probability and intensity among users. Confidence intervals (CIs) were calculated using the delta method under Wald assumptions, consistent with the AME estimation approach.25
Given the low KM utilization (182 participants, 2.6 %), propensity score matching (PSM) was not considered due to potential sample loss. As the study aimed to explore utilization patterns rather than causality, we used multivariate adjustments to control for confounders while maintaining full sample representativeness.
2.4.1. Multicollinearity assessment
To ensure the reliability of statistical analysis, Variance Inflation Factor (VIF) analysis was conducted on the negative binomial model in both KM and WM utilization using the car package in R studio (Supplement 2). All VIFs for WM utilization were within acceptable limits (VIF = 5). However, in the KM model, injury type exhibited a VIF of 7.6, suggesting moderate multicollinearity, possibly due to its association with the small KM user group (2.6 %). This suggests homogeneity within the subsample but did not compromise the overall model validity when comparing KM and WM utilization.
3. Results
3.1. Descriptive analysis
Table 1 presents the baseline characteristics of the second (n = 3,294) and the third (n = 3,691) PSWCI waves. To provide context for the subsequent logistic regression within the hurdle model, we examined healthcare utilization patterns between 2018 and 2023. The proportion of KM users increased from 2.0 % in 2018 to 3.1 % in 2023. Meanwhile, WM users comprised 86.1 % in 2018 and 84.9 % in 2023. This suggests that between 2018 and 2023, KM utilization may have become more accessible relative to WM use. By Pearson’s chi-square test for comparing two cohorts, the difference in utilization proportions between 2018 and 2023 was statistically significant (p = 0.0168).
Table 1.
Baseline Characteristics of Study Participants (N = 6,985; column %, within each survey).
| Baseline Characteristics | Second PSWCI (N = 3,294) |
Third PSWCI (N = 3,691) |
p-value | Total (N = 6,985) |
|
|---|---|---|---|---|---|
| Medical Utilization |
KM Users | 67 (2.0 %) | 115 (3.1 %) | 0.017* | 182 (2.6 %) |
| WM Users | 2,837 (86.1 %) | 3,133 (84.9 %) | 5,970 (85.5 %) | ||
| Non-clinic Users | 390 (11.8 %) | 443 (12.0 %) | 833 (11.9 %) | ||
| Gender | Male | 2,728 (82.8 %) | 3,062 (83.0 %) | 0.901 | 5,790 (82.9 %) |
| Female | 566 (17.2 %) | 629 (17.0 %) | 1,195 (17.1 %) | ||
| Age Group (in years) |
< 40 | 541 (16.4 %) | 590 (16.0 %) | <0.001* | 1,131 (16.2 %) |
| 40−49 | 682 (20.7 %) | 758 (20.5 %) | 1,440 (20.6 %) | ||
| 50−59 | 1,178 (35.8 %) | 1,027 (27.8 %) | 2,205 (31.6 %) | ||
| ≥ 60 years | 893 (27.1 %) | 1,316 (35.7 %) | 2,209 (31.6 %) | ||
| Education Level |
Middle School or Less | 1,184 (35.9 %) | 1,049 (28.4 %) | <0.001* | 2,233 (32.0 %) |
| High school | 1,525 (46.3 %) | 1,741 (47.2 %) | 3,266 (46.8 %) | ||
| College or Higher | 585 (17.8 %) | 901 (24.4 %) | 1,486 (21.3 %) | ||
| Marital Status |
Unmarried | 1,048 (31.8 %) | 1,375 (37.3 %) | <0.001* | 2,423 (34.7 %) |
| Married | 2,246 (68.2 %) | 2,316 (62.7 %) | 4,562 (65.3 %) | ||
| Smoking Status |
Non-smoker | 2,096 (63.6 %) | 2,425 (65.7 %) | <0.075 | 4,521 (64.7 %) |
| Smoker | 1,198 (36.4 %) | 1,266 (34.3 %) | 2,464 (35.3 %) | ||
| Drinking Status |
Non-drinker | 1,431 (43.4 %) | 1,818 (49.3 %) | <0.001* | 3,249 (46.5 %) |
| Drinker | 1,863 (56.6 %) | 1,873 (50.7 %) | 3,736 (53.5 %) | ||
| Work Type | Non-manual work | 584 (17.7 %) | 673 (18.2 %) | 0.606 | 1,257 (18.0 %) |
| Manual work | 2,710 (82.3 %) | 3,018 (81.8 %) | 5,728 (82.0 %) | ||
| Annual Earned Income (in million KRW) |
NA/non-active | 1,140 (34.6 %) | 1,333 (36.1 %) | <0.001* | 2,473 (35.4 %) |
| < 15 | 1,025 (31.1 %) | 934 (25.3 %) | 1,959 (28.0 %) | ||
| 15−24 | 531 (16.1 %) | 486 (13.2 %) | 1,017 (14.6 %) | ||
| ≥ 24 | 598 (18.2 %) | 938 (25.4 %) | 1,536 (22.0 %) | ||
| Private Medical Insurance | No | 1,407 (42.7 %) | 1,098 (29.7 %) | <0.001* | 2,505 (35.9 %) |
| Yes | 1,887 (57.3 %) | 2,593 (70.3 %) | 4,480 (64.1 %) | ||
| Employment Status |
Unemployed | 1,234 (37.5 %) | 1,350 (36.6 %) | <0.073 | 2,584 (37.0 %) |
| Re-employed | 1,101 (33.4 %) | 1,175 (31.8 %) | 2,276 (32.6 %) | ||
| Returned to original work | 959 (29.1 %) | 1,166 (31.6 %) | 2,125 (30.4 %) | ||
| Disability Registration | No | 2,874 (87.2 %) | 2,997 (81.2 %) | <0.001* | 5,871 (84.1 %) |
| Yes | 420 (12.8 %) | 694 (18.8 %) | 1,114 (15.9 %) | ||
| Hypertension | No | 2,827 (85.8 %) | 3,017 (81.7 %) | <0.001* | 5,844 (83.7 %) |
| Yes | 467 (14.2 %) | 674 (18.3 %) | 1,141 (16.3 %) | ||
| Diabetes | No | 3,077 (93.4 %) | 3,377 (91.5 %) | 0.003* | 6,454 (92.4 %) |
| Yes | 217 (6.6 %) | 314 (8.5 %) | 531 (7.6 %) | ||
| Other Chronic Diseases | No | 2,965 (90.0 %) | 3,315 (89.8 %) | 0.814 | 6,280 (89.9 %) |
| Yes | 329 (10.0 %) | 376 (10.2 %) | 705 (10.1 %) | ||
| Accident Type | Occupational accident | 3,092 (93.9 %) | 2,934 (79.5 %) | <0.001* | 6,026 (86.3 %) |
| Occupational disease | 185 (5.6 %) | 564 (15.3 %) | 749 (10.7 %) | ||
| Commuting accident | 17 (0.5 %) | 193 (5.2 %) | 210 (3.0 %) | ||
| Injury Type | Musculoskeletal conditions | 175 (5.3 %) | 423 (11.5 %) | <0.001* | 598 (8.6 %) |
| Fractures | 1,895 (57.5 %) | 1,898 (51.4 %) | 3,793 (54.3 %) | ||
| External Injuries | 1,085 (32.9 %) | 1,129 (30.6 %) | 2,214 (31.7 %) | ||
| Internal and Systemic conditions | 139 (4.2 %) | 241 (6.5 %) | 380 (5.4 %) | ||
| Injury Location |
Head and Neck | 340 (10.3 %) | 359 (9.7 %) | 0.441 | 699 (10.0 %) |
| Upper Limb | 1,410 (42.8 %) | 1,548 (41.9 %) | 2,958 (42.3 %) | ||
| Lower Limb | 585 (17.8 %) | 639 (17.3 %) | 1,224 (17.5 %) | ||
| Torso | 862 (26.2 %) | 1,018 (27.6 %) | 1,880 (26.9 %) | ||
| Other | 97 (2.9 %) | 127 (3.4 %) | 224 (3.2 %) | ||
| Rehabilitation Period |
< 3 months | 624 (18.9 %) | 507 (13.7 %) | <0.001* | 1,131 (16.2 %) |
| 3−6 months | 1,217 (36.9 %) | 1,159 (31.4 %) | 2,376 (34.0 %) | ||
| 6 months−1 year | 966 (29.3 %) | 1,141 (30.9 %) | 2,107 (30.2 %) | ||
| ≥ 1 year | 487 (14.8 %) | 884 (24.0 %) | 1,371 (19.6 %) | ||
(PSWCI: Panel Study of Workers’ Compensation Insurance; KM: Korean Medicine; WM: Western Medicine).
* p < 0.05 (by Pearson’s chi-square test).
The two cohorts differed significantly in several characteristics. Participants in 2023 were older, had higher education and income levels, were less likely to drink alcohol, and were more often covered by private medical insurance than those in 2018 (all p < 0.05). Clinically, the 2023 group had higher rates of disability, hypertension, diabetes, occupational disease, fractures, and longer rehabilitation periods, whereas non-significant variables—including sex, smoking, job type, employment status, other chronic diseases, and injury site—were comparable between cohorts.
3.2. KM utilization analysis
Table 2 presents the results of KM utilization based on the hurdle model. In the logistic regression, the low KM utilization rate (2.6 %) resulted in an intercept value near zero. In the probability component, KM utilization was more likely among women, those with private insurance, mid-income earners (15–24 million KRW), workers diagnosed with occupational disease, fracture cases, and lower-limb injuries. In contrast, injuries to other body regions were rarely treated with KM.
Table 2.
Results from the hurdle model for KM outpatient visits using pooled, unweighted data from the 2018 and 2023 cohorts (Logistic: n = 6,985; Negative Binomial: n = 182).
| Variable | Probability (Logistic Regression) |
Intensity (Negative Binomial Regression) |
Total Marginal Effect | |||
|---|---|---|---|---|---|---|
| OR (95 % CI) | p-value | IRR (95 % CI) | p-value | Visits (95 % CI) | ||
| PSWCI | Second PSWCI | Reference | - | Reference | - | Reference |
| Third PSWCI | 1.32 (0.95, 1.84) | 0.103 | 0.90 (0.59, 1.37) | 0.620 | 0.15 (-0.32, 0.62) | |
| Gender | Male | Reference | - | Reference | - | Reference |
| Female | 1.73 (1.14, 2.62) | 0.010* | 1.13 (0.66, 1.92) | 0.655 | 0.70 (0.02, 1.36)* | |
| Age Group (in years) |
< 40 | Reference | - | Reference | - | Reference |
| 40−49 | 0.84 (0.48, 1.47) | 0.547 | 1.52 (0.77, 3.00) | 0.228 | 0.20 (-0.52, 0.93) | |
| 50−59 | 0.93 (0.54, 1.58) | 0.779 | 1.11 (0.57, 2.15) | 0.754 | 0.01 (-0.63, 0.64) | |
| ≥ 60 years | 1.13 (0.62, 2.03) | 0.693 | 1.50 (0.70, 3.21) | 0.295 | 0.45 (-0.33, 1.22) | |
| Education Level | Middle School or Less | Reference | - | Reference | - | Reference |
| High school | 1.03 (0.70, 1.52) | 0.865 | 0.88 (0.53, 1.47) | 0.630 | -0.09 (-0.65, 0.46) | |
| College or Higher | 0.83 (0.49, 1.41) | 0.502 | 0.66 (0.34, 1.27) | 0.212 | -0.51 (-1.18, 0.16) | |
| Marital Status |
Unmarried | Reference | - | Reference | - | Reference |
| Married | 0.94 (0.67, 1.31) | 0.704 | 1.10 (0.74, 1.64) | 0.645 | 0.02 (-0.44, 0.49) | |
| Disability Registration | Non-disabled | Reference | - | Reference | - | Reference |
| Disabled | 0.94 (0.59, 1.49) | 0.781 | 1.62 (0.88, 3.00) | 0.120 | 0.47 (-0.31, 1.26) | |
| Smoking Status |
Non-smoker | Reference | - | Reference | - | Reference |
| Smoker | 0.96 (0.66, 1.39) | 0.832 | 1.03 (0.64, 1.65) | 0.901 | -0.02 (-0.54, 0.50) | |
| Drinking Status |
Non-drinker | Reference | - | Reference | - | Reference |
| Drinker | 1.04 (0.75, 1.45) | 0.808 | 0.69 (0.43, 1.12) | 0.133 | -0.29 (-0.75, 0.18) | |
| Annual Earned Income (in million KRW) |
NA/non-active | Reference | - | Reference | - | Reference |
| < 15 | 1.37 (0.89, 2.10) | 0.154 | 0.96 (0.54, 1.71) | 0.884 | 0.23 (-0.35, 0.80) | |
| 15−24 | 1.77 (1.06, 2.97) | 0.029* | 0.71 (0.35, 1.41) | 0.324 | 0.29 (-0.43, 1.00) | |
| ≥ 24 | 1.24 (0.73, 2.11) | 0.436 | 1.43 (0.69, 2.96) | 0.338 | 0.55 (-0.32, 1.42) | |
| Work Type | Non-manual work | Reference | - | Reference | - | Reference |
| Manual work | 0.75 (0.51, 1.11) | 0.146 | 0.85 (0.54, 1.35) | 0.495 | -0.44 (-1.03, 0.14) | |
| Employment Status | Unemployed | Reference | - | Reference | - | Reference |
| Re-employed | 0.96 (0.64, 1.43) | 0.841 | 0.99 (0.58, 1.70) | 0.973 | -0.04 (-0.61, 0.52) | |
| Returned to original work |
0.90 (0.58, 1.42) | 0.663 | 0.86 (0.47, 1.58) | 0.635 | -0.21 (-0.83, 0.40) | |
| Private Medical Insurance | No | Reference | - | Reference | - | Reference |
| Yes | 1.48 (1.04, 2.12) | 0.029* | 1.32 (0.83, 2.10) | 0.234 | 0.57 (0.14, 1.00)* | |
| Hypertension | No | Reference | - | Reference | - | Reference |
| Yes | 0.91 (0.59, 1.40) | 0.675 | 1.26 (0.74, 2.15) | 0.396 | 0.14 (-0.48, 0.75) | |
| Diabetes | No | Reference | - | Reference | - | Reference |
| Yes | 0.87 (0.46, 1.62) | 0.656 | 1.28 (0.55, 2.97) | 0.567 | 0.14 (-0.82, 1.11) | |
| Other Chronic Diseases | No | Reference | - | Reference | - | Reference |
| Yes | 1.08 (0.67, 1.75) | 0.743 | 1.27 (0.71, 2.30) | 0.422 | 0.32 (-0.42, 1.05) | |
| Accident Type | Occupational accident | Reference | - | Reference | - | Reference |
| Occupational disease | 4.45 (2.34, 8.45) | 0.000* | 1.32 (0.59, 2.94) | 0.504 | 2.64 (0.84, 4.44)* | |
| Commuting accident | 1.34 (0.65, 2.76) | 0.425 | 0.52 (0.21, 1.30) | 0.164 | -0.16 (-0.98, 0.66) | |
| Injury Type | Musculoskeletal conditions | Reference | - | Reference | - | Reference |
| Fractures | 2.26 (1.10, 4.63) | 0.026* | 0.99 (0.41, 2.35) | 0.976 | 0.67 (-0.16, 1.51) | |
| External Injuries | 1.42 (0.72, 2.79) | 0.306 | 1.07 (0.44, 2.65) | 0.877 | 0.29 (-0.50, 1.08) | |
| Internal and Systemic conditions | 1.42 (0.57, 3.56) | 0.451 | 0.97 (0.34, 2.78) | 0.949 | 0.19 (-0.87, 1.26) | |
| Injury Location |
Head and Neck | Reference | - | Reference | - | Reference |
| Upper Limb | 1.04 (0.51, 2.12) | 0.924 | 0.90 (0.34, 2.40) | 0.830 | -0.05 (-0.86, 0.75) | |
| Lower Limb | 2.30 (1.11, 4.75) | 0.025* | 1.54 (0.61, 3.87) | 0.356 | 1.31 (0.36, 2.26)* | |
| Torso | 1.15 (0.55, 2.40) | 0.717 | 1.49 (0.54, 4.13) | 0.444 | 0.46 (-0.45, 1.38) | |
| Other | 0.40 (0.09, 1.80) | 0.230 | 0.10 (0.01, 0.72) | 0.022* | -1.16 (-1.99, -0.33)* | |
| Rehabilitation Period | < 3 months | Reference | - | Reference | - | Reference |
| 3−6 months | 1.22 (0.73, 2.03) | 0.447 | 1.17 (0.59, 2.33) | 0.647 | 0.30 (-0.33, 0.93) | |
| 6 months−1 year | 1.37 (0.81, 2.32) | 0.235 | 1.04 (0.54, 2.00) | 0.903 | 0.32 (-0.31, 0.95) | |
| ≥ 1 year | 1.04 (0.55, 1.98) | 0.904 | 1.83 (0.76, 4.43) | 0.178 | 0.66 (-0.27, 1.58) | |
| Intercept | Intercept | 0.01 (0.00, 0.02) | <0.001* | 16.71 (3.30, 84.53) | <0.001* | |
CI: Confidence Interval; IRR: Incidence Rate Ratio; KM: Korean Medicine; OR: Odds Ratio; PSWCI: Panel Study of Workers’ Compensation Insurance; TME: Total Marginal Effect.
* p-value < 0.05.
Intercept corresponds to the predicted count when all covariates are at reference levels.
Model fit: AIC = 3309.7.
Among KM users, outpatient visits showed no notable variation across most covariates, except that injuries to other regions were associated with markedly fewer visits. The estimates for TME confirmed the same pattern: the largest increases in expected KM visits occurred in occupational-disease cases, lower-limb injuries, female and private medical insurance, while other variables showed minimal impact.
3.3. WM utilization analysis
As shown in Table 3, WM services remained dominant. WM use was more probable in older workers and those with hypertension, diabetes, or other chronic diseases, but less likely among smokers, drinkers, higher-income groups (over 24 million KRW), upper limb injuries, and workers who were re-employed or had returned to their jobs.
Table 3.
Results from the hurdle model for WM outpatient visits using pooled, unweighted data from the 2018 and 2023 cohorts (Logistic: n = 6,985; Negative Binomial: n = 5,970).
| Variable | Probability (Logistic Regression) |
Intensity (Negative Binomial Regression) |
Total Marginal Effect | |||
|---|---|---|---|---|---|---|
| OR (95 % CI) | p-value | IRR (95 % CI) | p-value | Visits (95 % CI) | ||
| PSWCI | Second PSWCI | Reference | - | Reference | - | Reference |
| Third PSWCI | 0.87 (0.76, 1.01) | 0.068 | 0.74 (0.69, 0.80) | <0.001* | -5.78 (-7.07, -4.48)* | |
| Gender | Male | Reference | - | Reference | - | Reference |
| Female | 1.04 (0.84, 1.29) | 0.738 | 1.02 (0.92, 1.14) | 0.693 | 0.40 (-1.44, 2.23) | |
| Age Group (in years) |
< 40 | Reference | - | Reference | - | Reference |
| 40−49 | 1.23 (0.99, 1.52) | 0.059 | 1.05 (0.93, 1.19) | 0.409 | 1.34 (-0.59, 3.28) | |
| 50−59 | 1.34 (1.08, 1.66) | 0.007* | 1.17 (1.03, 1.32) | 0.013* | 3.40 (1.47, 5.34)* | |
| ≥ 60 years | 1.24 (0.97, 1.60) | 0.089 | 1.16 (1.01, 1.33) | 0.030* | 3.10 (0.88, 5.32)* | |
| Education Level | Middle School or Less | Reference | - | Reference | - | Reference |
| High school | 1.03 (0.86, 1.23) | 0.739 | 0.91 (0.83, 1.00) | 0.0499* | -1.69 (-3.27, -0.10)* | |
| College or Higher | 1.19 (0.95, 1.50) | 0.138 | 0.84 (0.75, 0.95) | 0.005* | -2.55 (-4.52, -0.58)* | |
| Marital Status |
Unmarried | Reference | - | Reference | - | Reference |
| Married | 1.04 (0.89, 1.21) | 0.636 | 0.91 (0.84, 0.99) | 0.028* | -1.52 (-2.90, -0.14)* | |
| Disability Registration | Non-disabled | Reference | - | Reference | - | Reference |
| Disabled | 1.07 (0.86, 1.32) | 0.551 | 1.31 (1.18, 1.45) | <0.001* | 5.42 (3.43, 7.42)* | |
| Smoking Status |
Non-smoker | Reference | - | Reference | - | Reference |
| Smoker | 0.81 (0.70, 0.94) | 0.007* | 1.09 (1.00, 1.18) | 0.046* | 0.90 (-0.51, 2.30) | |
| Drinking Status |
Non-drinker | Reference | - | Reference | - | Reference |
| Drinker | 0.86 (0.74, 0.99) | 0.037* | 0.88 (0.81, 0.95) | 0.001* | -2.87 (-4.16, -1.57)* | |
| Annual Earned Income (in million KRW) |
NA/non-active | Reference | - | Reference | - | Reference |
| < 15 | 0.75 (0.61, 0.91) | 0.004* | 1.06 (0.96, 1.17) | 0.282 | 0.17 (-1.62, 1.96) | |
| 15−24 | 0.73 (0.57, 0.92) | 0.009* | 0.85 (0.75, 0.96) | 0.009* | -3.76 (-5.77, -1.76)* | |
| ≥ 24 | 0.72 (0.57, 0.91) | 0.005* | 0.82 (0.73, 0.93) | 0.002* | -4.29 (-6.19, -2.40)* | |
| Work Type | Non-manual work | Reference | - | Reference | - | Reference |
| Manual work | 1.09 (0.90, 1.31) | 0.367 | 1.01 (0.92, 1.12) | 0.790 | 0.52 (-1.16, 2.20) | |
| Employment Status | Unemployed | Reference | - | Reference | - | Reference |
| Re-employed | 0.77 (0.64, 0.92) | 0.005* | 0.65 (0.60, 0.72) | <0.001* | -8.30 (-9.89, -6.71)* | |
| Returned to original work |
0.80 (0.65, 0.98) | 0.034* | 0.69 (0.62, 0.76) | <0.001* | -7.54 (-9.35, -5.73)* | |
| Private Medical Insurance | No | Reference | - | Reference | - | Reference |
| Yes | 0.92 (0.79, 1.07) | 0.258 | 1.12 (1.04, 1.21) | 0.003* | 1.81 (0.53, 3.08)* | |
| Hypertension | No | Reference | - | Reference | - | Reference |
| Yes | 1.67 (1.32, 2.10) | <0.001* | 1.16 (1.05, 1.29) | 0.003* | 4.15 (2.34, 5.96)* | |
| Diabetes | No | Reference | - | Reference | - | Reference |
| Yes | 1.75 (1.23, 2.48) | 0.002* | 1.01 (0.89, 1.16) | 0.858 | 1.53 (-0.73, 3.79) | |
| Other Chronic Diseases | No | Reference | - | Reference | - | Reference |
| Yes | 1.56 (1.19, 2.06) | 0.001* | 1.08 (0.99, 1.19) | 0.083 | 2.81 (0.77, 4.85)* | |
| Accident Type | Occupational accident | Reference | - | Reference | - | Reference |
| Occupational disease | 0.99 (0.72, 1.37) | 0.959 | 1.10 (0.93, 1.30) | 0.278 | 1.50 (-1.41, 4.40) | |
| Commuting accident | 0.81 (0.56, 1.19) | 0.282 | 1.08 (0.88, 1.34) | 0.458 | 0.86 (-2.98, 4.69) | |
| Injury Type | Musculoskeletal conditions | Reference | - | Reference | - | Reference |
| Fractures | 0.82 (0.58, 1.16) | 0.259 | 1.14 (0.95, 1.37) | 0.172 | 1.61 (-1.21, 4.44) | |
| External Injuries | 0.95 (0.67, 1.34) | 0.767 | 1.04 (0.87, 1.26) | 0.648 | 0.47 (-2.33, 3.27) | |
| Internal and Systemic conditions | 0.67 (0.43, 1.04) | 0.073 | 0.88 (0.71, 1.10) | 0.274 | -3.10 (-6.41, 0.21) | |
| Injury Location |
Head and Neck | Reference | - | Reference | - | Reference |
| Upper Limb | 0.71 (0.53, 0.95) | 0.023* | 0.85 (0.74, 0.97) | 0.018* | -3.65 (-5.91, -1.38)* | |
| Lower Limb | 0.85 (0.61, 1.17) | 0.319 | 1.21 (1.04, 1.41) | 0.014* | 3.46 (0.73, 6.20)* | |
| Torso | 0.85 (0.63, 1.15) | 0.300 | 1.08 (0.94, 1.25) | 0.276 | 1.09 (-1.36, 3.54) | |
| Other | 0.95 (0.59, 1.53) | 0.827 | 0.91 (0.73, 1.13) | 0.400 | -1.74 (-5.26, 1.78) | |
| Rehabilitation Period | < 3 months | Reference | - | Reference | - | Reference |
| 3−6 months | 0.99 (0.81, 1.21) | 0.940 | 1.34 (1.20, 1.50) | <0.001* | 3.87 (2.44, 5.31)* | |
| 6 moths−1 year | 1.02 (0.82, 1.26) | 0.848 | 1.68 (1.50, 1.88) | <0.001* | 7.89 (6.26, 9.53)* | |
| ≥ 1 year | 1.08 (0.83, 1.40) | 0.582 | 2.03 (1.76, 2.33) | <0.001* | 12.09 (9.81, 14.37)* | |
| Intercept | Intercept | 9.76 (5.63, 16.92) | <0.001* | 15.40 (11.60, 20.46) | <0.001* | - |
CI: Confidence Interval; IRR: Incidence Rate Ratio; OR: Odds Ratio; PSWCI: Panel Study of Workers’ Compensation Insurance; TME: Total Marginal Effect; WM: Western Medicine.
* p-value < 0.05.
Intercept corresponds to the predicted count when all covariates are at reference levels.
Model fit: AIC = 52843.5.
Among WM visitors, outpatient visits significantly rose with disability, older age, private insurance, chronic conditions, lower-limb injuries, and longer rehabilitation periods, but declined with higher education, marriage, alcohol consumption, higher income (over 24 million KRW), re-employment or return-to-work status, and upper-limb injuries.
Correspondingly, TME figures indicated that medical-severity variables, such as disability and prolonged rehab period drove larger increases in WM visits, whereas favorable socioeconomic factors, such as higher education, higher income, and reemployment status were linked to fewer visits.
3.4. Comparison between KM and WM
Our pooled analyses included two PSWCI cohorts, introducing the possibility that certain statistically significant findings may reflect temporal variation. For instance, the cohort variable was related to a decrease in WM utilization. This may reflect the impact of COVID-19, suppressing the demands for outpatient visits. Additionally, occupational diseases were linked to higher KM utilization, possibly indicating greater use among patients in the chronic phase.
Overall, the number of KM users was relatively small, leading to wide confidence intervals, which limited statistical precision. These factors should be taken into account when interpreting the results, especially those pertaining to the KM subgroup. Unlike TME for WM use, several KM-related marginal effects correspond to fewer than one additional visit per year. This likely reflects the generally low utilization frequency (2.6 %) in the broader population. Nevertheless, these findings remain informative for identifying demographic groups with relatively higher usage and may offer insights for targeting or understanding healthcare choices among specific subpopulations.
4. Discussion
4.1. Summary of main findings
When comparing the two groups from 2018 and 2023, the 2019 policy changes did not lead to a statistically significant difference in KM utilization based on TME. However, WM utilization decreased in 2023 compared to 2018. Unlike KM, WM services may have faced greater disruption due to COVID-19 restrictions or shifting patient behavior. The statistical analysis also identified several demographic and economic factors influencing KM utilization. Aligning with earlier research,26, 27, 28, 29 female and private insurance were positively linked to KM utilization. KM visits were higher in occupational disease cases than occupational accidents. This correlation is likely due to the suitability of KM treatments for managing long-term symptoms rather than providing immediate, acute medical interventions.
4.2. Comparison with previous studies
In the past, KM utilization within WCI has only been analyzed through clinical records. A 2010 study30 found that 80.4 % of occupationally injured patients in a KM hospital also received WM services. Similarly, a 2003 study31 showed that 82.5 % of patients admitted to a KM hospital were transferred from WM hospitals after surgical procedures. For severe MSDs such as fractures,32,33 ruptures,34 and spinal cord injuries,35 previous studies highlight KM’s role in relieving symptoms and enhancing the quality of life in patients. These findings help contextualize the present study’s results, particularly in light of KM’s role in treating chronic MSDs. As conservative treatments, KM for MSDs may expand its role toward integrative medicine.
4.3. Clinical and research implications
The increasing proportion of MSDs likely contributed to higher KM participation, given KM’s role in MSD rehabilitation. In 2019, KCOMWEL streamlined the process for recognizing six MSDs as occupational diseases in manual workers, including cervical herniation, lumbar herniation, rotator cuff rupture, epicondylitis, meniscus cartilage rupture, and carpal tunnel syndrome.10 Most of these conditions overlap with the most common KM visits over the past decade,5 and this may reflect people’s preference for KM for MSDs management. Furthermore, KM’s growing use in occupational diseases suggests a shift in occupational healthcare priorities toward long-term symptom management.
4.4. Strengths and limitations
Our study’s main strength is simultaneously analyzing the utilization of Western and traditional medicine using the PSWCI data, which has not previously been applied in KM research. Several countries have already incorporated traditional medicine, particularly acupuncture, into workers’ compensation systems to enhance occupational health (Supplement 3). Even so, most countries have not conducted national-level panel surveys that link occupational health with traditional medicine use. KCOMWEL, as a centralized administrative body, enabled thorough analysis of demographic, socioeconomic, and clinical characteristics related to occupational health.
Nonetheless, there are several limitations regarding statistical analysis. Given the relatively small number of KM users, the first-stage logistic model may be susceptible to overfitting. Also, the large number of predictors means that some significant associations may have occurred by chance. This concern is particularly relevant for the KM model, in which the small sample size further reduces statistical precision and increases the risk of type I error. Due to computational constraints and methodological considerations, the delta method was used to calculate confidence intervals for the TMEs. However, future studies could explore bootstrapping as an alternative estimation method. In addition, because we employed a hurdle function from the pscl package in R, it was not possible to conduct weighted analysis concurrently. Future research may provide more generalizable results by incorporating survey weights.
Regarding research outcomes, we analyzed panel survey data only for the period following WCI coverage implementation. In other words, we were unable to assess actual healthcare costs and clinical records. Moreover, our KM analysis was limited to 182 out of 6,985 panel members, which may introduce selection bias. Future research should examine actual patient data to better characterize injured workers who use KM. The aggregation of additional data will enable researchers to evaluate its impact on pain management, long-term medical costs, and overall quality of life.
4.5. Conclusions and suggestions for future studies
Using a hurdle model with PSWCI data, we found that there was no statistically significant difference in the TME for KM utilization, whereas WM utilization showed a significant decrease over the same period. Due to the low overall proportion of KM users (2.6 %), only a few factors—such as being female, having private insurance, and sustaining a lower-limb joint injury—showed positive correlation with KM utilization, while many other variables did not reach statistical significance. In contrast, WM utilization showed negative correlation with higher education, higher income, and being reemployed, whereas it was positively linked to medical severity factors such as disability registration, rehabilitation period, and the presence of hypertension or other chronic diseases. These findings suggest that medical severity increases WM utilization, whereas socioeconomic factors related to returning to work are relevant to lower WM use.
Overall, these findings highlight the need for further research incorporating clinical outcomes and long-term data will strengthen the evidence base for integrating traditional medicine into occupational healthcare, ultimately promoting a more effective workers’ welfare system. Also, this study highlights the need for continued research into how policy shifts influence KM utilization patterns in workers’ compensation systems. Over the past decade, KM users and medical expenditure have increased more than tenfold under WCI, in parallel with both expanded coverage and a rise in occupational patients with MSDs. While this study primarily examined policy-level impacts, the growing KM use among injured workers may also reflect its clinical value—such as pain relief, functional improvement, and support for return-to-work. Accordingly, future growth will depend not only on policy design, but also on the accumulation of clinical evidence supporting KM’s role in occupational injury care.
Author contributions
SP: Data curation, Formal analysis, Investigation, Writing-Original draft preparation.
BJ: Conceptualization, Methodology, Supervision, Writing- Reviewing and Editing.
Conflict of interest
The authors declare that they have no conflicts of interest.
Funding
No funding was received for this work.
Ethics statement
This study was approved by the Institutional Review Board (IRB) of Kyung Hee University (No. KHSIRB-24-565(EA)). As the study analyzed publicly available data, patient consent was not required. All personal information was de-identified by KCOMWEL before public release. Due to the study's retrospective nature, the IRB waived the requirement for informed consent.
Data availability
The data that support the findings of this study are openly available at: https://www.pswci.or.kr/.
Acknowledgements
The authors would like to thank KCOMWEL for conducting the PSWCI.
Footnotes
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.imr.2025.101202.
Appendix. Supplementary materials
References
- 1.World Health Organization. WHO convenes first high-level global summit on traditional medicine to explore evidence base, opportunities to accelerate health for all. WHO Website. https://www.who.int/news/item/10-08-2023-who-convenes-first-high-level-global-summit-on-traditional-medicine-to-explore-evidence-base-opportunities-to-accelerate-health-for-all. Published 2023. Accessed February 10, 2025.
- 2.Park Y.R., Canaway R. Integrating traditional and complementary medicine with national healthcare systems for universal health coverage in Asia and the Western Pacific. Health Syst Reform. 2019;5(1):24–31. doi: 10.1080/23288604.2018.1539058. [DOI] [PubMed] [Google Scholar]
- 3.Park M.J., Lim B.M., Cha W.S., You MS. Institutional analysis on organizational changes of Korean medicine hospitals. J Korean Med. 2014;35(1):145–156. [Google Scholar]
- 4.Shin M.S., Shin B.C., Lee M.J., Kim H.J., Song Y.K., Song M.Y., et al. The utilization of physical therapies in the Korean medicine clinics: a survey for national health insurance planning. J Korean Med Rehab. 2009;19(3):101–109. [Google Scholar]
- 5.Health Insurance Review & Assessment Service (HIRA). Healthcare Big Data Hub. HIRA Website. https://opendata.hira.or.kr/op/opc/olapHifrqSickInfoTab1.do. Published 2024. Accessed February 10, 2025.
- 6.Sung A.D., Choi S.Y., Park M.J., Sung SH. A study on the characteristics of the awareness and satisfication[sic] level of musculoskeletal patients and non-musculoskeletal patients that use Korean medical institutions. J Soc Prev Korean Med. 2020;24(3):27–38. [Google Scholar]
- 7.Wang B.R., Choi I.Y., Kim K.J., Kwon YD. Use of traditional Korean medicine by patients with musculoskeletal disorders. PLoS One. 2013;8(5) doi: 10.1371/journal.pone.0063209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kim H.N., Yi E.H., Kim DS. The effects of occupation on the use of Korean Medicine: using population-based survey data. Eur J Integr Med. 2022;54 doi: 10.1016/j.eujim.2022.102169. [DOI] [Google Scholar]
- 9.Ministry of Employment and Labor. Medical care benefit criteria under the Industrial Accident Compensation Insurance Act [administrative rule]. 2022. Available from: https://www.law.go.kr/행정규칙/산업재해보상보험요양급여산정기준. Accessed Feb 10, 2025.
- 10.Korea Workers’ Compensation & Welfare Service. Expanded treatment coverage during special checkups for work-related risks of the six major musculoskeletal disorders. KCOMWEL Website. https://www.comwel.or.kr/_custom/kcom/_common/board/download.jsp?attach_no=1801865. Published 2019. Accessed February 10, 2025.
- 11.Ministry of Employment and Labor. Partial revision notice: criteria for recognizing work-relatedness of cerebrovascular, cardiovascular, and musculoskeletal diseases. Ministry of Employment and Labor Website. https://www.moel.go.kr/info/lawinfo/instruction/view.do?bbs_seq=20220402523 Published 2022. Accessed February 10, 2025.
- 12.Chae J., Lee C., Shin K. A study of the effect on the public and private insurances caused by National Health Insurance coverage of Chuna treatment. J Korean Insur Assoc. 2019 Apr;118:105–138. [Google Scholar]
- 13.Ministry of Health and Welfare. Chuna therapy is now covered by health insurance and medical benefits. Ministry of Health and Welfare website. https://www.mohw.go.kr/board.es?mid=a10503010100&bid=0027&tag=&act=view&list_no=348728&cg_code= Published 2019. Accessed February 10, 2025.
- 14.Korea Workers’ Compensation & Welfare Service. Public Data: designated medical institutions for workers’ compensation insurance. Public Data website. https://www.data.go.kr/data/3044320/fileData.do. Published 2024. Accessed February 10, 2025.
- 15.Ministry of Employment and Labor. Workers' compensation insurance annual report 2015. Sejong: Ministry of Employment and Labor; published 2017. p. 240-242. Available from: https://moel.go.kr/info/publict/publictDataView.do?bbs_seq=1483491829257. Accessed February 10, 2025
- 16.Ministry of Employment and Labor. Workers' Compensation Insurance annual report 2023. Sejong: Ministry of Employment and Labor; Published 2024. p. 258−260,336. Available from: https://www.moel.go.kr/info/publict/publictDataView.do?bbs_seq=20241001451. Accessed February 10, 2025
- 17.Lei X., Lin W. The impacts of health insurance on health care utilization among the older people in China. Soc Sci Med. 2013;85:59–65. doi: 10.1016/j.socscimed.2013.02.037. [DOI] [PubMed] [Google Scholar]
- 18.Korea Workers’ Compensation & Welfare Service Labor Welfare Research Institute. An introduction to panel study of workers’ compensation insurance (PSWCI). Published 2024. p. 2-4. Available from: https://www.pswci.or.kr/Introduction/DataSource/%EC%82%B0%EC%9E%AC%EB%B3%B4%ED%97%98%ED%8C%A8%EB%84%90%EC%A1%B0%EC%82%AC%20%EC%98%81%EB%AC%B8%20%EC%86%8C%EA%B0%9C%EC%9E%90%EB%A3%8C_%20An%20Introduction%20to%20PSWCI.pdf?rnd=96764.6888907834. Accessed February 10, 2025.
- 19.Korea Workers’ Compensation & Welfare Service. Survey design of the panel study of workers’ compensation insurance (PSWCI). Published 2023. Available from: https://www.pswci.or.kr/Introduction/Plan.aspx?rnd=376324.539248051. Accessed June 6, 2025.
- 20.Kim D., Lee Y.J., Jang B.H., Park J.S., Park S., D’Adamo C.R., et al. Analysis of factors associated with the use of Korean medicine after spinal surgery using a nationwide database in Korea. Sci Rep. 2023;13 doi: 10.1038/s41598-023-47454-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Belotti F., Deb P., Manning W.G., Norton EC. twopm: two-part models. Stata J. 2015 Apr;15(1):3–20. [Google Scholar]
- 22.Le D.D., Gonzalez R.L., Matola JU. Modeling count data for health care utilization: an empirical study of outpatient visits among Vietnamese older people. BMC Med Inform Decis Mak. 2021;21(1):265. doi: 10.1186/s12911-021-01619-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Liu X., Chen W., Chen T., Zhang H., Zhang B. Marginal effects and incremental effects in two-part models for endogenous healthcare utilization in health services research. Health Serv Outcomes Res Method. 2020;20:111–139. doi: 10.1007/s10742-020-00211-x. [DOI] [Google Scholar]
- 24.Bao Y. Predicting the use of outpatient mental health services: do modeling approaches make a difference? Inquiry. 2002;39(2):168–183. doi: 10.5034/inquiryjrnl_39.2.168. [DOI] [PubMed] [Google Scholar]
- 25.Arel-Bundock V., Greifer N., Heiss A. How to interpret statistical models using marginal effects in R. J Stat Softw. 2024 Nov;111(9):1–32. doi: 10.18637/jss.v111.i09. [DOI] [Google Scholar]
- 26.Huang C.W., Hwang I.H., Lee Y.S., Hwang S.J., Ko S.G., Chen F.P., et al. Utilization patterns of traditional medicine in Taiwan and South Korea by using national health insurance data in 2011. PLoS One. 2018;13(12) doi: 10.1371/journal.pone.0208569. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Alwhaibi M., AlRuthia Y., Meraya AM. Gender differences in the prevalence of complementary and alternative medicine utilization among adults with arthritis in the United States. Evid Based Complement Alternat Med. 2019;2019(1) doi: 10.1155/2019/8739170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kwon K.N., Chung W. Effects of private health insurance on medical expenditure and health service utilization in South Korea: a quantile regression analysis. BMC Health Serv Res. 2023;23(1):1219. doi: 10.1186/s12913-023-10251-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Park M., Ko SG. The effect of private health insurance coverage reduction on Korean medicine utilization. J Soc Prev Korean Med. 2015;19(3):57–66. [Google Scholar]
- 30.Park T.Y., Lee J.H., Moon S.J., Ko H.N., Ko Y.S., Song Y.S., et al. The descriptive analysis on the general characteristics of the inpatients with occupational injuries in an oriental medicine hospital. J Soc Prev Korean Med. 2010;14(3):117–127. [Google Scholar]
- 31.Park E.J., Cho MR. The clinical analysis on 40 patients of industrial accident. J Acupunct Res. 2003;20(3):75–85. [Google Scholar]
- 32.Kim M.K., Park K.S., Choi G.C., Yu J.E., Lee H.W., Kwon Y.S., et al. Long-term follow-up of inpatients with traumatic fractures who received integrative Korean medicine treatment: a retrospective analysis and questionnaire survey study. Medicine. 2023 Oct 13;102(41) doi: 10.1097/MD.0000000000034530. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Kim J.S., Lee G.B., Choi H.W., Kim D.Y., Kim D.J., Yoon D.J., et al. Integrative Korean medicine for pain and functional recovery in lumbar transverse process fractures: a case report and literature review. J Acupunct Res. 2025;42:46–52. doi: 10.13045/jar.24.0050. [DOI] [Google Scholar]
- 34.Kim T.R., Kim Y.J., Woo CH. 45 Cases report of Korean medicine rehabilitation after surgery. J Korean Med Rehab. 2017;27(1):83–89. [Google Scholar]
- 35.Lee G.H., Choi S.K., Jo J.H., Jin H.W., Jeon S.J., Lee J.H., et al. The clinical effects of applying an integrated rehabilitation protocol during the recovery phase for traumatic cervical spinal cord injury: a report of four patients. J Korean Med Rehab. 2024;34(3):107–114. doi: 10.18325/jkmr.2024.34.3.107. [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are openly available at: https://www.pswci.or.kr/.