The New Diagnosis-Related Group Reimbursement System and Laboratory Test Quality in Korea: Analysis of External Quality Assessment Results

Korea introduced a new diagnosis-related group (NDRG), which is a mixed-bundle reimbursement system. We evaluated the effects of NDRGs on laboratory test quality by analyzing data over three years (2016–2018) from the Korean Association of External Quality Assessment Service (KEQAS). A total of 42 NDRG-participating hospitals (CASE), 84 non-participating similar size-hospitals (CON-1), and 42 tertiary hospitals (CON-2) were included. We assumed the proportion of KEQAS results with a larger than 2 standard deviation index (SDI) to be a bad laboratory quality marker (BLQM). CASE BLQMs were lower than CON-1 BLQMs for more than 2 years in alkaline phosphatase (ALP), alanine aminotransferase (ALT), chloride, glucose, sodium, and total protein, and higher in creatinine. CASE BLQMs were higher than CON-2 BLQMs for more than 2 years in ALP, chloride, creatinine, glucose, lactate dehydrogenase (LDH), phosphorus, potassium, sodium, total calcium, total cholesterol, triglyceride, and uric acid. Mean SDIs for general chemistry tests were not significantly different depending on NDRG participation. However, the NDRG is currently a pilot program that compensates the amount of each institution’s reimbursement based on the fee-for-service system, and most participants were public hospitals. Thus, the effects of NDRGs on laboratory test quality should be re-evaluated after the NDRG program has stabilized and more private hospitals are participating.


Introduction
Increasing healthcare costs are a major concern in modern medicine. To reduce high medical costs and avoid unnecessary overtreatment, many countries have adopted diagnosis-related groups (DRGs) for reimbursements [1]. DRGs allow hospitals to receive fixed payment for the number and mix of patients they treat [2]. Accordingly, DRGs allow hospitals to reduce their costs per patient by cutting down the hospitalization period or intensity of services [3]. This generates conflicts of interest among physicians, mainly in terms of commitment to patient interests on the one hand and consideration of the hospital's economic interests on the other [4]. DRGs bear inherent risks of potential harm to patients by quality of health services to reduce the cost of healthcare. Previous studies showed that implementing DRGs decreased the duration of stay and out-of-pocket costs [5] as well as significantly reducing examination fees relative to fee-for-service (FFS) [6].
The compulsory national health insurance program of Korea was originally based on an FFS system [7]. After starting a pilot program in 1997 and the voluntary period starting in 2002, Korea implemented a mandatory DRG-based payment system for seven groups of specific operations/diseases in inpatient care on 1 July 2012 [8]. However, the DRG design was difficult to extend, due to the rigidity of the model [9]. Therefore, the new DRG (NDRG), a mixed model that combines the existing DRG system with FFS, began in April 2009, as a pilot program for 20 disease groups with three public hospitals. Under the NDRG system, basic services such as hospitalization and tests, general procedures, and drug treatments are grouped into DRG-based payment, but doctors' operations and procedures taken during hospitalization are not included into DRG-based payment and are separately reimbursed by the number of procedures. NDRGs expanded significantly in 2012 in terms of the number of disease groups and the number of participating hospitals; however, as it is still a pilot program, only 42 public hospitals covering 559 disease groups had participated up to 31 July 2018. As the system is still in the "introduction" period [9,10], there have been little data on the effects of NDRGs [11].
Laboratory tests are a large part of modern medicine. Nearly all inpatients, about half of patients in the emergency department, and nearly one-third of outpatients had laboratory results during their visit to the healthcare center [12], comprising a small portion (5-6%) of the total cost of healthcare [5,6]. Since poor laboratory results can significantly impact patients' care, monitoring and improving the quality of laboratory tests is crucial [13]. In order to improve the quality of testing, the clinical laboratory in Korea conducts quality control internally, voluntarily participates in the external quality assessment (EQA) program, and also voluntarily receives a laboratory accreditation through external on-site inspection by an independent third party every one or two years. Even though participation in EQA and laboratory accreditation were not mandatory, hospitals receive a certain percentage of quality incentives based on clinical pathologist recruiting, EQA scores, and laboratory accreditation scores while they are in the FFS reimbursement system [14]. However, this quality incentive is only indirectly reflected in the NDRG through an adjustment factor, so the effect of incentives is not very noticeable.
We hypothesized that the introduction of the NDRG system could negatively affect the quality of laboratory tests because the NDRG is a modified DRG. Although there was a recent report regarding significantly reduced examination fees after the introduction of DRG [6], there have been no reports about the quality of laboratory tests after the introduction of DRGs or NDRGs. Thus, in this study, we aimed to evaluate the effects of NDRGs on the quality of laboratory tests by analyzing EQA results as a quality marker.

Study Subjects
Medical institutions in Korea are classified as clinics, (small-to-medium) hospitals, long-term care hospitals, general hospitals, and tertiary hospitals according to the Medical Service Act [15]. Since the term "hospital" is also a generic noun that includes several types of hospitals, we decided to use the Korean Medical Service Act's term "hospitals" indicating "small-to-medium hospitals" to avoid ambiguity in this article. Hospitals should have more than 30 beds. General hospitals should have more than 100 beds with seven or more predefined medical departments (in case of general hospitals with 100-300 beds) or nine or more medical departments (in case of general hospitals with >300 beds) [15]. Tertiary hospitals should have at least 20 medical departments.
A total of 42 hospitals that participated in NDRGs before August 2018 were included in the study group (CASE). Additionally, 84 hospitals that did not participate in the NDRG system in July 2018 were selected for one control group (CON-1) by 1:2 matching based on number of hospital beds, and 42 tertiary hospitals not participating NDRGs were designated as another control group (CON-2) for reference comparison. The CASE group consisted of 35 public hospitals and seven private hospitals (33 general hospitals, eight small-to-medium hospitals, and one government institution). CON-1 consisted of all private hospitals (70 general hospitals and 14 small-to-medium hospitals). CON-2 consisted of 12 public hospitals and 30 private hospitals.

Data Source
The EQA results of the Korean Association of External Quality Assessment Service (KEQAS) database over three years (2016-2018) were analyzed for clinical laboratories, which were included as study subjects. EQA is a widely accepted method for evaluating clinical laboratories and enhancing their performance [16]. KEQAS was launched in 1976 and is the nation's top authorized EQA provider for standardization and quality management of the laboratory tests of domestic hospitals. KEQAS was accredited as an EQA provider by International Organization for Standardization/International Electrotechnical Commission (ISO/IEC) 17043 in 2015. A total of 1862 hospitals participated in KEQAS in 2018, which correspond to approximately 49.1% of hospitals (including small-to-medium hospitals, general hospitals, and tertiary hospitals) that submit health insurance claims for laboratory testing in Korea [17].
The proportion of results with a standard deviation index (SDI) more than 2.0 for each general chemistry test item was compared between CASE and control groups for each year. The SDI was calculated as the difference between laboratory result and the peer group mean divided by peer group standard deviation (SD) at each sample (SDI = (laboratory result−peer group mean peer group SD ). Results of more than 2 SDI (95% acceptable criteria) are generally regarded as poor results. Therefore, we assumed the proportion of laboratory results with larger than 2 SDI to be a bad laboratory quality marker (BLQM) for each test item. The peer group consisted of laboratories performing the analysis using the same methods, since it is expected that each method will have a similar matrix-related bias for a given specimen.
As a statistical parameter to assess bias in performance between CASE and control groups, we used the SDI of each laboratory's results. The geometric mean SDIs for each general chemistry test item between CASE and control groups were compared for each year.

Statistical Analysis
BLQM of CASE and control groups were compared by chi-square tests (CASE vs. CON-1 and CASE vs. CON-2). Mean SDIs among three groups (CASE and two control groups) were compared using ANOVAs, followed by post hoc analysis using Duncan's multiple range test to control the type I comparison-wise error rate. SDI values were reported as the geometric mean with 95% confidence intervals. p-values were based on two-sided comparisons, and p < 0.05 was considered statistically significant.

Ethics Statement
The study was approved by the Institutional Review Board of Konkuk University, in Korea (approval number: 7001355-201904-E-092).

General Characteristics of Study Subjects
The general characteristics of study subjects are shown in Table 1. Table 1. General characteristics of study subjects. Abbreviation: N, number; CON-1, control 1; CON-2, control 2; NDRG, new diagnosis-related group; LMF, Laboratory Medicine Foundation. CON-1 and CON-2 subjects were selected among the NDRG non-participating hospitals for the EQA result analysis period (2016-2018). CON-1 subjects were matched according to the number of beds in NDRG participating hospitals and CON-2 subjects were selected as the tertiary hospitals for reference comparison. † LMF accreditation is a laboratory accreditation program through external on-site inspection by LMF as an independent third party every one or two years. This shows the low proportion (40.1%) in CASE group and consistent high proportion in CON-1 (69.0%) and CON-2 (100%) groups.

Proportion of Results of More than 2 SDI (BLQM) in General Chemistry Tests
BLQMs differed among groups, according to test items or year. The BLQMs of CASE were lower than those of CON-1 for more than two years in ALP, ALT, chloride, glucose, sodium, and total protein, and higher for creatinine. The BLQMs of CASE were higher than those of CON-2 for more than two years in ALP, chloride, creatinine, glucose, LDH, phosphorus, potassium, sodium, total calcium, total cholesterol, triglyceride, and uric acid. There were no test items for which the BLQMs of CASE were lower than those of CON-2 ( Figure 1).

Differences in Mean SDIs for General Chemistry Tests
Mean SDIs of general chemistry tests differed among groups, according to test items and year. Mean SDIs in the CASE group were higher than those of CON-1 for ALP, total calcium, and uric acid, but lower for sodium. For total calcium and total protein, mean SDIs were higher in the CASE group than those for CON-2; however, ALP and ALT were lower than those for CON-2 ( Table 2).  CON-2, control 2. * p value by chi-square test was lower than 0.05 between CASE and CON-1. † p value by chi-square test was lower than 0.05 between CASE and CON-2. ALP: alkaline phosphatase, ALT: alanine aminotransferase; AST: aspartate transaminase; BUN: blood urea nitrogen; GGT: γ-glutamyl transferase; LDH: lactate dehydrogenase.

Differences in Mean SDIs for General Chemistry Tests
Mean SDIs of general chemistry tests differed among groups, according to test items and year. Mean SDIs in the CASE group were higher than those of CON-1 for ALP, total calcium, and uric acid, but lower for sodium. For total calcium and total protein, mean SDIs were higher in the CASE group than those for CON-2; however, ALP and ALT were lower than those for CON-2 ( Table 2).

Discussion
In this study, we evaluated the effects of pilot NDRGs on laboratory test quality by analyzing general chemistry test data from KEQAS from 2016 to 2018. To the best of our knowledge, this is the first report on this issue. There were some significant differences but in inconsistent directions; that is, some tests were higher quality in the CASE group, while other tests were higher quality in the CON-1 group.
By implementing DRGs, hospitals lowered laboratory operating costs by limiting laboratory growth and development [19], since costs became the biggest problem for hospital laboratories [20]. As a result, diagnostic tests requiring high operating costs such as complicated or new tests would be avoided [20], and efforts to maintain testing quality were predicted to decrease. However, in this study, we could not find any worsening in laboratory test quality, at least in terms of EQA data, after NDRGs were introduced. These findings should be very cautiously interpreted, however, because this does not necessarily mean that NDRG had no adverse effect on the laboratory test quality. The reimbursement scheme of the current pilot NDRG model is unique, as NDRG participating institutions were reimbursed commensurate with FFS cost, by the "Adjustment Factor" for each institution [10]. Therefore the current pilot NDRG did not encourage healthcare providers to cut unnecessary procedures/tests for cost containment [10].
Our findings could also be due to the characteristics of the study population. The majority of the CASE group were public hospitals. Public hospitals have many differences in purpose, operations, and patient demographics compared with private hospitals [11]. Public hospitals are less sensitive to profitability and tend to put more importance on the fulfillment of their mission than on efficiency of care [11]. The characteristics of public hospitals and private hospitals may have caused selection bias that impaired the study's intrinsic validity. However, that bias was impossible to avoid because there were no private hospitals that participated in NDRG during the evaluation period. In the future, it would be helpful to re-investigate this issue with private hospitals, which may be more sensitive to and dependent upon profits.
Further, EQA data may not be a sensitive enough indicator reflecting laboratory quality changes after the new payment system was introduced. Although EQA is widely used for monitoring medical laboratory performance [16], it reflects quality only at specific points in time, not the laboratory's overall quality. Additionally, quality improvement in Korea has already reached a fairly high level, due to decades-long efforts in EQA participation.
Last, the study did not fully control for environmental factors other than hospital size. For example, test quality may be positively affected by the laboratory accreditation program participation and ability of working clinical pathologists [18]. As shown in Figure 1, we found that the proportion of bad-quality test results were definitely lower in tertiary hospitals (CON-2) than in the CASE group. These findings may be due to the effects of laboratory accreditation. Jang et al. [18] reported that hospitals with laboratory accreditation showed significantly better results over four years for all general chemistry tests (p < 0.0001) compared to hospitals without accreditation. The researchers stressed the need to establish a laboratory accreditation program to standardize diagnostic testing. Further research on the effects of NDRGs is required, using various metrics that can measure overall quality of laboratory services.
Laboratory Medicine Foundation (LMF) accreditation participation shows a low proportion (40.1%) in the CASE group and consistently high proportions in the CON-1 (69.0%) and CON-2 (100%) groups (Table 1). Laboratory accreditation is a good tool for monitoring and improving the total quality of a laboratory. Participation in accreditation is burdensome, however, because it requires time, effort, and money. Therefore, regulations or benefits should encourage participation. In Korea, participants found some incentives ("Comprehensive Verification Fee") under the FFS system [21] but cannot receive them under NDRGs. Perhaps that is why the participation rate in the CASE group is low.
Laboratory participation in quality assurance systems, such as laboratory accreditation and EQA, may contribute to reducing false-positive or -negative results and unnecessary test requests, thus ensuring patient safety and improving cost-effectiveness [22]. Laboratory quality improvement must be considered in any reimbursement system, and this should be true also in the current and future NDRG model. We suggest that adequate quality indicators be developed and included in NDRG reimbursement schemes to achieve this critical aim.

Conclusions
We found some significant differences in the quality of tests between CASE and CON groups, but in inconsistent directions; some tests were higher quality in the CASE group, while other tests were higher quality in the CON-1 group. However, these findings should be interpreted cautiously since the current NDRG program is a pilot program with compensations commensurate to the FFS, and most participants were public hospitals. The effects of NDRGs on laboratory test quality should be re-evaluated after the formal NDRG program is implemented with more private hospitals. We suggest adequate quality indicators be developed and included in the NDRG for laboratory quality improvement.