An Evaluation of the Patient Clinical Complexity Level (PCCL) Method for the Complexity Adjustment in the Korean Diagnosis-Related Groups (KDRG)


 ObjectiveTo evaluate the performance of the Patient Clinical Complexity Level (PCCL) mechanism, which is the patient level complexity adjustment factor within the Korean Diagnosis-Related Groups (KDRG) patient classification system, for explaining the variation of resource consumption within Age Adjacent Diagnosis-related groups (AADRGs).MethodsWe used the inpatient claims data from a public hospital in Korea from January 1, 2017 to June 30, 2019, with 18,846 claims and 138 Age Adjacent Diagnosis-related groups (AADRGs). The differences in the total average payment between the four PCCL levels for each AADRG was tested using ANOVA and Duncan’s post-hoc test. The three patterns of the differences with R-squared were: the PCCL reflected the complexity well (Valid); the average payment of PCCL 2, 3, 4 was greater than PCCL 0 (Partially Valid); the PCCL did not reflect the complexity (Not Valid).ResultsThere were 9 (6.52%), 26 (18.84%), and 103 (74.64%) ADRGs included in VALID, PARTIALLY VALID and NOT VALID, respectively. The average R-squared in VALID, PARTIALLY VALID, and NOT VALID was 32.18%, 40.81%, and 35.41% respectively, with the average R-squared for all patterns of 36.21%.ConclusionsAdjusting using PCCL in the KDRG classification system exhibited low performance to explain the variation of resource consumption within Age Adjacent Diagnosis-related groups (AADRGs). As the KDRG classification system is used for reimbursement under the New DRG-based PPS pilot project with plans for expansion, there should be an overall review of the validity of the complexity and rationality of using the KDRG classification system.


Healthcare system and payment method in Korea
In Korea, more than 90% of hospitals are privately owned (12) with more complex patient case-mix in private hospitals than public hospitals. The main method for payment is the fee-for-service model with no separate payments between hospitals and doctors by the National Health Insurance System.
There are two types of DRG payment systems for patient classi cation in Korea that roots from the same patient classi cation system (13,14): (1) the mandatory DRG-based Prospective Payment System (PPS) for seven diseases, and (2) the new DRG-based PPS for public hospitals. The mandatory DRG-based PPS, including payments for both hospitals and doctors, targets seven relatively simple surgical disease groups and was introduced rst in July 2012 to certain clinics and hospitals. It has since been extended to all medical institutions since July 2013. Under the pilot project, the New DRG-based PPS targeted public hospitals with doctors' procedures, expensive therapeutic materials, and some expensive drugs paid separately by fee-for-service payment in the system. Since 2018, the New DRG-based PPS has been extended to private hospitals through voluntary participation.

Overview of the complexity adjustment of the Korean Diagnosis-Related Groups
The mechanisms for re ecting the complexity of the Korean Diagnosis-Related Groups (KDRG) system, which was developed based on the United States Re ned DRG (US RDRG) and the Australian Re ned DRG (AR-DRG), are as follows (15,16); (1) The patient's complications and comorbidities (CC) are assigned a severity score based on the CC list, (2) If there are multiple secondary diagnoses, remove the duplicates by applying an exclusion list, and (3) adjust the severity of the disease group and re ne the disease group using the patient clinical complexity level (PCCL) calculation formula that calculates the cumulative effects of the multiple diagnoses. The PCCL was designed to prevent similar diseases being calculated more than once and intended to re ect the cumulative effect of patient's comorbidities (17,18). The PCCL value is calculated per patient episode, and Re ned DRG (RDRG) per Age Adjacent DRG (AADRG) is determined by considering the statistical criteria and the minimum number of counts (15,16 (19). However, we were not able to evaluate the accuracy of payment at the patient level. As the New DRG-based PPS extends to private hospitals that have more complex patients than public hospitals in Korea, ensuring accurate payment at the patient-level is of an importance. As such, this follow-up study aims to validate the accuracy of payment at the patient-level using the New DRG-based PPS in a public hospital.

Statistical analyses
The general characteristics of the data were reported as mean ± SD or as proportions for gender, age, types of insurance, length of stay, and payment amount. We also showed data characteristics according to Major Diagnostic Category (MDC) in KDRG.
Only the AADRGs that has adequate sample sizes were selected to report. We used Gpower 3.1 (20) to calculate the minimum sample sizes per AADRGs. The alpha was set to 0.05 and the power to 0.8. Effect size was estimated from standard deviation within each group of each AADRG, the sample size, and mean of log-transformed payment amount from the actual data. For example, in AADRG I6821 where the number of groups = 4, SD within each group = 0.2656, the average log-transformed payment amount of 6.32068, 6.45057, 6.7249 and 7.03847, and sample size of 97, 20, 7, 3, respectively. The estimated effect size was 0.5360507 and the minimum sample size was 44. AADRG I6821 was selected to report because the actual sample size of was 127.
To evaluate the performance of the PCCL scores to explain the complexity of the patient, we performed a one-way Analysis of Variance (ANOVA) and Duncan's post-hoc test using PCCL scores as an independent variable and the log-transformed payment amount as the dependent variable (Supplement 1: The diagram of analysis method). The R 2 value of the ANOVA was presented for the explanatory power of the PCCL on the payment amount.

Pattern analysis
Based on the same criteria as our previous research, we categorized the results of the Duncan's post-hoc test by AADRGs into three different validity patterns: Valid, Partially valid, and Not valid (Supplement 2: Criteria used to classify validity patterns).
The VALID pattern included the AADRGs, in which the average payment amount increased signi cantly along with increase in the PCCL scores. B6623 in Supplement 3 is a good example. For the PARTIALLY VALID pattern, the average payment amount of PCCL 0 was signi cantly less than the lowest average payment amount of other PCCLs. Duncan's post-hoc test for the payment amount of E7202 in Supplement 3 showed that the average payment amount of PCCL 3 and PCCL 4 were not statistically different from that of PCCL 2, but different from that of PCCL 0. We considered them inappropriate, but better than NOT VALID. In the NOT VALID pattern, the average payment amount of PCCL 0 is statistically equal to or greater than the average payment amount of other PCCLs. J6002 in Supplement 3 showed that the average payment amount of PCCL 0 is statistically same as that of PCCL 2.

General characteristics
The number of AADRGs and inpatient claims in the raw and analysis data at the MDC level is shown in Table 1. Of the 532 AADRGs, 138 (25.94%) AADRGs were included for analysis in 18,846 (70.36%) claims.

The validity pattern analysis
The summary of the validity pattern analysis evaluated for the validity of the PCCL scores is shown in Table 2. The average payment amount increased signi cantly with increase in the four PCCL scores (0, 2, 3, 4) or had a 'VALID' pattern in nine AADRGs (6.52%). There were 26 AADRGs (18.84%) that were 'PARTIALLY VALID' or had average payment amount of PCCL 0 that was signi cantly less than the lowest average amount of other PCCL scores and the 103 AADRGs as 'NOT VALID' (74.64%) that did not re ect the complexity between average payment and PCCL score or not valid suggesting that average amount of PCCL 0 was not signi cantly different from those of other PCCLs.

Discussion
This is the rst study to evaluate the mechanism of patient level complexity adjustment in KDRG. Our results showed that using PCCL for the new DRG-based PPS exhibited low performance. A study conducted in Australia reported a newly developed complexity adjustment mechanism, since the existing PCCL measure developed using limited data on length of stay had not been revised since its rst introduction of usage (18). Similarly to our study, this study also reported poor performance using PCCL complexity adjustment on their hospital cost data.
Low performance of PCCL adjustment in determining average payment using the KDRG may potentially be due to various factors used to calculate PCCLs, such as the CC list, CCLs (15) and CC exclusion list (21), which have not been updated since the introduction of such things, as stated in the our previous study (19). Another reason for the poor performance of PCCL adjustment may be due to inaccuracy of secondary diagnoses coding (22). The current coding guideline used in Korea is based on other countries for statistical purpose to nd out the prevalence and mortality of the disease and not for DRG-based payments (23). It is currently revised and issued by National Statistical O ce under the Ministry of Economy and Finance, not by the Ministry of Health and Welfare. This administrative structure makes it di cult to re ect clinical reality in various healthcare elds in the guideline.

Limitations
There are limitations to this study. The results of this study are not generalizable to total patient population paid for New DRG-based PPS, since the inpatient claims were derived from a single medical institution, which was a general hospital and one of the reference institutions based on calculating the base DRG fee for the New DRG-based PPS pilot project. Our research showed poor performance of complexity adjustment mechanism in the KDRG system, despite the hospital conducted this research has a greater proportion of patients with more common and moderate complexity diseases than tertiary hospitals. This suggests that the performance may be worse in hospitals with more complex patient case-mix.
Furthermore, not all of the AADRGs were evaluated because we were limited to the number of DRGs found in our inpatient claims database. Lastly, we assessed the validity of using the PCCL adjustment with the KDRG system on the medical charges and not the cost. The charge for fee-for-service has set up including payments for hospital and doctor under government control and used as a proxy to identify resource consumption in Korea. Furthermore, due to the small sample sizes included per AADRGs, we may have overestimated AADRGs with NOT VALID pattern analysis. However, we calculated the appropriate size of data by AADRGs using Gpower ensuring statistical power.

Signi cance
In most countries, DRG is mainly used as a basis for budget allocation (24). In Korea, however, predetermined DRG fee for each disease group is used to directly pay health care providers for their services. As of 2020, there are 37 private hospitals participating in the new DRG-based PPS pilot project with the government providing up to 30% policy participation incentives to hospitals. By 2022, however, participation incentives for the new DRG-based PPS is expected to decrease. Thus, hospitals will be reimbursed for inpatient services solely on the DRG-speci c fees calculated based on the cost currently being collected by government. The most accurate and appropriate compensation using the new DRGbased PPS can be determined with stable patient classi cation system and a reasonable complexity adjustment mechanism. Experts argue for quickly replacing the fee-for-service system with the New DRGbased PPS to stabilize the rapid increase of national medical expenditure and increase health insurance coverage (25). With increasing participation of private hospitals in the New DRG-based PPS pilot project and expansion of the new DRG-based PPS to 200 medical institutions by 2022, there is importance in ensuring payment accuracy using the new DRG-based PPS (26).

Conclusion
Poor performance of PCCLs, a mechanism for the patient-level complexity adjustment, in the KDRG system suggest that there should be an overall review of the validity and rationality of using the PCCLs in the KDRG classi cation system for reimbursement. According to hospitals that have participated in the pilot project, the hospitals have negative revenue after excluding participation incentives. This can be interpreted as that the compensation for the provision of medical services is not covered the medical charges calculated using the PCCL adjustment in the KDRG classi cation system, but rather by the participation incentive money that is to be discontinued in the near future.
Although changes in the payment mechanism for providers is inevitable, but stabilization and rationality of the system's components must be ensured, as the payment system is a factor that can affect the providers, insurers and ultimately the patients. Therefore, when designing systems and implementing policies, policy makers should take a more cautious approach considering their long-term impact.

Declarations
Ethics approval and consent to participate Not applicable Consent for publication Not applicable Availability of data and materials The datasets generated during and/or analysed during the current study are not publicly available due the characteristic of data owned by the medical institution but are available from the corresponding author on reasonable request.
SJ is the rst author of the paper, reviewed related papers, analyzed the data, and wrote most part of the paper. BY and KH contributed to the data management, data analysis and interpretation of the results. SM reviewed and gave helpful comments on English version of the paper. SI directed the overall study and is the guarantor for the study. All authors read and approved the nal manuscript.