An empirical study on diagnosis related group of hospitalization costs for lumbar disc herniation patients in Chinese medicine hospitals, China

Mengen Chen, Haojia Hou, Youshu Yuan, Xiaoxi Zhang, Hao Yan, Pu Ge, Yushi Che, Jingyu Yang, Zhiwei WangResearch Square 2024 · DOI

Abstract

Background: Lumbar disc herniation (LDH) is a frequent and multiple disease that imposes a heavier global economic burden. As China's diagnosis related group (DRG) payment system for Chinese medicine health insurance has entered the initial stage, this study explored the factors influencing the hospitalization costs and established a DRG case classification scheme for LDH patients in Chinese medicine hospitals.

Methods: A total of 2,857 LDH patients were hospitalized in Qingyang City Hospital of Chinese Medicine from January 2017 to June 2022. Univariate analysis and a multiple linear regression model were used to determine the primary factors influencing hospitalization costs. These factors were then utilized as classification node variables in an E-CHAID decision tree model to categorize LDH patients into DRGs. The outcomes were evaluated, and standard costs were computed for each group.

Results: Hospitalization costs were mainly affected by age, admission pathways, diagnosis and treatment based on Chinese medicine evidence, complications and comorbidities, surgery and operation, and length of stay (LOS), with LOS being the most crucial factor. The decision tree model resulted in 10 DRGs. The intra-group coefficient of variation (CV) ranged from 0.23 to 0.91, and the inter-group reduction in variation (RIV) was 0.48.

Conclusions: The E-CHAID decision tree model grouping results are robust, showing good homogeneity within each DRG and notable heterogeneity between DRGs. These findings provide valuable insights to advance reforms in Chinese medicine DRG-based healthcare payment methods.

Introduction

As China transitions its healthcare system toward a Diagnosis Related Group (DRG) payment model—where hospitals are paid a fixed amount based on the patient's diagnosis rather than individual services—Traditional Chinese Medicine (TCM) faces a unique challenge. Standard Western DRG models often fail to account for the specific diagnostic techniques, herbal treatments, and non-surgical therapies central to TCM. Lumbar Disc Herniation (LDH) is a high-prevalence condition that imposes a massive economic burden globally. This paper seeks to bridge the gap by creating a TCM-specific DRG framework for LDH, ensuring that payments are fair and costs are controlled.

Methods

The study utilized a dataset of 2,857 LDH patient records from the Qingyang City Hospital of Chinese Medicine, spanning January 2017 to June 2022. To account for inflation, all costs were adjusted using the Gansu Province Healthcare Consumer Price Index (CPI) with 2016 as the base year. The researchers employed a three-stage statistical pipeline:

  1. Univariate Analysis: Mann-Whitney U and Kruskal-Wallis H tests identified patient characteristics significantly correlated with costs.
  2. Multiple Linear Regression: Significant variables were fed into a regression model to determine the primary drivers of expenditure.
  3. E-CHAID Decision Tree: Using the primary factors as nodes, an Exhaustive Chi-squared Automatic Interaction Detection (E-CHAID) model branched patients into 10 homogenous groups.

Results

Primary Cost Drivers

Multiple linear regression identified several key variables influencing hospitalization costs. Length of stay (LOS) emerged as the single most influential variable (Beta = 0.692). Other significant factors included age, admission pathways, and whether the patient underwent surgery. Interestingly, patients receiving "diagnosis and treatment based on TCM evidence" had higher costs, likely due to more targeted and individualized care plans.

Multiple linear regression analysis of factors influencing hospitalization costs.

DRG Classification and Validation

The E-CHAID model successfully categorized patients into 10 distinct DRGs. The quality of these groups was validated using the Coefficient of Variation (CV) for internal consistency and Reduction in Variation (RIV) for external heterogeneity. All CV values were below 1.0 (ranging from 0.23 to 0.91), indicating high internal homogeneity. The RIV value was 0.48, exceeding the 0.40 threshold required for a valid classification model.

Classification results and statistical evaluation of the 10 established DRGs.

The study established standard costs for each DRG, ranging from approximately 3,091 RMB to 31,930 RMB, providing a benchmark for hospitals to identify and investigate "outlier" cases of overspending.

Discussion and Conclusion

The findings suggest that the most effective way for TCM hospitals to remain sustainable under DRG systems is to optimize rehabilitation timelines, as LOS is the pivotal factor in cost. The E-CHAID decision tree model provides a robust framework for grouping patients, showing good homogeneity within groups and notable heterogeneity between them. While limited by its single-center scope, this study offers a scientific basis for formulating standard costs and advancing TCM healthcare payment reforms across China.

Summary updated

Abstract

<title>Abstract</title> Background Lumbar disc herniation (LDH) is a frequent and multiple disease that imposes a heavier global economic burden. As China's diagnosis related group (DRG) payment system for Chinese medicine health insurance has entered the initial stage. Our study explored the factors influencing the hospitalization costs and established DRG cases classification scheme of LDH patients in Chinese medicine hospitals. Methods A total of 2,857 LDH patients were hospitalized in Qingyang City Hospital of Chinese Medicine from January 2017 to June 2022. Using univariate analysis and multiple linear regression model to determine the primary factors influencing hospitalization costs for LDH patients. These factors were then utilized as classification node variables in a decision tree model to categorize LDH patients into diagnosis related groups (DRGs). The outcomes of these groupings were evaluated, and standard costs were computed for each DRG. Results Hospitalization costs of patients with LDH were mainly affected by age, admission pathways, diagnosis and treatment based on Chinese medicine evidence, complications and comorbidities, surgery and operation, and length of stay, with the latter being the most crucial one. The decision tree model resulted in 10 DRGs with corresponding standard costs. The intra-group evaluation metric coefficient of variation (CV) was appropriate (required to be less than 1, actual range:0.23–0.91), and the inter-group evaluation metric reduction in variation (RIV) was suitable (needed to be greater than 0.4, actual value is 0.48). Significant variations exist in disease related weight (RW) and standard costs across all groups. Conclusions The E-CHAID decision tree model grouping results are reasonably robust, showing good homogeneity within each DRG and notable heterogeneity between DRGs. Based on these groupings, scientifically formulate standard costs for each DRG and set upper limits, providing valuable insights to advance reforms in Chinese medicine DRG-based healthcare payment methods.
BibTeX
@inproceedings{chen2024empirical,
  author = {Mengen Chen and Haojia Hou and Youshu Yuan and Xiaoxi Zhang and Hao Yan and Pu Ge and Yushi Che and Jingyu Yang and Zhiwei Wang},
  title = {{An empirical study on diagnosis related group of hospitalization costs for lumbar disc herniation patients in Chinese medicine hospitals, China}},
  booktitle = {Research Square},
  year = {2024},
  doi = {10.21203/rs.3.rs-4961494/v1},
}
Figure