Materials and methods
2.1 data source
This study employed a natural experiment design with a nationwide, populations-based dataset to examine the impacts of the Taiwan version of the DRG payment scheme. In 2009, 486 hospitals were contracted with the BNHI, accounting for more than 95% of all hospitals in Taiwan, and more than 99% of Taiwan residents were enrolled in the compulsory health insurance program. The analysis presented in this study is based on the nationwide NHI claims dataset from 2009 to 2010 in Taiwan.
2.2 DRG payment
Beginning with its inauguration on March 1, 1995, the BNHI implemented a case-payment scheme, which was a simplified version of a prospective payment system similar to the DRG in the United States. The case-payment scheme reimbursed a predetermined, fixed amount for each patient admitted into a hospital for certain surgical procedures, such as appendectomy or Caesarean section. However, this scheme did not adjust for age, sex, or discharge status. Patients underwent those procedures would be reimbursed by a fixed amount of reimbursement; however, a certain proportion of patients with complication/co morbidity or consumed higher resources during a hospital stay could be reimbursed by fee-for-services schemes via application. The total number of surgical procedures under the case-payment scheme increased between 1995 and 2009 from 3 to 54
The BNHI developed the Taiwan version of DRG payments scheme based on the 18th version of DRG provided by center for medicare and Medicaid services (former health care financing administration). It reimburses a fixed rate per admission while accounting for age and sex, discharge status, and the presence of complications or co morbidities. Currently, the Taiwan version of DRG consists of 1029 codes and excludes several categories, such as psychiatric disorders, cancer treatments etc. The DRG payment scheme was introduced gradually by various groups beginning in 2010 and is scheduled to be completed in 2014. In order to minimize the impact of DRG payment, in 2010 the BNHI started with the 164 categories of DRGs which were derived and sub-categorized from the original 54 categories in the case-payment scheme. Inpatient services that will adopt the DRG payment in the second year (2011) or later were reimbursed by fee-for-services schemes in 2010 as usual.
2.3 The participants
Using the nationwide NHI claims data, we identified subjects who had a major diagnosis listed in the cardiovascular surgical DRG category (Major diagnostic classification [MDC] 5, disease and disorders of the circulatory system). A total of 67 surgical DRG codes were listed in MDC 5. Forty-six surgical DRG codes were implemented in 2010, and the other DRG categories were to be implemented the following year. In order to increase the homogeneity of the study subjects, we included only those patients underwent coronary artery bypass graft surgery (CABG) or percutaneous transluminal coronary angioplasty (PTCA) in the analysis. These patients accounted for 60% of patients under surgical DRG categories in MDC 5 and formed the intervention group. Patient who underwent a cardiovascular-related procedure that was not paid by a DRG scheme (which was to be included in the second year) constituted the comparison group. Because the DRG payment system was launched on Januay 1, 2010, we defined “January 1, 2009 to December 31, 2009” as the pre-DGR period and “January 1, 2010 to December 31, 2010” as the post-DRG period.
This study aimed to examine the impacts of DRG payment on the changes in health care providers’ behavior. We assumed that health care providers would adopt different treatment strategies for patients in the DRG or in the comparison group. To minimize the potential influence of selection bias, we used two strategies in sample selection: selecting patients from the same group of surgeons and using PSM approaches. First , to enhance the comparability between the two groups, we selected patients treated by the same group of surgeons and examined these surgeons’ behavior pre-and post the implementation of the DRG payment. Subjects treated by surgeons who had performed relatively few cardiovascular-related procedures (fewer than 3 patients per year) in the two study groups before or after the implementation of the DRG payments were excluded. This restriction excluded only a small portion of subjects from the analysis. The intervention group included 485 surgeons with 10,892 patients in the post-DRG period. The comparison group included 28,436 patients in the pre-DRG period and 29,035 patients in the post-DRG period.
Because the subjects in the pre- and post-DRG periods were not the same group of patients, we employed PSM process to increase the homogeneity of the subjects before and after the DRG payments [41]. We performed two separate PSM process using logistic regression models for two sets of patients, one for selecting the pre- vs. post-DRG subjects in the intervention