Introduction

Over the past few decades, many countries have experienced rapid increases in healthcare costs, primarily driven by the overuse of medical services under the Fee-For-Service (FFS) model [1,2]. In response, many countries have gradually implemented the Diagnosis-Related Groups (DRG) payment system to control escalating healthcare expenses while improving hospital efficiency and service transparency [3]. As a form of case-based payment or activity-based funding, DRG groups patients based on their characteristics, allowing healthcare payers to provide fixed compensation to medical institutions and facilitating comparisons between them [4].

Developed countries such as those in Europe and the United States have accumulated substantial experience in the implementation of DRG. Some studies have concluded that the introduction of the DRG payment system may result in a range of anticipated and unanticipated changes in healthcare costs and service efficiency [5,6]. Theoretically, DRG systems are expected to enhance hospital efficiency by providing fixed payments that incentivize cost-effective practices, as hospitals have a strong financial motivation to reduce the use of medical resources and select the least costly inputs. Given these incentives, the intensity and duration of care for patients are anticipated to decrease, leading to shorter lengths of stay and potentially reducing overall healthcare costs. However, empirical evidence is mixed; while some studies report significant reductions in hospital costs [7–9], others have found that DRG reforms have not consistently resulted in cost savings [10]. Moreover, the reduction in the length of stay, although effective in controlling costs, has raised concerns about the quality of care, with some studies suggesting an increase in readmission rates as a consequence of shorter hospital stays [11].

As one of the fastest-growing developing countries, China’s healthcare costs have surged dramatically due to inefficient use of medical services under the FFS model [12]. In response, many Chinese cities have adopted healthcare reforms, with DRG serving as a representative model [13–16]. In 2018, following an initiative by the Meishan Municipal Government, the city was approved by the Sichuan Provincial Human Resources and Social Security Department as a pilot city for the DRG-based payment reform. In January 2019, 29 Grade II and Grade III hospitals were selected to participate in the DRG reform. The inclusion criteria for these hospitals were based on their capacity to meet the technical and administrative requirements necessary for DRG implementation, including adequate hospital scale, information infrastructure, coding ability, and overall management capacity. Primary care institutions were excluded at this stage due to limited capabilities. This government-imposed mandate effectively created an exogenous shock, as hospitals had no discretion in deciding whether to participate, thus providing a clear and abrupt policy change for evaluation.

As the DRG reform progressed, hospitals gradually adapted to the new payment system, refining their cost control strategies. Given the potential for these reforms to influence hospital efficiency and costs in both expected and unforeseen ways, we introduce a second period of DRG reform beginning in January 2021 to further evaluate the sustainability of the DRG system’s impact. However, it is important to clarify that this second period does not correspond to any formal policy change, but rather serves as a continuation of observation following the initial implementation.

Compared to countries in Europe and America, China’s DRG reform is relatively recent. Given the disparities in healthcare service capacity and resources, the reform experiences of developed countries may have limited applicability for guiding the implementation of DRG in developing countries.

To the best of our knowledge, this study represents the longest evaluation of the DRG system in Meishan, China, and serves as a valuable supplement to existing research. We provide a more comprehensive analysis of trends following the introduction of DRG, addressing the limitations of previous studies with shorter evaluation periods, and offering policy recommendations to optimize the DRG payment system.

Methods

Variables

Our study includes five outcome variables: (1) total hospital costs; (2) patient cost-sharing: In China’s medical insurance system, patient cost-sharing consists of three components: first, expenses outside the scope of basic medical insurance coverage; second, expenses within the coverage scope that require patient co-payment at a certain percentage; and third, out-of-pocket expenses, including deductibles and costs outside the coverage scope; (3) patient sharing ratio; (4) length of stay: used to measure hospital service efficiency; (5) 30-day readmission rate: used to assess changes in healthcare quality. To effectively mitigate the impact of price fluctuations on the research data, this study uses 2017 as the base year and adjusts the cost-related data from 2018 to 2022 to the same price level as 2017, based on the regional Consumer Price Index (CPI). Table 1 summarizes the measurement methods for the outcome variables and control variables.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Variables name and variable measurement.

https://doi.org/10.1371/journal.pone.0325041.t001

Empirical approach

To assess the impact of the DRG reform implemented in January 2019 and to examine the potential ongoing effects of the policy, we hypothesize the occurrence of a second policy shock in January 2021. Although this second shock is hypothetical and does not reflect any actual policy changes, it enables us to explore the potential long-term effects and further refine our understanding of the DRG reform’s impact. We use the Interrupted Time Series (ITS) model to analyze changes in various outcome variables before DRG reform (January 2017 to December 2018), during the first period of DRG reform (January 2019 to December 2020), and during the second period of DRG reform (January 2021 to December 2022). The model is specified as follows:

(1)

where Y_t represents the outcome variable for month t. T_t is the time elapsed since the beginning of the study (coded 1–72). DRG_1t and DRG_2t are dummy variables representing pre-intervention (coded 0) and post-intervention (coded 1) periods. T_tDRG_1t and T_tDRG_2t are interaction terms for time and the intervention. X_t is a vector of control variables at the year-month level, including age, gender, insurance type, CCI, hospital level, and hospital ownership. We use fourier transform to control for seasonality [17]. β₀ represents the baseline level of the outcome variable when T_t = 0, β₁ represents the potential trend before the DRG reform, β₂ represents the immediate change from the initial policy shock, β₃ the monthly change following the initial policy shock, β₄ represents the immediate change from the hypothesized second policy shock, and β₅ represents the monthly change from the hypothesized second policy shock.

Considering that the variables for total hospital costs, patient cost-sharing, patient sharing ratio, and length of stay exhibit skewed distributions, we employ a Generalized Linear Regression (GLM) model with a log link function for estimation [18]. For the count data (30-day readmission rate), we utilize a Negative Binomial Model (NBM) to account for overdispersion.

Meaningful changes in clinical practice under a new payment system take time to capture the delayed effects. Consequently, we conducted a lagged analysis of the initial policy shock, rather than examining immediate changes following the policy implementation. In addition, we performed a value-for-money estimation using the GLM without log link. Given that the DRG reform may affect patients across different age groups in distinct ways [19], we performed a stratified analysis by age. All statistical analyses were carried out using Stata 17.0, with a significance level set at α = 0.05.

Results

Table 2 presents the sample characteristics and outcome variables of our study, which include 1,617,608 inpatient discharge cases across the three periods.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Descriptive statistics of patient characteristics and outcome variables before, Phase I and Phase II implementation of DRG.

https://doi.org/10.1371/journal.pone.0325041.t002

We observed that after the DRG reform, the proportion of male patients increased, while the proportion of elderly patients aged 65 and above initially decreased but then increased. More than 80% of patients were covered by the resident insurance scheme and primarily sought care at public hospitals. Compared to the pre-reform period, total hospital costs increased during the first period after the DRG reform but decreased in the second period. Patient cost-sharing, patient sharing ratio, and length of stay all decreased significantly in both post-reform periods, while the 30-day readmission rate increased noticeably.

Table 3 summarizes the baseline trend, immediate changes and monthly changes of outcome variables. The Fig 1 presents changes in outcome variables during the three periods.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Interrupted time series (ITS) analyses for total hospital costs, patient cost-sharing, patient sharing ratio, length of stay and 30-day readmission rate of hospitalized patients before and after the DRG.

https://doi.org/10.1371/journal.pone.0325041.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Monthly trends in total hospital costs, patient cost-sharing, patient sharing ratio, length of stay and 30-day readmission rate.

Notes: The first dashed line represents the initial policy shock (implementation of the DRG payment reform in January 2019), while the second dashed line indicates the second policy shock (a hypothetical intervention assumed for further observation of outcome changes). The red trend line is derived from segmented regression predictions of the time series model. Outcomes were adjusted for gender, age, insurance type, Charlson Comorbidity Index, hospital level, hospital ownership and seasonality.

https://doi.org/10.1371/journal.pone.0325041.g001

The monthly trend of total hospital costs significantly increased by 0.39% (95% CI, 0.19%-0.59%) before the DRG reform. In the first period of DRG reform, we found no immediate change in total hospital costs (0.02%, 95% CI, -4.29% to 4.33%), but a monthly decrease of 1.23% (95% CI, 0.88%-1.59%). In the second period, total hospital costs immediately and significantly increased by 6.09% (95% CI, 1.63%-10.55%), with a monthly increase of 0.32% (95% CI, -0.04% to 0.68%), though the result was not statistically significant (Table 3 and Fig 1A).

Patient cost-sharing significantly increased by 0.28% per month (95% CI, 0.07%-0.49%) before the introduction of DRG. In the first period of DRG reform, the immediate change was not significant (2.61%, 95% CI, -1.90% to 7.12%), but a significant monthly decrease of 1.46% (95% CI, 1.09%-1.83%). In the second period, patient cost-sharing showed a significant immediate increase of 7.90% (95% CI, 3.23%-12.56%), but the monthly change was not statistically significant, with an increase of 0.22% (95% CI, -0.15% to 0.60%) (Table 3 and Fig 1B).

The patient sharing ratio decreased significantly by 0.11% per month (95% CI, 0.01%-0.20%) before DRG reform. In the first period of DRG reform, the immediate change was a significant increase of 2.58% (95% CI, 0.51%-4.66%), followed by a significant monthly decrease of 0.23% (95% CI, 0.06%-0.40%). In the second period, both the immediate change (1.78%, 95% CI, -0.36% to 3.92%) and the monthly change (0.10%, 95% CI, -0.27% to 0.08%) in patient sharing ratio were not statistically significant (Table 3 and Fig 1C).

Before the introduction of DRG, we observed no significant monthly change in length of stay (0.02%, 95% CI, -0.14% to 0.18%). In the first period of DRG reform, the immediate change was not significant (-2.0%, 95% CI, -5.45% to 1.44%), but a significant monthly decrease of 0.56% (95% CI, 0.27%-0.84%) was observed. In the second period, the immediate change increased significantly by 3.65% (95% CI, 0.09%-7.21%), while the monthly change (0.18%, 95% CI, -0.11% to 0.47%) was not statistically significant (Table 3 and Fig 1D).

We found that the 30-day readmission rate increased by 0.89% per month (95% CI, 0.43%-1.27%) before the introduction of DRG. In the first period of DRG reform, the immediate change increased significantly by -8.03% (95% CI, -15.57% to -0.48%), but the monthly change was not significant (-0.11%, 95% CI, -0.73% to 0.50%). In the second period, both immediate and monthly changes were decreased, but none of these were statistically significant (Table 3 and Fig 1E).

S1 and S2 Tables present the results of the lagged analysis and age-stratified analysis. We found that, with a 6-month lag, the immediate changes related to costs shifted from an increase to a decrease after the introduction of DRG. The age-stratified analysis revealed that in the first period after the introduction of DRG, the greatest monthly reductions in total hospital costs (1.43%, 95% CI, 1.01%-1.86%) and length of stay (0.83%, 95% CI, 0.52%-1.14%) were observed among patients aged 65 and above. In the second period, the largest monthly increases were observed in total hospital costs (0.63%, 95% CI, 0.27%-1.00%) and length of stay (0.48%, 95% CI, 0.21%-0.75%). S3 Table presents the results of a value-for-money estimation based on ITS analysis. Additionally, the number of inpatients per period, as reported in Table 2, can be used in conjunction with these monthly averages to estimate the overall financial impact of the policy.

Discussion

This study used ITS to assess the impact of DRG reform on medical costs, service efficiency, and healthcare quality. Currently, in the field of health policy evaluation, the Difference-In-Differences (DID) method seems to be favored by researchers [20]. DID is typically used to estimate the average treatment effect of an intervention and requires meeting the parallel trends assumption [21]. Therefore, ITS serves as a suitable and effective analytical framework in the absence of a control group. It not only constructs counterfactual scenarios by establishing potential trends through time series data [17], but also enables flexible adjustments for underlying trend factors, such as seasonal fluctuations, cyclical trends, and lagged effects, thereby improving the precision of intervention effect estimates [22].

This study found that the monthly trends of total hospital costs, patient cost-sharing, patient sharing ratio, and length of stay declined during the first period of DRG reform. In the second period, total hospital costs, patient cost-sharing, and length of stay showed a nonsignificant upward trend. The two periods after the introduction of DRG slowed the increase in the 30-day readmission rate, though this result was not statistically significant. These findings suggest that improvements in cost and efficiency indicators are associated with the DRG reform, while no significant improvements were observed in quality indicators.

Before the introduction of DRG, the main payment system for medical insurance was FFS, which led to excessive prescriptions and unnecessary tests [23,24]. The first period of DRG reform effectively reduced medical costs and length of stay, a finding consistent with conclusions from other pilot cities in China [25,26] as well as from studies in Europe and America [27–30]. The DRG reform may have stimulated changes in clinical practices, as the fixed payment structure created marginal incentives that heightened hospitals’ awareness of cost control and efficiency [30].

Reducing the burden on patients has always been an important goal of healthcare reform in China [31]. We found that, during the first period of DRG reform, both patient cost-sharing and the patient sharing ratio showed a significant decreasing trend, which is consistent with research findings in China [32,33]. Notably, we did not observe an increase in the patient sharing ratio, as total hospital costs and medical insurance fund expenditures declined more rapidly than patient cost-sharing. Of course, the substantial cost reduction was likely driven by the synergistic effects of other policies accompanying the DRG reform [34], such as the zero-markup policy for drugs and medical supplies.

We observed that during the second period of the DRG reform, there were no significant monthly changes in cost and efficiency indicators, although these indicators increased compared to the first period. We speculate that two factors may explain this: On the one hand, hospitals progressively adapted to the DRG payment system, transitioning from an initial focus on cost control to prioritizing the core goal of standardizing medical practices. As a result, the control of costs and efficiency under the DRG system may have stabilized, a conclusion supported by previous studies [35]. On the other hand, hospitals may have been concerned that excessive cost control would lead to a continuous reduction in DRG payment standards, prompting them to raise total hospital costs to mitigate this effect [36].

Our results indicate that while the introduction of DRG somewhat slowed the increase in 30-day readmission rates, the monthly change was not statistically significant, consistent with findings from France [37]. The impact of DRG on healthcare quality, particularly readmission rates, remains a contentious issue [11,33,34,38]. This suggests that changes in readmission rates may not be directly attributed to the implementation of a specific policy [39] but could instead be influenced by patients’ healthcare needs or socioeconomic status [37].

In the lag analysis, we found that the introduction of DRG resulted in an immediate reduction in cost following a lag of 6 months. This may be because hospitals need time to adapt to new medical practices [5]. As a new payment system, DRG necessitates systematic training for healthcare managers to enhance cost awareness. In the early period of the reform, this effect was primarily reflected in a slower rate of cost growth, a finding consistent with a study conducted in Beijing [40].

It is a matter of concern that DRG-based payment systems may introduce financial incentives that drive unintended hospital behaviors. We observed a reduction in the proportion of patients aged 65 and above and the most substantial monthly reductions in total hospital costs and length of stay during the first period of DRG implementation. Older adults tend to have higher rates of comorbidities and functional impairments, which typically lead to higher costs compared to other age groups, often exceeding the DRG payment standard. To address potential financial pressures on hospitals treating complex elderly patients, Meishan Healthcare Security Administration has introduced an additional payment policy to mitigate the unintended risks associated with DRG [41].

Besides, one potential concern is patient selection (“cream skimming”), where hospitals preferentially admit patients with lower severity conditions who require fewer resources but yield the same DRG reimbursement. Another common issue is upcoding, in which hospitals exaggerate patient severity by recording additional complications to qualify for higher DRG payment categories [42]. Additionally, hospitals might adopt early discharge strategies, reducing the length of stay to lower per-case costs, which could lead to an increase in readmission rates [11]. A further possibility is resource reallocation, where hospitals compensate for revenue losses by shifting patients to outpatient services, rehabilitation centers, or self-financed treatments [43]. While our study does not directly measure these strategic responses, further evidence is needed to verify the occurrence of unintended behaviors.

Conclusions

The introduction of DRG in the early period effectively reduced total hospital costs, patient cost-sharing, patient sharing ratio, and length of stay, but did not significantly improve the rising trend in 30-day readmission rates. As the reform continued, the impact of DRG on cost control and service efficiency stabilized. However, hospitals may have adopted certain unintended behaviors in response to DRG incentives, which require further research to fully validate. Our results remind policymakers to strengthen the regulation of clinical practices, improve the DRG payment system, and continuously monitor trends in healthcare quality.

Supporting information