Medical insurance payment system reform is one of the crucial measures for controlling irrational medical practices (1). The payment systems can be categorized into retrospective and prospective payment systems. The fee-for-service (FFS) payment method is the most traditional retrospective payment method (2). However, it is prone to moral risks such as overtreatment due to information asymmetry and provider-induced demand (3–5).

Prospective payment systems encompass capitation, per-diem, and diagnosis-related group (DRG) payment systems (6–8). Among these, the DRG payment is a globally recognized approach that ensures the quality of medical care while facilitating better cost control (9, 10). Originating in the United States, the DRG payment groups patients into different diagnostic groups based on age, sex, number of days in the hospital, clinical diagnosis, severity of illness, comorbidities, and complications. The medical insurance department calculates the standard payment for each DRG and reimburses hospitals accordingly (11, 12). Compared to FFS, DRG payment contributes more effectively to reducing unreasonable medical expenditures, improving the efficiency of medical insurance fund utilization, and enhancing the management capabilities of hospitals. However, this approach also carries the risk of patient disincentives and reduced quality of care (13, 14).

Since the establishment of the National Healthcare Security Administration (NHSA) in 2018, China’s medical insurance fund has grown steadily in both scale and function, as illustrated in Figure 1. Between 2018 and 2024, medical insurance fund revenue increased from 2.11 trillion RMB to 3.48 trillion RMB, representing a CAGR of 8.71%. Over the same period, medical insurance fund expenditure rose from 1.78 trillion RMB to 2.97 trillion RMB, with a slightly higher CAGR of 8.95%. While both revenue and expenditure have expanded significantly, the persistent faster growth of expenditure—evident in specific years such as 2020, 2022, and 2023—signals escalating fiscal pressure. This pressure is further compounded by the increase in total health expenditure, which rose from 5.91 trillion RMB in 2018 to 9.06 trillion RMB in 2023, with a CAGR of 8.99%, reflecting strong and sustained demand for health financing. Facing growing fiscal pressure, the NHSA prioritized payment system reform as a critical strategy to standardize medical practices and, on that basis, reduce unnecessary expenditures and strengthen the sustainability of the healthcare financing system. In 2021, the NHSA issued the Three-Year Action Plan for DRG/DIP Payment Reform, aiming to extend DRG or DIP-based payment models to all eligible inpatient care institutions by the end of 2025.

Diagnosis-Intervention Packet (DIP) is a prospective payment system independently developed by the NHSA, drawing inspiration from the DRG payment system. The DIP payment system utilizes mathematical models to analyze and categorize disease data, diagnostic procedures, and treatment methods comprehensively, enabling efficient management and reimbursement grouping (15). Its approach, features, and payment model are similar to those of the DRG payment. However, a key distinction lies in their grouping methodologies: while the DRG payment relies on clinical experience as its foundation through expert judgment and selection of clinical pathways for “one group of multiple diseases” or “one group of multiple operations,” the DIP payment emphasizes statistical analysis of objective real-world data by exhaustively clustering disease diagnoses and surgical procedures from historical cases into groups characterized by “one group of one operation for one disease” (16–18).

As China deepens its healthcare reforms, especially in payment system restructuring, evaluating the effectiveness of newly implemented mechanisms such as DIP is both timely and policy-relevant. Guangzhou, the capital city of Guangdong Province, is renowned for its robust economy with the fourth-largest gross domestic product (GDP) in China. Importantly, it was also among the first to implement the DIP payment system, launching the reform in January 2018. This study aims to evaluate whether the DIP payment reform has achieved its intended effects on inpatient medical service delivery. Specifically, it examines the reform’s impact on three key dimensions: (1) the average inpatient cost per admission, (2) the length of hospital stay (LOS) and 30-day readmission rate, and (3) patients’ out-of-pocket (OOP) payments. While the primary objective is to assess the effectiveness of the DIP model in controlling medical costs and alleviating patient financial burden, the study also extends its inquiry to the internal logic and structural design of the DIP system—particularly in the event that the expected policy outcomes are not realized. By integrating empirical evaluation with a mechanism-based analysis, this research aims to offer a more comprehensive understanding of how payment system design shapes implementation outcomes and informs future policy refinement.

The DIP payment system in China operates through four sequential stages, as illustrated in Figure 2.

Stage 1 involves disease grouping and relative weight (RW) calculation. The medical insurance authority utilizes historical inpatient data from the previous three years to formulate DIP groups based on primary diagnoses (ICD-10), procedures (ICD-9-CM-3), and case volume thresholds. Each group is then assigned a RW, calculated as the weighted average of inpatient costs across these three years (using a 1:2:7 weighting for years T − 3, T − 2, and T − 1, respectively).

Stage 2 covers inpatient admission and RW assignment. During the current year, newly admitted patients are mapped to pre-defined DIP groups based on their principal diagnosis and procedures at discharge, coded using ICD-10 and ICD-9-CM-3. If a patient’s inpatient cost falls within the defined cost range for the group (e.g., 0.5 to 2 times the historical weighted average in Guangzhou), the standard RW is applied; otherwise, the RW is adjusted according to local outlier policies. RWᵢ represents the sum of all individual case RWs for patients treated at hospital i.

Although medical insurance funds are typically disbursed on a monthly basis according to each hospital’s accumulated RWᵢ from the preceding month, the final reimbursement is determined through a year-end adjustment that accounts for the hospital’s total service provision over the entire fiscal year. Therefore, this study focuses on the year-end settlement process and does not address monthly payment mechanisms. In this context, Stage 3 involves the determination of the year-end point rate. The regional medical insurance budget is first divided by RRₜ (the proportion of total inpatient costs reimbursed by the insurance fund, calculated as total inpatient cost minus total OOP payments divided by total inpatient costs), resulting in the total distributable expenditure. This amount is then divided by RWₜ (the sum of RWᵢ across all designated hospitals) to derive the year-end point rate used for final settlement.

Stage 4 covers the calculation of each hospital’s income under the DIP system. A hospital’s payment standard (PSᵢ) is calculated by multiplying its RWᵢ by the year-end point rate. Since PSᵢ is calculated based on the total expenditure benchmark rather than the insurance-funded benchmark, the medical insurance authority deducts the non-reimbursable portion—namely, the OOP payments—from PSᵢ when determining the amount actually payable by the fund.

In China, two methods are commonly applied to implement this deduction: the subtraction method, which subtracts the actual OOP amount from PSᵢ, and the multiplication method, which applies the hospital’s reimbursement ratio to PSᵢ to estimate the reimbursable share. This amount represents the portion covered by the insurance fund. This insurance-covered amount, when combined with the OOP payments collected directly from patients, constitutes the hospital’s total income under the DIP system. If this total income exceeds the hospital’s total inpatient costs for all discharged cases, the institution is deemed to have operated at a “surplus” under the DIP system; conversely, if the income falls short, the hospital is considered to have incurred a “deficit.” Through this surplus-deficit mechanism, the medical insurance authorities aim to incentivize cost containment and promote reductions in overall inpatient costs.

Figure 3 illustrates the financial flow under the DIP payment system. Initially, the healthcare security administration sets the regional global budget and defines the DIP group classification and corresponding base RWs. Upon discharge, patients pay OOP expenses under the FFS scheme. Hospitals then submit diagnostic and procedural codes, which are used to assign each case to a DIP group and determine its RW. Throughout the year, each hospital accumulates its RW_i by summing the RWs of all inpatient cases. At the end of the year, the administration calculates the Year-End Point Rate and the PS_i for each hospital—steps detailed in Figure 2. The final insurance payment amount is then computed using either a subtraction or multiplication method, as specified by regional policy design.

We employed a set of outcome indicators categorized into three analytical dimensions to evaluate the impact of the DIP payment reform. First, in the dimension of inpatient medical expenditures, we examined the inpatient cost per case and further decomposed it into three components—medication expenditures, medical consumables expenditures, and diagnostic and therapeutic service expenditures—to capture changes in expenditure composition. Second, in the dimension of patient financial burden, we examined OOP payments per case, representing the share of hospitalization costs directly borne by patients (see Figure 4 for the structural relationship between cost components). Third, in the dimension of hospital service efficiency, we included LOS, defined as the number of inpatient days per admission, as an indicator of resource utilization. Additionally, we included the 30-day readmission rate to reflect short-term quality of care and treatment outcomes. To account for the effect of inflation, all expenditure-related variables were adjusted to 2017 constant prices using the Urban Residents Consumer Price Indices by Health Care, as published by the Guangzhou Statistics Bureau. Annual CPI values relative to the previous year were 106.1 in 2018, 103.8 in 2019, and 100.9 in 2020. Cumulative CPI multipliers were calculated to convert all values to 2017-based real terms. Subsequently, natural logarithmic transformations were applied to stabilize variance and improve comparability.

We used single-group interrupted time series analysis to compare the changes in various indicators before and after the implementation of the DIP payment reform (19–21). We constructed interrupted time series models in months, with January 2018 as the policy intervention month, involving 42 time points in three years from January 2017 to June 2020, including 12 months before and 30 months after the intervention. The single-group interrupted time series model is as follows. Y t = β 0 + β 1 T t + β 2 X t + β 3 T t X t + ε t .

In this model, Y t represents the indicators of interest in this paper for each month, and T t is the monthly linear time trend over 42 months. X t is the policy intervention variable, with January 2018 serving as the boundary. The 12 months prior to the intervention were assigned a value of 0, and the 30 months following the intervention were assigned a value of 1. T t X t is a continuous variable indicating the time variable after the implementation of the DIP payment reform, with a value of 0 before the intervention and a value of 1 after the intervention.

β 0 represents the baseline level of the indicators, β 1 reflects the monthly slope of each indicator before the implementation of the DIP payment reform, and β 2 represents the change at the first point after the implementation of the DIP payment reform (i.e., January 2018), indicating the short-term effect of the policy. β 3 represents the change in the monthly trend of the outcome indicator following the implementation of the DIP payment reform, relative to the pre-intervention slope, reflecting the long-term effect of the policy. To account for potential autocorrelation in the time series data, we employed the Prais-Winsten estimation method with robust standard errors and assessed model adequacy using the Durbin-Watson statistic. Additionally, we conducted the Ljung-Box tests on the residuals to examine whether they satisfied the white noise assumption.

Table 1 summarizes the descriptive characteristics of inpatient cases included in this study. A total of 472,958 admissions were recorded during the pre-reform period (January to December 2017), and 1,841,989 during the post-reform period (January 2018 to June 2020). The proportion of male patients remained relatively stable, accounting for 45.48% before the reform and 46.91% after. The average age of inpatients decreased from 66.87 years (SD = 16.66) to 59.65 years (SD = 18.60), and the mean length of stay declined from 10.09 days (SD = 10.12) to 9.50 days (SD = 9.79). The 30-day readmission rate was 5.37% before the reform and 4.61% after the reform. The average total inpatient cost was 15,320.16 RMB (SD = 23,143.62) in the pre-reform period and increased to 16,973.84 RMB (SD = 25,608.85) in the post-reform period. Among its components, medication expenditures decreased from 4,965.40 RMB (SD = 9,324.10) to 4,320.24 RMB (SD = 7,673.69), medical consumables expenditures increased from 2,925.69 RMB (SD = 10,032.37) to 3,629.66 RMB (SD = 11,710.53), and diagnostic and therapeutic service expenditures increased from 7,429.07 RMB (SD = 9,534.62) to 9,023.94 RMB (SD = 11,416.40). Additionally, out-of-pocket (OOP) payments per case rose from 3,388.28 RMB (SD = 5,972.84) to 4,216.85 RMB (SD = 7,673.69), indicating a potential increase in patient financial burden. Based on this dataset, we will conduct further empirical analysis to assess the causal impact of the DIP reform.