An unintended complementarity: how the outpatient pooling policy increases inpatient services in China Authors: Jianhui Wang, Mengyao Zhang, Miao Yu Published: 2026 ABSTRACT AbstractThis study investigates the relationship between outpatient and inpatient services within China’s tiered healthcare system following the implementation of outpatient pooling reform. Data were sourced from official administrative healthcare service records across counties in Sichuan Province (2010–20) and individual survey data from the China Health and Retirement Longitudinal Study (2011–18). Using a staggered difference-in-differences design with an imputation model, we identified an unexpected increase in hospital admissions post-reform, which was contrary to the policy’s intent to optimize healthcare resource allocation. We found that increased use of outpatient services was positively associated with demand for inpatient care, demonstrating a complementary effect. Mechanism analysis revealed that the rising hospital admissions were partly attributable to two factors: (i) expansion of outpatient services in primary care institutions generating an induced hospitalization effect that contributed to the rise in inpatient volume and (ii) hospitals improving inpatient service efficiency by shifting selected diagnostic and treatment tests to the outpatient account. These findings confirm the complex interaction between outpatient and inpatient services and underscore the need for a more effective design of the medical insurance payment system to enhance the efficiency of China’s healthcare services.Keywords: outpatient pooling reform, inpatient services, complementary effect, tiered diagnosis and treatment, China China’s Basic Resident Insurance outpatient pooling reform unexpectedly increased hospital inpatient service volume, revealing a complementary rather than substitutive relationship between outpatient and inpatient care. China’s Basic Resident Insurance outpatient pooling reform unexpectedly increased hospital inpatient service volume, revealing a complementary rather than substitutive relationship between outpatient and inpatient care. The complementary relationship was partly driven by an induced hospitalization effect: the reform increased outpatient visits at primary care institutions, some of which subsequently flowed upward to higher-level hospitals as inpatient admissions. The complementary relationship was partly driven by an induced hospitalization effect: the reform increased outpatient visits at primary care institutions, some of which subsequently flowed upward to higher-level hospitals as inpatient admissions. Hospitals further amplified inpatient admissions by reconfiguring service delivery, shifting selected diagnostic procedures from inpatient to outpatient settings, and expanding inpatient service capacity. Hospitals further amplified inpatient admissions by reconfiguring service delivery, shifting selected diagnostic procedures from inpatient to outpatient settings, and expanding inpatient service capacity. The unintended complementarity highlights the need for coordinated policy design that aligns outpatient benefit depth, provider payment incentives, and primary care capacity to achieve system-wide efficiency. The unintended complementarity highlights the need for coordinated policy design that aligns outpatient benefit depth, provider payment incentives, and primary care capacity to achieve system-wide efficiency. Introduction Since the early 2000s, healthcare systems in low- and middle-income countries have increasingly faced the persistent challenge of excessively high hospitalization rates. Such patterns not only lead to inefficient allocation of healthcare resources but also place substantial financial pressure on health insurance systems and may ultimately compromise patient outcomes (Yip and Hsiao 2014, Miao et al. 2018). At the same time, primary care systems, responsible for delivering outpatient services and public health functions, often lack sufficient capacity to meet population health needs. This imbalance contributes to unnecessary hospital admissions and exacerbates congestion in higher-level facilities (Ricketts et al. 2001, Lau et al. 2021). Against this backdrop of misconfiguration of healthcare services, expanding access to outpatient services has been widely regarded as a key policy instrument to substitute for inpatient care, improve system efficiency, and contain costs (Vitikainen et al. 2010, Du et al. 2022). Many countries have therefore sought to strengthen the gatekeeping role of primary care and promote tiered healthcare systems (Forrest 2003), in which patients are encouraged to seek first-contact care at lower levels (Liu et al. 2025). Nevertheless, the effectiveness of such reforms ultimately depends on the relationship between outpatient and inpatient care across different levels of the healthcare system (Elek et al. 2019). In certain countries with advanced primary care institutions (such as the USA and many countries in Europe), general outpatient care is largely provided by primary care institutions. These organizations increase accessibility of medical care at a local level, better manage chronic conditions, and provide rehabilitation follow-up services. They therefore contribute to health recovery and ultimately reduce the risk of future hospitalization (Reiss-Brennan et al. 2016, Elek et al. 2019). While a substantial body of literature suggests a substitutive relationship, where improved outpatient access reduces hospital admissions (Vitikainen et al. 2010), emerging evidence points to a more complex and sometimes contradictory pattern (Jiang et al. 2025, Jing 2025). Studies from Japan and China indicate that increased outpatient utilization may, under certain conditions, lead to higher inpatient demand, revealing a complementary relationship between the two (Takaku 2016, Lau et al. 2021). These findings suggest that outpatient expansion may not automatically alleviate hospital pressure, particularly in contexts where primary care capacity is limited or institutional incentives are misaligned (Kaestner and Khan 2012, He 2022). Despite this growing debate, existing research has paid limited attention to the conditions and mechanisms under which such complementarity arises (Fortney et al. 2005). China provides a critical context in which to examine this issue. The country has long experienced unusually high hospitalization rates, reaching 17.3%, significantly above the average observed in major economies (OECD 2025). This pattern reflects structural imbalances in the healthcare system (Sun et al. 2016, Wang et al. 2020), including historically limited outpatient coverage under basic medical insurance schemes, which incentivized patients to seek inpatient care even for relatively minor conditions (He 2022, Zhang et al. 2025). To address these distortions, the Chinese government introduced an outpatient pooling reform under the Basic Medical Insurance for Residents [Where “residents” refers to urban non-employed populations, rural populations, children, the elderly, low-income households and special groups in China.] (RBMI-OP reform), aiming to reduce cost sharing for outpatient care, encourage primary care utilization, and ultimately decrease avoidable hospital admissions. The reform mainly included: (i) the establishment of outpatient pooling funds to alleviate the financial burden on service users; (ii) differentiated reimbursement rates, providing higher reimbursement for visits to primary care institutions to encourage people to attend primary care first; and (iii) threshold standards, outpatient reimbursement cap, and percentages of outpatient reimbursement to regulate fund use. Following the policy change, cost sharing for outpatient care was reduced, and cost sharing for inpatient admissions and other treatments was maintained at the historical level. Despite these intentions, the effects of the RBMI-OP reform remain underexplored. Existing studies have focused primarily on a parallel reform in the employee insurance scheme (UEBMI), which has been shown to generate a substitution effect between outpatient and inpatient services (He 2022, Jiang et al. 2025, Wang and Ye 2026). In contrast, the RBMI-OP reform differs substantially in terms of funding levels and benefit design, particularly the relatively low outpatient reimbursement caps. These institutional features may alter patient behaviour and provider incentives in ways that do not necessarily align with the expected substitution effect. Against this background, this study asks how and under what conditions outpatient insurance expansion generates complementarity, rather than substitution, between outpatient and inpatient care. To be concrete, this study examines whether the RBMI-OP reform increased hospital inpatient utilization, through what mechanisms this effect occurred, and how local medical resource endowment and outpatient reimbursement caps moderated the reform effect. To answer these questions, we draw on county-level administrative data from Sichuan Province (2010–20) and individual-level data from the China Health and Retirement Longitudinal Study (CHARLS, 2011–18). Using a staggered difference-in-differences (DID) design, we find that the RBMI-OP reform leads to a delayed but persistent increase in hospital inpatient admissions, surgical cases, and outpatient-to-inpatient conversion rates. This study identifies two key mechanisms underlying this pattern: (i) an induced hospitalization effect, whereby increased outpatient utilization at the primary care level generates additional referrals to hospitals; and (ii) a reconfiguration of hospital service delivery, in which diagnostic services are shifted from inpatient to outpatient settings, thereby expanding inpatient admissions capacity. Furthermore, we show that the complementary effect is more pronounced in regions with greater medical resource endowment and in areas with lower outpatient reimbursement caps. This study contributes to the literature in three ways. First, it challenges the prevailing assumption that expanding outpatient coverage necessarily substitutes for inpatient care, demonstrating instead that complementarity may emerge under specific institutional conditions. Second, it advances the understanding of outpatient–inpatient interactions by identifying both demand-side and supply-side mechanisms through which policy reforms shape healthcare utilization. Third, it highlights the critical role of policy design, particularly reimbursement caps and provider incentives, in expanding the number of hospital admission. These findings offer important implications for health system reforms in China and other middle-income countries seeking to strengthen primary care while containing hospital demand. Conceptual framework This study develops a conceptual framework to examine the potentially counterintuitive effects of outpatient insurance expansion on hospital inpatient service utilization. In the first place, it is necessary to clarify the concept of complementarity with regard to the above. In contrast to a substitution relationship (where an increase in the utilization of one type of service leads to a decrease in the utilization of another), a complementary relationship refers to a situation where an improvement in the quality of one type of healthcare service leads to a concurrent increase in the utilization of another (Lau et al. 2021, Zhu et al. 2025). This study examines whether payment reforms for outpatient services may give rise to an anomalous phenomenon, namely an increase in the utilization of inpatient services. Considering the RBMI-OP reform, the reimbursement rates for outpatient services have been increased, in particular in primary care facilities, thereby encouraging patients to seek initial consultations at the primary level. Nevertheless, within China’s non-compulsory tiered healthcare system, it remains uncertain whether patients will continue to seek care at primary care facilities. On the demand side, due to differences in perceptions of healthcare quality and trust, patients often prefer to choose higher-level hospitals (Li et al. 2017, Lu et al. 2019). On the supply side, healthcare providers respond strategically to financial incentives embedded in the payment system, adjusting admission and referral decisions accordingly (Ellis and Ruhm 1988). These dynamics ultimately redirect patients back to hospitals, indirectly contributing to a rise in hospital admissions (Zhou et al. 2021, Wang et al. 2023). Accordingly, this study proposes the following basic hypotheses: Hypothesis 1 (H1) The RBMI-OP reform increases demand for hospital inpatient services. The complementary effect is expected to be driven by two mechanisms. First, the policy significantly reduces the out-of-pocket expenses of outpatient visits, especially at primary care institutions, thereby promoting the utilization of services at this level (Forrest 2003, Elek et al. 2019, Jing 2025). However, such newly generated medical demands may ultimately be diverted from primary care institutions to higher-level hospitals: This is because patients who either reach their insurance reimbursement cap or require complex diagnosis are incentivized to seek care via inpatient admissions at higher-level hospitals. In this scenario, expanded access to outpatient care induces greater utilization of inpatient services, as primary care institutions redirect people to hospitals again—whether driven by clinical necessity or financial constraints. Hypothesis 2 (H2) The RBMI-OP reform generates an induced hospitalization effect, whereby increased outpatient utilization at primary care institutions leads to higher inpatient admissions at hospitals. Second, the policy changes also interact with the supply-side response of hospitals. Under the global budget prepayment system, Chinese public hospitals operate under a ‘no compensation for overspending, retention of surplus’ rule, which creates incentives to control per-episode inpatient costs (MOHRSS 2012). In this context, shifting diagnostic and auxiliary procedures from inpatient to outpatient settings allows hospitals to reduce inpatient cost per case while maintaining or expanding service volume. The RBMI-OP reform further strengthens this incentive by increasing outpatient reimbursement, making outpatient service provision more financially viable. In addition, national performance assessment systems and policy guidelines encourage hospitals to improve operational efficiency by shortening length of stay (LOS) and increasing bed turnover rates (State Council General Office 2015, 2022, National Health Commission 2024). These institutional pressures reinforce hospitals’ incentives to reallocate services across care settings. As a result, outpatient expansion may free inpatient capacity, accelerate bed turnover, and enable hospitals to admit more patients overall. Hypothesis 3 (H3) The RBMI-OP reform incentivizes hospitals to shift diagnostic and auxiliary procedures from inpatient to outpatient settings, thereby freeing inpatient bed capacity, accelerating bed turnover, and increasing overall inpatient admission volumes. Figure 1 shows the article’s analytical framework. The analytical framework for this study. Materials and methods Study setting This study focuses on Sichuan Province, a large and socioeconomically diverse province in southwestern China. In 2020, Sichuan had a population of 83.71 million and a per capita gross domestic product (GDP) of ∼59 000 RMB, ranking 16th among China’s 31 provinces. Its level of economic development is broadly comparable to the national average, making it a representative setting for examining healthcare system dynamics in middle-income regions. Sichuan has a relatively well-developed healthcare system, comprising 2435 hospitals and 79 491 primary care institutions (Sichuan Statistical Bureau 2020). In terms of primary care capacity, Sichuan’s number of primary care beds per 1000 population (0.8) and health personnel per 1000 population (1.9) are close to the national averages (0.9 and 2.1, respectively), and the share of primary care visits in total consultations (52%) is broadly consistent with the national average (∼50%; National Bureau of Statistics of China 2026). However, at the same time, the province exhibits substantial intra-provincial variation in economic development and healthcare resource distribution, ranging from highly urbanized metropolitan areas to resource-constrained rural and mountainous regions. This heterogeneity provides a useful setting for examining how institutional and resource conditions shape healthcare utilization. It is observed in China that the establishment of an outpatient pooling system is closely associated with local economic status and the distribution of healthcare resource (Zhang and Chen 2025, Zhang et al. 2025). Given the significant disparities in its economic and healthcare conditions, Sichuan started RBMI-OP reform early within its population in 2009. Pioneer cities such as Chengdu and Yibin became among the earliest adopters nationally. Subsequently, other cities in Sichuan also successively established the outpatient pooling system for residents’ medical insurance. These characteristics make Sichuan an interesting case study to assess the effects of RBMI-OP reform in hospitals’ admissions. Table 1 summarizes the implementation timeline and key features of the outpatient pooling system across cities. Although most regions provide higher reimbursement rates for primary care visits to encourage first-contact care, policy design varies substantially across locations. For instance, Yibin City reimburses 70% of outpatient service expenditure at community health centres but only 60% at tertiary hospitals. In Meishan City, outpatient reimbursement is restricted to contracted medical institutions through patient–provider agreements, reinforcing first-contact care at primary care institutions. In addition, in China's most regions, the highest reimbursement rate is provided when people directly visit primary care institutions. The reimbursement rate is reduced by 10%–20% or patients may be ineligible if they go to higher-level hospitals without a referral (Lu et al. 2019). Table 1Summary of RBMI-OP reform across cities in Sichuan Province, China.CityYear of reformCoverage scope of pooling fundsThreshold standard (CNY/year)Outpatient reimbursement cap (CNY/year)Outpatient reimbursement for PCI (%)Chengdu2009Primary care–20060Zigong2011Primary and secondary care–10070Panzhihua2012Primary and secondary care–Tier 1: 100Tier 2: 30070Luzhou2011All levels care–13070Deyang2011Primary care–Tier 1: 100Tier 2: 20075Mianyang2015All levels care–9070Guangyuan2015Primary and secondary care–11050Suining2014All levels care–9080Neijiang2011Primary care–17070Leshan2011Primary care–16060Nanchong2015Primary care–12060Meishan2012All levels care–10060Yibin2010All levels care–15070Guangan2020All levels care–21050Dazhou2011All levels care–12070Ya’an2009Primary care50Tier 1: 120Tier 2: 30080Bazhong2009All levels care–10050Ziyang2019Primary care–16880The information on this table is collected by the authors. Due to the limitations of observation data, four cities, including Chengdu, Bazhong, Ya’an, and Yibin, were not included in empirical analysis. PCI, primary care institution. Summary of RBMI-OP reform across cities in Sichuan Province, China. The information on this table is collected by the authors. Due to the limitations of observation data, four cities, including Chengdu, Bazhong, Ya’an, and Yibin, were not included in empirical analysis. PCI, primary care institution. The design of benefits also varies substantially across cities in Sichuan, reflecting local differences in policy implementation. The outpatient reimbursement caps range from 90 RMB (13 USD) in Mianyang City to 300 RMB (42 USD) for the second-tier plan in Panzhihua City. The percentage of outpatient reimbursement at primary care institutions is generally between 50% and 80%, and most areas have abolished the threshold standard to encourage initial contact at the primary care level. Our focus is the outpatient pooling reform under the insurance scheme for rural and urban residents. In 2024, 71.39% of the total population was enrolled in this scheme, making it China’s most significant basic medical insurance programme (National Healthcare Security Administration 2025). The effects of RBMI-OP reform on the healthcare system therefore warrant in-depth examination. Data This study integrates four major data sources. First, county-level administrative data on healthcare service utilization (2010–20) are obtained from the ‘Sichuan Health Statistical Yearbook’. Second, official policy documents provide detailed information on the timing of RBMI-OP reform implementation across the prefecture-level cities. Third, socioeconomic and healthcare system indicators are drawn from the ‘Sichuan Statistical Yearbook, China Statistical Yearbook, and China City Statistical Yearbook’. Finally, individual-level data from the CHARLS (2011–18) are used to analyse changes in healthcare-seeking behaviour among residents aged 45 and above. Variables We constructed three dependent variables: ‘Hospital inpatient counts’, ‘Hospital surgical cases’, and ‘Conversion Rate from Outpatient to Inpatient Care’. The hospital inpatient count was a key indicator of a hospital’s service volume and reflects the scale of inpatient service use over a given period. It was estimated from the logarithm of the number of discharges. The hospital surgical cases captured the most technology and resource-intensive components of inpatient services and reflect the intensity of inpatient operations. It was estimated from the logarithm of the number of surgeries performed on hospitalized patients. The conversion rate from outpatient to inpatient care reflected the number of admissions per 100 outpatient visits, capturing the conversion ratio between outpatient and inpatient services at the hospital level. An increase in this indicator suggests a larger share of outpatient visits was converted into inpatient admissions, reflecting a higher utilization intensity of hospital inpatient services. This variable was also logarithmically transformed in empirical analysis. The key explanatory variable in this study was the RBMI-OP reform. A binary variable was constructed to capture the timing of reform implementation. Specifically, if city i carried out the reform in year t, then the RBMI-OP reformit equals 1. To leave enough transitive time and reflect timely policy effects, this study adjusted the policy year t to t + 1 for reforms officially issued between July and December of year t, following the conventional practice in prior studies (Chen et al. 2022). Supplementary Figure S1 portrayed the distribution of the treatment and control groups across the years. To account for potential confounding factors, this study selected the following control variables guided by the Andersen behavioural model of health services utilization (Andersen, 1973). For predisposing characteristics, we included the proportion of the population aged 65 and above and the gender ratio to control for the effects of population ageing and gender structure on the disease spectrum and healthcare demand. For enabling resources, GDP per capita captures the level of economic development and its influence on hospital operations and care-seeking behaviour (Zheng et al. 2025); the number of hospital beds per 1000 population controls for the potential demand-inducing effect of supply-side expansion; the proportion of fiscal medical assistance (defined as the proportion of government fiscal subsidy revenue in the total revenue of healthcare institutions) reflects the influence of government health investment on hospital behaviour (Li et al. 2018); and the urbanization rate controls for the confounding effects of urban–rural disparities in insurance benefits and service accessibility. For need factors, we included perceived need factors like self-assessed health and evaluated need factors like activities of daily living (ADL; assessed by 6 items: dressing, bathing, eating, transferring, toileting, and continence; each scored 0–3, higher score indicates worse function) index when we pursued our analysis of the mechanisms related to individual medical service demands. All control variables were matched to the year of observation. All the variables are measured at the county level except when analysing individual medical care demand behaviour using CHARLS data (where the unit of observation is the individual). Empirical strategy We adopt a quasi-experimental design using a staggered DID framework to estimate the impact of the RBMI-OP reform on inpatient service utilization. The baseline specification is: where Yit denotes the outcome variable of county i in year t; and the binary treatment variable Dit equals 1 if county i has implemented the reform in year t, and 0 otherwise; αi and βt show county and year fixed effects; and εit is the error term. The coefficient τ captures the average treatment effect of the reform on inpatient outcomes before and after implementation. However, recent studies have revealed that the conventional practice of implementing DID studies via two-way fixed-effect regressions may introduce inherent biases or even incorrect signs when treatment is adopted at different times and treatment effects are heterogeneous (Callaway and Sant’Anna 2021, Goodman-Bacon 2021, Baker et al. 2022, Borusyak et al. 2024). In staggered adoption settings, standard two-way fixed-effects models may compare newly treated units not only with untreated units but also with units that were treated earlier. Such comparisons are valid only under strong assumptions of homogeneous treatment effects across cohorts and over time. When these assumptions are violated, the estimator may assign inappropriate or negative weights to some treatment effects, thereby distorting the estimated average effect. This concern is particularly relevant in our study because the RBMI-OP reform was implemented sequentially across cities in Sichuan Province. Some cities had already adopted the reform before the observation period, while most implemented it between 2010 and 2020. Given this staggered rollout and the relatively long observation window, potential bias arising from heterogeneous treatment effects cannot be ignored. To address this concern, we applied the imputation-based DID estimator proposed by Borusyak et al. (2024), which has been increasingly adopted in recent empirical studies (Pan 2025, Papp and Oremus 2025). This approach estimates untreated potential outcomes using observations that have not yet been treated, imputes counterfactual outcomes for treated observations, and then calculates treatment effects as the difference between observed and imputed outcomes. Compared with conventional two-way fixed-effects models, this estimator allows for heterogeneous treatment effects across cohorts and over time, provided that the assumptions of parallel trends and no anticipation hold. The corrected specification is presented as follows: where Yit and Dit remain the same as Equation (1). Ait and Xit are known non-stochastic vectors that extend the general model of Y(0). Ait′λi incorporates county fixed effects but additionally allows interactions with observable covariates that are unaffected by treatment status; Xit′δ accommodates time fixed effects and allows any time-varying covariates. The target parameter of the treatment effect to be estimated in this model is defined as τ=Γitθ, where Γ is a known N1×(N1−M) matrix of full column rank and θ is a (N1−M)×1 vector of unknown parameters. Under assumptions of parallel trends and no-anticipation effects, any linear estimator τw^ of τw that is unbiased under arbitrary treatment-effect heterogeneity can be obtained via imputation. The specific estimation steps are as follows: For every treated observation, estimate expected untreated potential outcomes Ait′λi+Xit′δ by some unbiased linear estimator Yit^(0) using data from the untreated observations. For every treated observation, estimate expected untreated potential outcomes Ait′λi+Xit′δ by some unbiased linear estimator Yit^(0) using data from the untreated observations. For each treated observation, set τit^=Yit−Yit^(0). For each treated observation, set τit^=Yit−Yit^(0). Estimate the target by a weighted sum τw^=∑itεΩ1witτit^. Estimate the target by a weighted sum τw^=∑itεΩ1witτit^. While relaxing the assumption of treatment-effect homogeneity, the DID imputation method imposes stricter requirements on the data structure of sample units included in the analysis: First, sample units must have untreated observations within the study window. Second, to avoid the spurious identification of long-run effects, untreated units must be available in every period. Units failing to meet these conditions were excluded from the estimation sample. In the empirical study, starting from an initial panel of 1452 county-year observations across 18 prefecture-level cities in Sichuan Province (2010–20), the sample units applied the following exclusion criteria required by the DID_imputation estimator: First, 451 observations from four cities (Chengdu, Bazhong, Ya’an, and Yibin) were excluded because these cities had already implemented the RBMI-OP reform prior to the observation window (2009 or 2010). Second, observations from 2020 were partially excluded (net exclusion: 91 observations) because all counties in Sichuan had completed the reform by that year. This yielded a final analytical sample of 910 county-year observations. All analyses were conducted using Stata 16. Descriptive statistics are reported in Table 2. Table 2Descriptive statistics.VariablesObs.MeanStd. dev.MinMaxIndependent variable RBMI-OP reform9100.6900.4600.0001.000Dependent variables Hospital inpatient counts (logarithmic form)91010.6500.9006.28012.690 Hospital surgical cases (logarithmic form)9109.1801.1304.44012.040 Conversion rate from outpatient to inpatient care (logarithmic form)9102.0500.3400.8503.030Control variables Ageing rate91013.6901.5109.13017.350 Hospital beds per thousand people9101.5201.0200.4206.430 GDP per capita (logarithmic form)91010.3200.5108.88011.730 Fiscal medical assistance9100.1800.0900.0500.590 Urbanization rate9100.3100.1900.0500.980 Gender ratio9101.0600.0400.9301.180Obs., observations; std. dev., standard deviation. Descriptive statistics. Obs., observations; std. dev., standard deviation. Results Baseline results The regression results show a positive relationship between establishing outpatient pooling funds and hospital inpatient services (Table 3). Following RBMI-OP reform, there was an 18.06% (e0.166≈118.06%) increase in hospital inpatient counts [95% confidence interval (CI) 0.096–0.236, P < .01], and a 23.12% (e0.208≈123.12%) increase in hospital surgical cases (95% CI 0.122–0.295, P < .01). The conversion rate from outpatient to inpatient care was also predicted to increase by 24.86% (e0.222≈124.86%) following the reform. These results suggest that although the RBMI-OP reform reduced the cost sharing of outpatient services, the volume of hospital inpatient care also increased, with a higher proportion of outpatient visits resulting in hospital admission. This confirms H1. Table 3Impact of RBMI-OP reform on hospital inpatient service volumes.(1)(2)(3)Hospital inpatient countsHospital surgical casesConversion rate from outpatient to inpatient careRBMI-OP reform0.166***0.208***0.222***(0.036)(0.044)(0.041)Ageing rate−0.244−0.274−0.306**(0.222)(0.226)(0.147)Hospital beds per thousand people0.1320.278***0.016(0.090)(0.086)(0.062)GDP per capita−0.2020.279−0.044(0.287)(0.383)(0.192)Fiscal medical assistance−2.762−3.812**−0.603(1.728)(1.786)(0.702)Urbanization rate−0.0426−0.6390.551(0.930)(1.057)(0.491)Gender ratio6.362**1.4082.810(3.207)(4.222)(3.874)Year-fix effectsYesYesYesCounty-fix effectsYesYesYesCounty numbers999999Observations910910910Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Impact of RBMI-OP reform on hospital inpatient service volumes. Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Dynamic effects and lagged impact The analysis of dynamic effects reveals a significant temporal lag in the impact of the outpatient pooling reform (Table 4). In the initial period (Years 0–2) immediately following the policy’s implementation, there was no statistically significant change in any of the three hospital inpatient indicators. However, a positive and statistically significant effect on hospital inpatient volume emerged in Years 3–5. These findings indicate that the system adjustment is not immediate and takes several years to fully materialize. Table 4Dynamic effects of RBMI-OP reform on hospital inpatient service volumes.(1)(2)(3)Hospital inpatient countsHospital surgical casesConversion rate from outpatient to inpatient careCurrent year for policy−0.006−0.0110.003(0.025)(0.040)(0.018)Lagged 1 year0.029−0.0250.007(0.029)(0.047)(0.029)Lagged 2 years0.0260.0040.001(0.034)(0.049)(0.036)Lagged 3 years0.189***0.221***0.193***(0.037)(0.054)(0.038)Lagged 4 years0.191***0.217***0.280***(0.041)(0.054)(0.042)Lagged 5 years0.330***0.468***0.478***(0.060)(0.074)(0.076)ControlsYesYesYesYear-fix effectsYesYesYesCounty-fix effectsYesYesYesObservations910910910Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. The reference period is the year immediately preceding reform implementation. ‘Current year for policy’ refers to the year of reform implementation. ‘Lagged k years’ refers to k years after reform implementation. Dynamic effects of RBMI-OP reform on hospital inpatient service volumes. Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. The reference period is the year immediately preceding reform implementation. ‘Current year for policy’ refers to the year of reform implementation. ‘Lagged k years’ refers to k years after reform implementation. Robustness analysis The DID approach relies on the assumption of parallel trends and no-anticipation effects. We employed an event-study approach (Borusyak et al. 2024) to test the critical parallel trends assumption (Fig. 2). Prior to the implementation of the outpatient pooling reform, the coefficients capturing the differences between treatment and control groups were statistically insignificant. This confirms that the two groups shared a common trend in hospital inpatient service volume before the policy intervention. Parallel trends test results. This figure presents the dynamic effect coefficients estimated from the event-study analysis, used to verify the parallel trends assumption of the DID identification strategy. The horizontal axis represents time relative to reform implementation, where pre5 is the base year, tau0 denotes the year of the reform, tau1 is the year following the reform, and pre1 is the year preceding the reform. The vertical axis shows the estimated coefficients of the RBMI-OP reform variable. When the coefficients are significantly different from zero, this indicates marked differences between the treatment and control groups. To further rule out the possibility that the observed effects were driven by unobserved time trends or spurious correlation, a placebo test was conducted by randomly reassigning the treatment years/counties and recalculating the average treatment effect 500 times (Liu et al. 2021). The resulting distribution of the placebo coefficients was tightly clustered around zero (Fig. 3), while the estimated policy effect lay well outside this distribution. This outcome strongly supports the causal inference that the increase in hospital inpatient service volume is attributable to the reform. Placebo test results (500 simulated iterations). This figure displays the kernel density distribution of estimated coefficients from 500 random assignments of placebo treatment years and counties. The vertical dashed line indicates the estimated coefficient for the actual policy shock, confirming that the baseline results are not driven by unobserved factors or random chance. We additionally conducted a series of robustness checks to verify the credibility of the main findings, including sensitivity analyses of the treatment effect, exclusion of potential spatial spillover effects, and controls for concurrent policy interventions (namely, the case-based prospective payment reform (DRG/DIP reform), Integration of Urban and Rural Residents Basic Medical Insurance (URRBMI) reform, Medical Community Construction, Critical Illness Insurance expansion, and hierarchical medical system reform). We also excluded alternative explanations, such as case-mix changes and the potential confounding effects of the COVID-19 pandemic on inpatient volumes. Furthermore, alternative confounding factors, including differentiation of hospital samples and the development of primary care services, were considered. All robustness checks are reported in the Supplementary Materials, and the main regression results remained stable across all specifications. Mechanism analysis Induced hospitalization effect The reduction in cost sharing for outpatient care may have indirectly induced the expansion of hospital services through changes in patient flow. We examine the induced hospitalization effect at both county and individual levels. At the county level, Table 5 presents the effects of the RBMI-OP reform on service volumes across health tiers. Columns (1) and (2) show that the reform significantly increased outpatient visits at primary care institutions by 7.68% (e0.074≈107.68%) while significantly reducing inpatient admissions at primary care institution by 45.94% (e−0.615≈54.06%). These results confirm that the reform effectively incentivized residents to seek outpatient services at the primary care level. Table 5The induced hospitalization effects after RBMI-OP reform (macro level).(1)(2)(3)(4)PCI outpatient admissionsPCI inpatient admissionsDistribution ratio of hospital outpatient serviceDistribution ratio of hospital inpatient serviceRBMI-OP reform0.074**−0.615***−0.0030.113***(0.038)(0.177)(0.007)(0.025)ControlsYesYesYesYesYear-fix effectsYesYesYesYesCounty-fix effectsYesYesYesYesObservations910910910910The definition of the core variable is as follows: (i) primary care institution (PCI) outpatient admissions = Log(The number of outpatient and emergency visits at PCIs); (ii) PCI inpatient admissions = Log(The number of PCI hospital admissions); (iii) Distribution ratio of hospital outpatient service = The number of outpatient and emergency visits at hospital/(The number of outpatient and emergency visits at PCIs + The number of outpatient and emergency visits at hospital); (iv) Distribution ratio of hospital inpatient service = The number of hospital admissions/(The number of PCI admissions + The number of hospital admissions). Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. The induced hospitalization effects after RBMI-OP reform (macro level). The definition of the core variable is as follows: (i) primary care institution (PCI) outpatient admissions = Log(The number of outpatient and emergency visits at PCIs); (ii) PCI inpatient admissions = Log(The number of PCI hospital admissions); (iii) Distribution ratio of hospital outpatient service = The number of outpatient and emergency visits at hospital/(The number of outpatient and emergency visits at PCIs + The number of outpatient and emergency visits at hospital); (iv) Distribution ratio of hospital inpatient service = The number of hospital admissions/(The number of PCI admissions + The number of hospital admissions). Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Columns (3) and (4) further examine the distribution of outpatient and inpatient services across provider tiers. These indicators capture the proportion of hospital outpatient and inpatient services relative to total regional healthcare utilization across both hospitals and primary care institutions, serving as macro-level proxies for patients’ revealed preferences. While the change in the distribution ratio of hospital outpatient services was not statistically significant, the distribution ratio of hospital inpatient services increased by 11.96% (e0.113≈111.96%). This pattern suggests that incremental patient flows generated by the reform were primarily absorbed by hospital inpatient services rather than hospital outpatient departments, consistent with the induced hospitalization effect. At the individual level, we draw on data from the CHARLS (2011–18) to examine individual healthcare utilization decisions. As shown in Table 6, the reform significantly increased both the probability and frequency of outpatient visits among middle-aged and elderly residents (Columns 1 and 2), with particularly pronounced effects at the primary care level (Columns 3 and 4). Column (5) further shows that the reform increased the probability of hospital admission among this population by 4.1%, suggesting a structural shift towards more specialized inpatient services. These individual-level findings further corroborate the induced hospitalization mechanism: expanded access to primary care outpatient services indirectly drives the observed increase in hospital inpatient volumes. Taken together, evidence from both the macro-level patient flow analysis and the micro-level individual behavioural data converges in support of H2. Table 6The induced hospitalization effects after RBMI-OP reform (micro-level).(1)(2)(3)(4)(5)Outpatient visitOutpatient visit frequencyPCI outpatient visitPCI outpatient visit frequencyHospitalization statusRBMI-OP reform0.262***1.020***0.105***1.105***0.041**(0.002)(0.154)(0.010)(0.182)(0.016)ControlsYesYesYesYesYesYear-fix effectsYesYesYesYesYesIndividual-fix effectsYesYesYesYesYesObservations26922692269226922692The control variables for the micro-data include individual-level factors (age, marital status, self-assessed health, ADL score, income, and household registration) and city-level factors (ageing rate, number of beds per thousand people, per capita GDP, fiscal medical assistance, urbanization rate, and gender ratio). Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. PCI, primary care institution. The induced hospitalization effects after RBMI-OP reform (micro-level). The control variables for the micro-data include individual-level factors (age, marital status, self-assessed health, ADL score, income, and household registration) and city-level factors (ageing rate, number of beds per thousand people, per capita GDP, fiscal medical assistance, urbanization rate, and gender ratio). Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. PCI, primary care institution. Reconfiguration of outpatient and inpatient services To empirically examine the hospital service restructuring mechanism, we draw on medical expenditure indicators, which have been widely used as sensitive proxies for hospital operational behaviour (Fu et al. 2018). To be concrete, we focus on the share of diagnostic expenditure within total medical costs in both inpatient and outpatient settings, as this component is more readily adjustable by hospitals in response to changing reimbursement incentives (Zhu et al. 2022). As shown in Table 7, the reform was followed by a significant increase in the share of diagnostic service expenditure on outpatient services (β = 0.020, P < .05) and a significant decrease in the share of diagnostic service expenditure on inpatient services (β = −0.032, P < .05). Column (3) further indicates that the proportion of outpatient diagnostic expenditure to inpatient diagnostic expenditure increased significantly after the reform (β = 0.018, P < .01), suggesting that diagnostic costs expanded more rapidly in the outpatient setting than in the inpatient setting. Overall, these three results indicate that there was a systematic migration of diagnostic services from inpatient to outpatient settings. Table 7Distributional shifts in hospital diagnostic service expenditures following the RBMI-OP reform.(1)(2)(3)Share of outpatient diagnostic service expenditureShare of inpatient diagnostic service expenditureProportion of outpatient expenditure to inpatient expenditure in diagnostic serviceRBMI-OP reform0.020**−0.032***0.018***(0.009)(0.008)(0.003)ControlsYesYesYesYear-fix effectsYesYesYesCounty-fix effectsYesYesYesObservations910910910The definition of the core variable is as follows: (i) Share of outpatient diagnostic service expenditure = Per-episode diagnostic expenditure/Per-episode total medical expenditure (outpatient); (ii) Share of inpatient diagnostic service expenditure = Per-episode diagnostic expenditure/Per-episode total medical expenditure (inpatient); (iii) Proportion of outpatient expenditure to inpatient expenditure in diagnostic service = Per-episode outpatient diagnostic expenditure/Per-episode inpatient diagnostic expenditure. Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Distributional shifts in hospital diagnostic service expenditures following the RBMI-OP reform. The definition of the core variable is as follows: (i) Share of outpatient diagnostic service expenditure = Per-episode diagnostic expenditure/Per-episode total medical expenditure (outpatient); (ii) Share of inpatient diagnostic service expenditure = Per-episode diagnostic expenditure/Per-episode total medical expenditure (inpatient); (iii) Proportion of outpatient expenditure to inpatient expenditure in diagnostic service = Per-episode outpatient diagnostic expenditure/Per-episode inpatient diagnostic expenditure. Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Table 8 provides further evidence linking this service reconfiguration to improvements in hospital operational efficiency. Two commonly used indicators, LOS and bed turnover rate, are employed to capture hospital efficiency (Qian et al. 2019, Zhu and Yu 2025). Column (1) shows that a higher outpatient diagnostic expenditure share is significantly associated with a shorter average LOS (β = −0.413, P < .01), while Column (2) further indicates the service restructure has led to a higher bed turnover rate (β = 0.654, P < .01). Table 8Impact of diagnostic service reallocation on hospital operational efficiency.(1)(2)Average length of hospital stayBed turnover rateShare of outpatient diagnostic service expenditure−0.413***0.654***(0.125)(0.202)ControlsYesYesYear-fix effectsYesYesCounty-fix effectsYesYesObservations910910The definition of the core variable is as follows: (i) Average length of hospital stay = Total inpatient days/Total number of discharges; (ii) Bed turnover rate = Number of discharged patients/Average number of hospital beds per year. All columns are estimated using a fixed-effects panel model with county and year fixed effects. Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Impact of diagnostic service reallocation on hospital operational efficiency. The definition of the core variable is as follows: (i) Average length of hospital stay = Total inpatient days/Total number of discharges; (ii) Bed turnover rate = Number of discharged patients/Average number of hospital beds per year. All columns are estimated using a fixed-effects panel model with county and year fixed effects. Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Overall, the evidence from Tables 7 and 8 indicate that the development of the outpatient pooling system incentivized hospitals to reorganize service delivery in ways that simultaneously expanded outpatient diagnostic volume and improved inpatient admission capacity, thereby increasing the number of hospital admissions. This provides support for H3. Heterogeneity analysis To assess how contextual factors modify the impact of the policy, we carried out heterogeneity analysis using regional resource endowments and reimbursement policy design (Table 9). First, we find that the positive complementary effect on hospital inpatient services was conditional, based on local medical resource endowment (Table 9). In regions with more medical resources (above-average number of hospital beds per 1000 population), there was a significant positive increase across all three inpatient use indicators. In contrast, in regions with weaker medical resources, inpatient admissions increased, but the increase in technology-intensive surgical cases and the outpatient referrals were less pronounced than in the higher-resource group. These findings show that the complementary effects are more likely to happen in regions with greater medical resources, probably because hospitals there are better equipped to absorb the induced inpatient demands and deal with technology- and resource-intensive surgeries. Table 9Results of heterogeneity analysis.Lower groupHigher groupInter-group coefficient differenceLocal medical resource endowment Hospital inpatient counts0.124***0.119***−0.005(0.038)(0.045) Hospital surgical cases0.0610.209***0.148**(0.054)(0.051) Conversion rate from outpatient to inpatient care0.152***0.327***0.175***(0.042)(0.054)Outpatient reimbursement caps Hospital inpatient counts0.204***0.043−0.162***(0.036)(0.063) Hospital surgical cases0.273***−0.002−0.275***(0.048)(0.063) Conversion rate from outpatient to inpatient care0.257***0.108*−0.149**(0.044)(0.065)Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Results of heterogeneity analysis. Robust standard errors are given in parentheses. ***, **, and * mark significance at the 1%, 5%, and 10% levels, respectively. Second, we observe that the cap lines of outpatient reimbursement rates across different counties introduced significant heterogeneity (Table 9). Cities with outpatient reimbursement caps above the provincial mean (136.24 RMB ∼20 USD, per year, as reported in Table 1) are classified as high-cap areas, while those below or equal to this threshold are classified as low-cap areas. (For cities that implement two tiers of outpatient reimbursement caps, we uniformly selected the first-tier outpatient reimbursement cap as the baseline for regional outpatient reimbursement caps.) The increase in inpatient services was more pronounced in low-cap areas. In contrast, in high-cap areas, hospital inpatient counts and surgical cases do not show statistically significant increases, and the rise in outpatient-to-inpatient conversion rates is smaller than the 29.3% (e0.257 ≈ 129.30%) increase observed in low-cap areas. These findings suggest that the complementary effects will be diminished if the outpatient reimbursement caps are further increased. This is primarily because enhanced outpatient coverage reduces the financial barriers for patients seeking comprehensive care at the primary level, decreasing the subsequent demand for inpatient services that might otherwise arise from inadequate initial treatment. Discussion The findings of this study challenge the view that expanding outpatient coverage necessarily relieves pressure on inpatient services. Consistent with studies showing that increased outpatient spending can be associated with more hospital admissions (Kato and Goto 2017, Lau et al. 2021), the RBMI-OP reform jointly increased hospital inpatient counts, hospital surgical cases, and the conversion rate from outpatient to inpatient care, confirming the existence of a complementary effect between outpatient pooling reform and the growth of inpatient services. This result stands in notable contrast to the substitution effect observed following China’s parallel employee insurance reform (He 2022, Wang and Ye 2026). The divergence underscores how institutional features, particularly the generosity of insurance benefits, provider incentive structures, and primary care capacity, critically shape the outpatient–inpatient relationship. Dynamic event study estimates further reveal that the complementary effect emerged with a lag of 2–3 years following reform implementation, suggesting that the outpatient–inpatient complementarity is not an immediate policy response but rather the cumulative product of gradual adaptation on both the demand and supply sides. On the demand side, residents require time to develop awareness of policy changes and to adjust their healthcare-seeking behaviour accordingly. As shown in prior research, utilization responses to health insurance subsidies often take several years to fully manifest even after enrolment responses have materialized (Asuming 2013). On the supply side, the reconfiguration of hospital services, such as shifting diagnostic procedures across care settings, requires organizational adaptation, including workflow redesign and coordination across departments (English et al. 2026). These gradual adjustments are consistent with prior evidence that the effects of health system reforms often materialize over extended periods (Walker et al. 2019, Wattal et al. 2024). This study further advances the literature by proposing two theoretically distinct mechanisms, which are the induced hospitalization effect and hospital service reconfiguration, to account for the observed complementarity, thereby extending Fortney et al.’s (2005) foundational discussion of the relationship between outpatient and inpatient services within the institutional context of China’s healthcare system. The first mechanism, the induced hospitalization effect, provides a demand-side explanation for the observed complementarity. This mechanism directly challenges Forrest et al.’s gatekeeping perspective. While the gatekeeping model assumes that primary care acts as a filter that reduces unnecessary escalation to higher-level care, our findings suggest that this function remains relatively weak in the Chinese context. Although the RBMI-OP reform substantially increased outpatient visits at primary care levels, it simultaneously generated an increase in inpatient admissions at higher-level hospitals. This pattern leads us to argue that primary care gatekeeping policies, when unaccompanied by adequate institutional capacity or sufficient reimbursement depth, may function less as a substitute for higher-level care than as a conduit towards it (Forrest 2003). The second mechanism, hospital service reconfiguration, provides a supply-side explanation and is broadly consistent with Fortney et al.'s observation that inpatient and outpatient services can complement one another through the sharing of diagnostic procedures. In the Chinese institutional context, operating under the convergent pressures of global budget prepayment, public hospital performance assessment, and hierarchical medical system mandates, the RBMI-OP reform incentivized hospitals to shift diagnostic and treatment procedures from the inpatient to the outpatient setting. This reconfiguration freed inpatient bed capacity, accelerated bed turnover, and ultimately expanded overall admission volumes. This finding also aligns with Vitikainen et al. (2010), who demonstrated that expanding outpatient service utilization can shorten average length of hospital stay and improve the operational efficiency of inpatient services. The heterogeneity analysis further sheds light on the conditions under which the complementary effect intensifies. First, we find that in areas with relatively abundant medical resources, the complementary effect is more pronounced, reinforcing the hospital admission capacity pathway and suggesting that well-resourced hospitals are better positioned to absorb and respond to the additional patient demand generated by the reform (Kato and Goto 2017). In the Chinese healthcare context, tertiary hospitals possess the most abundant concentration of medical resources and specialist expertise. In the absence of mandatory hierarchical medical care system, patients generally exhibit a strong preference for tertiary hospitals (Lu et al. 2019). There is therefore a possibility that, following the RBMI-OP reform, patients are incentivized to utilize the formal referral system at the primary care level, as doing so confers preferential reimbursement rates for subsequent tertiary hospital visits. Consequently, in areas where primary care institutions lack sufficient capacity to absorb the additional demand generated by the reform, the policy contributes to a more pronounced complementary effect, particularly in resource-rich areas where tertiary hospitals have sufficient capacity to accommodate upward patient flows. A further important finding concerns the moderating role of the outpatient reimbursement caps, which interacts closely with the induced hospitalization mechanism. Where the ceiling is relatively low, the reimbursable scope of outpatient care is constrained, leaving a portion of patients’ care needs inadequately addressed within the outpatient setting. Under these circumstances, both patients and providers face direct institutional incentives to shift care to the inpatient setting, where substantially higher reimbursement thresholds apply and a more comprehensive package of services can be covered. In such contexts, the reform is therefore more likely to intensify upward patient flows towards inpatient services. Conversely, where ceilings are set at a more generous level, patients are better positioned to have their care needs met within the outpatient framework, weakening the institutional incentive to substitute inpatient for outpatient care and thereby attenuating the induced hospitalization effect. Our findings suggest that the depth of outpatient reimbursement coverage is a critical policy lever: without sufficiently generous ceiling levels, outpatient insurance expansion risks inadvertently reinforcing, rather than reducing, inpatient demand. From a global perspective, research on the outpatient–inpatient relationship has predominantly focused on developed country settings, where expanded outpatient coverage is generally found to substitute for inpatient care (Elek et al. 2019). The Chinese experience documented in this study offers important insights for middle-income countries that also intend to pursue similar reform trajectories. While the original policy intent of expanding outpatient insurance coverage is to alleviate the pressure on inpatient care (Bindman et al. 1995), our findings suggest that an immature institutional environment can systematically reverse this expectation, turning an intended substitution effect into an unintended complementarity. Specifically, in contexts characterized by underdeveloped primary healthcare infrastructure and insufficiently generous outpatient reimbursement, primary care institutions are unable to fulfil their gatekeeping role; increased investment in outpatient services may instead unleash latent demand for medical care that flows upward to the inpatient system, further straining the inpatient system (Fortney et al. 2005, Zhu et al. 2025). Beyond demand-side factors, the outpatient–inpatient relationship is also critically shaped by the incentive structures facing healthcare providers. Where hospital payment systems such as fee-for-service or global budget prepayment with surplus-retention mechanisms may reward service volume expansion, hospitals have both the motivation and the capacity to absorb and amplify the additional patient demand generated by outpatient insurance expansion, further reinforcing complementary growth across care settings. Therefore, policymakers in middle-income countries considering outpatient insurance expansion should attend carefully to these institutional preconditions, particularly primary care capacity, the generosity of outpatient reimbursement ceilings, and provider payment incentive structures, to avoid unintended increases in inpatient utilization. Conclusion This study examined the impact of the RBMI-OP reform on hospital inpatient service utilization using county-level administrative data from Sichuan Province (2010–20) and CHARLS micro-level data (2011–18). The reform increased inpatient admissions by 18.06%, surgical cases by 23.12%, and the outpatient-to-inpatient conversion rate by 24.86%. This complementary effect operated through two pathways: an induced hospitalization effect arising from increased primary care outpatient utilization under limited primary care capacity and a reconfiguration of hospital services under the convergent pressures of global budget prepayment, performance assessment, and hierarchical healthcare mandates. The latter shifted selected diagnostic procedures from inpatient to outpatient settings, thereby increasing inpatient throughput. Heterogeneity analysis further showed that the complementary effect was more pronounced in regions with abundant medical resources and in areas with lower outpatient reimbursement caps. The complementary effect identified in this study should be interpreted with caution. On the one hand, the reform expanded the scope of outpatient reimbursement, promoted the utilization of primary care services, and concentrated hospital resources on severe and complex cases, thereby directly facilitating patient diversion in line with the hierarchical medical system. On the other hand, it remains unclear whether the observed increase in inpatient admissions was driven by genuine clinical need or by financial incentives associated with limited outpatient reimbursement caps. Rising inpatient volumes may also indicate a potential loss of allocative efficiency and place additional pressure on health insurance funds. The study has three policy implications. First, outpatient coverage and reimbursement rates for residents should be improved. The highest outpatient reimbursement cap in Sichuan Province (300 RMB, ∼42 USD, per year) is not commensurate with rising healthcare costs and may contribute to reimbursement-driven hospitalization. Second, improving the quality and service capacity of primary care systems is essential for reducing the siphoning effect of higher-level hospitals. Although differentiated reimbursement rates are designed to guide residents towards primary care, incomplete service facilities and weak service quality make it difficult to retain patients at this level. Strengthening the gatekeeping function of primary care institutions is therefore necessary. Third, more diversified financing mechanisms for medical insurance funds could help alleviate the dual expenditure pressure on both inpatient and outpatient services generated by the reform. Policy design should be tailored to regional economic conditions and residents’ healthcare needs, and implemented in a phased and measured manner. In addition, this study still has room for improvement. First, the study concerns potential sample selection bias. The analysis covers healthcare institutions in Sichuan Province only, and the findings should therefore be interpreted as suggestive evidence from a representative province, with a modest ambition of generalization. Second, the analysis is mainly conducted at the county and individual levels and lacks hospital-level data to directly observe how outpatient pooling reforms affect provider behaviour across different institutional levels. Third, although we find that the RBMI-OP reform increased residents’ use of outpatient services overall, we are unable to assess healthcare performance, particularly health outcomes. Fourth, while we document an increase in inpatient service utilization following the reform, the use of secondary administrative data prevents more detailed mechanism analysis, including distinguishing between necessary and unnecessary hospitalizations or identifying direct evidence of hospital behavioural change. Moreover, the patient behaviours collected from CHARLS data prevent analysis on the population under the age of 45. Future research using multi-province datasets, hospital-level data, and patient-level cost data is needed to address better precision and external validity. Supplementary Material