Implementing system dynamics in hospital services to improve operational efficiency: An empirical research study

Abstract

Objective

This study investigates operational efficiency in public general hospitals using a system dynamics (SD) model, focusing on resource allocation, patient flow and policy interventions. It explores the interactions between human resources, financial subsidies and patient visitation rates and their impact on hospital performance.

Methods

An SD model was developed to simulate various scenarios, incorporating data on hospital capacity, staffing and financial inputs, along with patient flow dynamics. Key performance metrics, including bed occupancy rate (BOR), average length of stay (ALOS), average cost per visit and workload index, were analysed under different policy scenarios.

Results

Simulation of four policy scenarios revealed that increased fiscal subsidies (scenario 2) consistently improved operational efficiency by reducing ALOS, staff workload and cost per visit. In contrast, scenarios involving human resource cuts or rapid patient growth triggered adverse feedback loops that undermined performance.

Conclusions

The SD model effectively captures the dynamic interactions within hospital operations and enables assessment of policy interventions over time. Enhanced government funding contributes most positively to efficiency, while demand surges and resource reductions introduce system strain and performance trade-offs.

Introduction

Efficient management of public hospitals is crucial for ensuring sustainable, high-quality healthcare services. However, the growing scale of healthcare institutions, while intended to achieve economies of scale and reduce patient waiting times, has simultaneously introduced greater operational complexity, complicating resource management, service delivery and financial sustainability [1, 2]. These challenges adversely affect key stakeholders, including patients (through reduced service quality), hospital administrators (through resource management difficulties) and policy-makers (through escalating healthcare expenditures). Addressing these systemic inefficiencies is therefore critical to enhancing healthcare delivery, improving patient satisfaction and ensuring the financial sustainability of public healthcare systems.

Traditional evaluation methods, such as data envelopment analysis (DEA) and regression models, often fail to capture the nonlinear, time-dependent interactions inherent in healthcare systems [3, 4]. These approaches typically assess static efficiency snapshots, neglecting the long-term dynamic feedback loops that profoundly influence hospital performance.

A growing body of research advocates for the use of system-based modelling approaches to better capture the complexity of healthcare environments [5]. Since the 1960s, various simulation methods – such as discrete event simulation (DES) and agent-based modelling (ABM) – have been applied to healthcare problems [6, 7]. Although these methods offer valuable insights, they are often constrained by a micro-level focus, modelling individual processes or events without adequately reflecting the broader systemic interactions across departments and time horizons [8]. Moreover, DES models often rely on highly granular, process-specific data and require specialized technical expertise, making them difficult to build, validate and generalize in the context of large, complex hospital systems [9, 10].

System dynamics (SD) modelling provides a distinct advantage in this context by offering a macroscopic, feedback-oriented view of hospital systems [11]. SD captures the cumulative and delayed effects of decisions, facilitates the examination of policy interventions over long periods, and supports the understanding of emergent behaviours [12]. While SD has been widely used in areas such as disease modelling and healthcare supply chains [13, 14], its application to systematically evaluate operational efficiency, particularly concerning resource allocation and policy innovation in large general hospitals, remains limited [15]. This gap is critical, especially in public hospital systems facing rising demand, budget constraints and pressure to maintain service quality under resource scarcity.

To evaluate hospital operational efficiency under complex and dynamic conditions, this study employs an SD modelling framework. The model captures the interdependencies among resource allocation, patient flow, financial inputs, and service outcomes over time. By simulating different policy scenarios, it provides insights into optimizing hospital performance and supports evidence-based decision-making for healthcare administrators.

Specifically, this study contributes to the literature by:

• (1) Constructing an operational efficiency evaluation framework tailored to the complexities of large public hospitals

• (2) Validating the model against multi-year historical data to ensure realistic representation

• (3) Using scenario analysis to explore the long-term impacts of different policy levers (e.g. resource reductions, subsidy increases, patient inflow growth)

• (4) Highlighting the dominating feedback loops that drive hospital system behaviour under varying conditions

Unlike prior SD applications that primarily focus on disease transmission or micro-level resource bottlenecks, this study provides a macro-integrated efficiency evaluation framework validated by real-world hospital data and multi-level policy simulations. Our findings offer both theoretical and practical value by advancing SD applications in healthcare management research and providing actionable insights for hospital administrators seeking to optimize resource utilization, improve patient services and enhance system resilience.

Literature review

Evolution of SD in healthcare modelling

SD was first developed as a method to understand the nonlinear behaviour of complex systems over time. Initially applied to industrial and corporate systems, SD soon gained traction in social and public policy fields owing to its ability to model feedback loops, time delays and accumulations (stocks and flows) inherent in large-scale systems [16]. The foundational principles of SD involve causal loop diagrams, stock-flow structures and simulation to understand how policy decisions interact with system structures to produce intended or unintended consequences [17]. These characteristics are particularly well suited to the healthcare domain, where long-term impacts, nonlinearities and dynamic interdependencies are pervasive.

Early applications of SD in healthcare primarily focused on the spread and control of infectious diseases, population health dynamics and epidemic models [15]. Over time, the focus broadened to encompass healthcare delivery, resource management and health policy analysis. For instance, SD models have been used to examine chronic disease progression, the effect of preventive health interventions and the allocation of healthcare workforce and infrastructure [14]. This transition reflects a shift from micro-level clinical modelling to system-level operational and strategic decision-making. In the last two decades, there has been a notable increase in SD studies that target hospital performance, patient flow and service delivery redesign as healthcare systems face mounting cost pressures and quality demands [5].

Compared with discrete event simulation (DES) and agent-based modelling (ABM), SD offers a macro-level, aggregate perspective that is particularly effective for long-term strategic planning and policy analysis [11]. DES is typically used for modelling operational details such as queue management or scheduling within specific hospital departments, and ABM is ideal for capturing heterogeneous agent behaviours in localized environments. In contrast, SD focuses on the interplay between key structural elements of a system over extended time horizons, making it better suited for examining cumulative policy effects, feedback mechanisms and resource sustainability [18, 19]. Moreover, SD models are generally less data-intensive at the micro level and can integrate expert knowledge, historical data trends and policy assumptions into comprehensive system representations [20]. This makes them more adaptable for use in complex, data-limited environments such as public hospitals.

SD in health system planning and policy evaluation

SD has been widely applied in national and regional health system planning and policy evaluation. Leerapan et al. (2021) employed SD to project Thailand’s physician workforce needs under Universal Health Coverage, revealing how hospital-centric care models exacerbate shortages unless shifted towards primary care [21]. Similarly, Rafiei et al. (2018) modelled Iran’s neurosurgeon gap, attributing maldistribution – not absolute shortages – as the critical issue [22]. These studies underscore SD’s capacity to simulate long-term workforce imbalances and test policy interventions such as training expansions or incentive realignments.

SD also informs service delivery restructuring and expenditure optimization. Peng et al. [23] developed a whole-system model for cardiovascular disease (CVD) in Australia, demonstrating that environmental interventions (e.g. promoting healthy lifestyles) yielded greater reductions in hospitalizations and deaths than acute care improvements alone [23]. Their findings highlighted trade-offs between immediate clinical outcomes and sustained population health gains – a nuance often missed by static models. Meanwhile, Occhipinti et al. (2021) applied SD to mental health policy in New South Wales, identifying misaligned federal and state priorities as a barrier to suicide prevention, thus advocating for integrated funding pools [24]. Such models integrate epidemiological, economic and behavioural data to evaluate policy resilience against demographic shifts or external shocks.

Public health interventions increasingly leverage SD to anticipate unintended consequences. Currie et al. (2018) reviewed environmental health policies, noting SD’s underutilization despite its potential to clarify micro- and macro-level decision-making trade-offs [25]. In infectious disease control, Tian et al. (2024) combined SD with agent-based modelling to optimize coronavirus disease 2019 (COVID-19) testing and contact tracing in Saskatchewan, revealing how early, decentralized interventions reduced hospital surges [26]. These applications demonstrate SD’s versatility in bridging clinical, operational and societal dimensions – enabling policy-makers to balance efficiency, equity and sustainability.

Applications of SD in hospital operations and resource management

At the hospital level, SD models excel in unravelling feedback loops and delays that disrupt patient flow and resource allocation. Lam et al. (2022) designed an agile SD framework for bed management during Singapore’s COVID-19 surges, rapidly adapting models to evolving policies and occupancy rates [27]. Their approach underscored SD’s real-time utility in crisis response, where iterative simulations informed capacity expansions and staff redeployments. Similarly, Catsis et al. (2023) modelled NHS cardiovascular waiting lists post pandemic, exposing bottlenecks in referral systems and outpatient services [28]. By simulating so-called what-if scenarios (e.g. increasing consultant appointments), they revealed counterintuitive effects, such as downstream disengagement due to prolonged wait times – a dynamic invisible to linear analyses.

Staffing and workflow optimization are other critical foci. Shire et al. (2020) used participatory SD with hospital pharmacists to map interruptions, fatigue and safety trade-offs, demonstrating how workload spikes propagate through rework cycles [29]. Their interactive dashboard enabled stakeholders to visualize how staffing adjustments impacted productivity, fostering consensus on operational changes. Likewise, Zimmerman et al. (2016) applied group model-building to VA mental health services, aligning front-line staff and administrators on evidence-based therapy implementation barriers [30]. These studies highlight SD’s role in translating localized inefficiencies into systemic solutions, emphasizing human factors alongside structural constraints.

Evaluating hospital performance and limitations of SD applications

While existing studies have demonstrated SD’s operational utility within hospital settings, particularly in patient flow, workforce and resource allocation, there is growing interest in evaluating how effectively SD models capture broader institutional performance. Furthermore, as SD is increasingly adopted for strategic and policy-level hospital decision-making, it is necessary to examine its methodological robustness and practical limitations in such contexts.

By capturing complex feedback loops and time delays inherent in healthcare systems, SD models provide valuable insights into the dynamic interactions among various hospital components. For instance, Supriyanto and Suryani (2019) developed an SD model to optimize inpatient room utilization, focusing on key performance indicators such as bed occupancy rate (BOR), length of stay (LOS), turn over interval (TOI) and bed turn over (BTO). Their simulation demonstrated that strategic adjustments could lead to optimal utilization of inpatient services, thereby enhancing overall hospital efficiency [31]. Similarly, Teymourifar and Trindade (2024) introduced the concept of dynamic resectorization to balance regional healthcare systems. Their SD-based approach aimed to redistribute healthcare resources effectively across multiple hospitals, considering factors such as service quality and accessibility. The simulation outcomes indicated that periodic resectorization could significantly improve the utility of healthcare systems by aligning capacity with patient demand [32]. Furthermore, Goldsmith and Siegel (2011) explored the integration of Health Information Systems (HIS) with SD modelling to enhance hospital operations. Their research highlighted how combining SD with HIS data could identify inefficiencies in hospital processes, leading to informed decision-making and improved patient safety [33]. These studies underscore the efficacy of SD modelling in capturing the intricate dynamics of hospital operations, enabling healthcare administrators to predict the outcomes of policy changes, optimize resource allocation and ultimately improve service quality and efficiency.

In summary, while existing SD applications in healthcare have addressed specific operational challenges such as patient flow, staffing and infrastructure bottlenecks, few studies have developed comprehensive models that simulate hospital operational efficiency under multiple policy interventions. This research aims to bridge that gap by integrating diverse management levers into a unified simulation framework that captures long-term, system-wide effects. In doing so, it extends the scope of SD applications from isolated performance factors to holistic, policy-oriented efficiency evaluations.

Methods

Research design

The primary objective of the model is to evaluate the long-term impacts of key hospital policy levers – such as fiscal subsidies, human resource adjustments and demand shocks – on operational efficiency. Specifically, the model aims to support evidence-based decision-making by simulating resource-performance trade-offs under dynamic conditions.

To design and evaluate effective strategies for improving operational efficiency in hospital services, we employed a mixed-methods approach that integrates qualitative insights with quantitative simulation modelling. The overall research process consisted of two main components: (1) stakeholder-engaged model building and (2) SD simulation for scenario testing and policy evaluation.

The qualitative component involved a series of participatory workshops with hospital administrators, front-line medical staff and operations managers. These workshops aimed to identify key inefficiencies and feedback structures within existing service delivery processes. Through facilitated group model building (GMB), we developed a shared understanding of the dynamic interactions among patient flow, staffing, resource allocation and service quality.

Building upon the qualitative system conceptualization, we constructed a stock-and-flow SD model that encapsulated the core operational processes of inpatient care, including admissions, discharges, bed occupancy and personnel deployment. The quantitative component was designed to simulate various policy scenarios, such as staffing adjustments, process redesign or capacity expansion, using both baseline hospital data and expert-informed parameter estimates.

To ground the model in a real-world context, we selected one of the largest public general hospitals in Hainan Province as the focal case. This hospital ranks within the top 10% in the national performance evaluation of tertiary public general hospitals, making it a representative and robust site for policy simulation and strategic assessment.

The integrated workflow of this study is shown in Fig. 1, demonstrating how stakeholder-derived system structures were translated into a quantitative simulation framework.

Fig. 1.

Workflow of the mixed-methods research design for SD-based hospital efficiency improvement

SD model

In essence, the development of the SD model in this study integrates both qualitative and quantitative methodologies to examine hospital operational efficiency. The initial phase of modelling involves a systematic definition of the research objectives and boundaries, followed by the identification of key system components – namely, human, financial and physical resources – which are critically associated with healthcare service provision and operational performance [34–36].

Drawing upon the foundational principles of SD modelling [37, 38], we first mapped the essential components and their interdependencies using a causal-loop diagram (CLD). This qualitative modelling step enables the identification of feedback structures – both reinforcing and balancing loops – that characterize dynamic system behaviour. The CLD serves as a conceptual framework, illustrating how factors such as bed occupancy, average length of stay (ALOS), workload and treatment costs influence hospital efficiency over time.

Following this, the model was translated into a stock-flow diagram (SFD), wherein variables were classified as stocks, flows or auxiliaries [39]. Stocks, such as total bed capacity or cumulative patient volume, are defined by their ability to accumulate over time and be measured at discrete intervals. Flows represent the rates of change in these stocks (e.g. admission or discharge rates), while auxiliary parameters support the computation of interactions among variables.

The final stage of model development involved the quantification of the SFD by assigning initial values to stocks and formulating mathematical relationships among all model components. This resulted in a system of equations representing the dynamic evolution of hospital operations. The simulation was then performed to explore the impact of selected policy and management levers on efficiency outcomes [36, 39]. To ensure model relevance, we selected parameters that: (1) demonstrate a high degree of interconnection with other system components and (2) directly influence hospital performance. The following core indicators were incorporated:

BOR: Measures the proportion of occupied hospital beds over a given period [40, 41]. It reflects patient turnover and resource utilization efficiency.

ALOS: Indicates the median number of days a patient spends in the hospital [41]. It is widely recognized as a proxy for inpatient care efficiency and plays a critical role in determining bed turnover, patient flow and overall operational capacity, which are essential components in system dynamics modelling of hospital performance.

Average Cost per Visit: Captures the mean expenditure per outpatient or inpatient visit, shaped by service intensity, clinical efficiency and financial management [42–44].

Index of Workload: Represents the clinical burden on healthcare staff, serving as a proxy for operational stress and care quality [45].

Model boundaries

As a foundational step in SD modelling, the definition of model boundaries is directly aligned with the objective of this study: to explore how large public hospitals respond to internal and policy-induced pressures in optimizing operational efficiency [46]. Unlike outpatient services, which typically involve shorter care cycles and decentralized decision-making, inpatient operations represent a highly structured and resource-intensive process. Thus, the model specifically focuses on inpatient care delivery, where management decisions regarding resource allocation, patient throughput, and financial performance are tightly interlinked and amenable to simulation.

To build the SD model for the given case study, we chose the parameters on the basis of the following: (1) patient flow dynamics, which track admission, length of stay and discharge behaviour [47]; (2) resource deployment, encompassing hospital beds, clinical staffing and fiscal subsidies; and (3) financial operations, which integrate operating income, personnel cost and cost–efficiency targets [48]. These parameters reflect the hospital’s endogenous mechanisms for achieving service continuity under constrained resources.

Key stakeholders included in the model are hospital administrators, medical staff and policy actors. Given the emphasis on hospital-internal efficiency, external patient referral systems and inter-hospital transfers were excluded. Moreover, variables with minimal variance during the simulation period, such as high-value fixed assets and specialist training cycles, were omitted to streamline model complexity.

The model assumes that hospital management seeks to optimize efficiency through cost–effective resource deployment while maintaining service quality. These boundaries ensure a focused yet comprehensive simulation of the feedback structures governing hospital operational behaviour.

Feedback structure and causal loop diagram

Following the delineation of system boundaries, we conducted a comprehensive analysis of the internal causal relationships among the variables enclosed within the defined scope. As a complex and dynamic system, hospital inpatient operations are influenced by the continuous interplay of patient flow, staffing levels, financial inputs and service quality. This interdependence lends itself well to causal-loop diagramming, which visually represents the feedback mechanisms governing system behaviour.

On the basis of the stakeholder-engaged group model building (GMB) sessions described in Section 3.1, we identified key factors that influence operational efficiency. These include BOR, ALOS, index of staff workload, patient admission volume, treatment costs and financial subsidies. The relationships among these factors were then mapped into a causal-loop diagram (CLD), where arrows indicate causal direction, and polarities (“+” for positive, “–” for negative) denote the nature of the causal influence.

As shown in Fig. 2, several reinforcing and balancing feedback loops were identified. For example, one key reinforcing loop (R) illustrates how an increase in patient admissions raises bed occupancy, thereby enhancing hospital revenue and enabling greater investment in staff or beds – thus further supporting service capacity and patient intake. Conversely, a critical balancing loop (B) shows how increased workload can degrade service quality, leading to prolonged ALOS and reduced system throughput.

Fig. 2.

Causal loop diagram. ALOS average length of stay, BOR bed occupancy rate, CPV cost per visit, PNR/PDR patient–nurse ratio/patient–doctor ratio

Stock-flow diagram

Expanding on the causal relationships discussed earlier, the stock-flow model constructed for evaluating the efficiency and quality of public general hospitals encapsulates the intricate dynamics of the hospital system by integrating essential stocks, flows and auxiliary variables. This model not only delineates the core resources but also captures the complexities of their interactions and how they collectively drive hospital performance.

Figure 3 shows a schematic diagram of the hospital operational system and the interrelationships between key variables. The system is centred around the hospital’s revenue, which is influenced by various internal processes and feedback loops. These include the hospital’s service delivery, patient inflow and resource management. Over time, these dynamic variables interact to determine the overall performance and financial sustainability of the hospital [37].

Fig. 3.

Representation of the hospital operational system

Central to the model are the stocks [49, 50], which represent critical resources such as the number of staff, expenses and revenues – each being a reservoir that is continuously altered by various flow variables. These flow variables, which include rates of staff recruitment and attrition as well as revenue generation and expenditure, are the driving forces that either replenish or deplete the stocks [51], thereby reflecting the hospital’s capacity to sustain its operations and maintain service quality. Complementing these flows are auxiliary variables that, although not directly altering the stocks, significantly influence the behaviour of the system by modulating the flow rates.

The stock-flow diagram, shown in Fig. 4, visually encapsulates these dynamic interactions, providing a nuanced understanding of how each component within the hospital system interrelates and influences overall performance. This model serves as a sophisticated tool for analysing how targeted interventions – whether in staffing, financial management or quality improvement – can lead to cascading effects across the entire system, thereby enabling more effective decision-making aimed at optimizing both efficiency and service quality in public general hospitals.

Fig. 4.

SD model of the hospital operational system

SD model parametrisation

The parametrisation of the SD model was achieved through a methodical process that included comprehensive estimation, a thorough literature review and iterative experimental adjustments. The SD equations were developed by integrating established methodologies from relevant studies and adapting them to the specific context of this research. For variables where precise values were challenging to establish, optimal parameters were determined through continuous adjustment during model calibration. Given the lack of unified dimensional units for the model’s variables, a dimensionless normalization approach was employed, scaling key factors that influence efficiency values in a 0–1 range to ensure consistency and comparability [52].

In this study, the factors influencing hospital operational efficiency exhibit varying degrees of impact and are embedded within complex causal feedback structures. Therefore, assigning appropriate weights to key variables is essential to enhance the structural rigour and simulation credibility of the SD model. Existing weighting approaches are broadly categorized into subjective and objective methods [53]. Subjective methods incorporate expert knowledge – typical techniques include the analytic hierarchy process (AHP) and the Delphi method – while objective methods rely on the intrinsic properties of data, such as entropy weighting and CRITIC, and are more suitable for large-scale, highly structured datasets [54, 55].

Considering that several key variables – such as workload intensity, perceived quality of care and subsidy efficiency – lack robust empirical baselines, this study adopted AHP to incorporate expert insight into the weighting process. A multidisciplinary expert panel comprising hospital administrators, policy-makers and front-line clinicians was convened to perform pairwise comparisons of core indicators. The resulting consistency-verified judgment matrix was used to derive relative weights, which were subsequently embedded into the operational efficiency formula.

The detailed model equations and their corresponding explanations are presented in Table 1.

Table 1.

Equation design and explanation of the SD model

Equation design	Explanation
$\text{Operational}e\text{fficiency}=\frac{\left({\text{w}}_1,\times ,\text{BOR},+,{\text{w}}_2,\times ,\text{Workload Index},\_,\text{normalized}\right)}{\left({\text{w}}_3,\times ,\text{ALOS},\_,\text{normalized},+,{\text{w}}_4,\times ,\text{Average Cost},\_,\text{normalized}\right)}$	Efficiency is defined as the weighted ratio of output (bed utilization and service volume capacity) to input (length of stay and average cost). Workload index appears only on the output side to reflect institutional capacity, thus avoiding conceptual redundancy and ensuring interpretability
$\text{BOR}=\frac{\text{Inpatient Reserve}}{\text{Total Actual Bed}-\text{Days Available}}$	Bed occupancy rate, reflecting bed utilization efficiency
$\text{Workload Index}=\frac{\text{Total Patient Visits}}{\text{Total Medical Staff}}$	Indicator of per-staff patient load, reflecting workload intensity
$\text{ALOS}=\frac{\text{Inpatient Days}}{\text{Inpatient Visits}}$	Average length of stay, measuring patient flow efficiency
$\text{Average cost}=\frac{\text{Reserve}}{\text{Number of Patient Visits}}$	Mean cost per patient, combining outpatient and inpatient costs
$\text{Revenue}=\int [(\text{Reserve}-\text{Expense})]\text{dt}$	Hospital revenue accumulation over time based on reserve dynamics
$\text{Reserve}=f(\text{Revenue},\text{Outpatient volume},\text{Government subsidy})$	Dynamic reserve pool driven by revenue, outpatient flow and subsidy
$\text{Number of outpatient visits}=\int [\text{Change of outpatient}+\text{Growth term}]\text{dt}$	Cumulative outpatient volume influenced by revenue-linked growth
$\text{Total medical staff}=\text{Number of doctors}+\text{Number of nurses}-\text{adjustment}$	Total staffing adjusted for cost-feedback effects

Normalization constants correspond to approximate upper bounds of workload index, ALOS and average cost observed during the baseline year (2018), ensuring all inputs and outputs are scaled comparably within the range of 0–1

Results

Model validation

Model validity refers to the capability of the model to accurately replicate the real-world system it represents, and it is essential for all simulation models to undergo rigorous validation testing [56]. Numerous methodologies exist for validating SD models. In this study, we selected specific historical data from 2018 to 2023 to directly compare with the simulation outcomes. Four variables with complete data over the 3-year period were chosen for comprehensive model validation: revenue, BOR, ALOS and outpatient visits. According to previous studies, a relative error margin of 5% is generally considered acceptable [57, 58]. As presented in Table 2, the average error remained below 5%, indicating that the model is functioning well and meets the required standards. Therefore, the model is considered effective and suitable for trend simulation and analysis.

Table 2.

Model validation: relative error between simulation results and historical data

Year	Medical service revenue (%)	BOR (%)	Outpatient visits (%)	ALOS (%)
2018	3.27	0.77	1.35	2.32
2019	2.02	4.13	4.20	0.08
2020	4.68	0.81	4.25	1.09
2021	3.25	0.24	1.40	2.61
2022	2.81	2.89	3.08	1.72
2023	1.98	4.77	1.45	2.14
Mean	3.00	2.27	2.62	1.66

Intervention scenarios

In assessing the operational dynamics of large public general hospitals, this study integrates an SD model to simulate the impact of strategic resource management under four distinct intervention scenarios. These scenarios are designed to explore the implications of policy changes and resource allocation on hospital efficiency and service delivery [35], enabling policy-makers and hospital administrators to make informed decisions.

Scenario 0 serves as the control or baseline, providing a foundational comparison for the impacts observed in other scenarios. Scenario 1 simulates an atypical policy adjustment, as hospitals generally avoid reducing human resource budgets. However, under ongoing public hospital reform in China, some institutions have experienced wage cuts or delayed payments to medical staff owing to financial constraints [59]. This scenario models a 10% annual reduction in the share of human resource spending, aiming to reflect such reform-driven pressures and examine their potential impacts on service delivery and staff workload. Scenario 2: Enhanced governmental fiscal allocations possess the potential to mitigate the reliance of public hospitals on conventional revenue channels, thereby curtailing the prevalence of excessive medical treatments. In light of this, certain researchers propose the institution of a compensatory doubling mechanism for public hospitals [60]. Given the existing policy framework and the historical trajectory of financial subsidies provided to case-study hospitals, this analysis proceeds to model a scenario wherein the velocity of subsidy augmentation is projected to double. Scenario 3 simulates an accelerated increase in patient visits, with an added 10% annual growth beyond current trends. This scenario reflects the rising healthcare demand driven by Hainan’s Health Island and Free Trade Port development [61], testing the hospital’s capacity to respond to increased service pressure.

Figures 5, 6, 7, 8 and 9 depict key metrics for evaluating hospital performance under different scenarios from 2018 to 2035. Each scenario models a specific intervention – ranging from human resource reductions to financial subsidy increases – and their impacts on operational outcomes.

Fig. 5.

Simulation result of ALOS

Fig. 6.

Simulation result of BOR

Fig. 7.

Simulation result of workload index

Fig. 8.

Simulation result of average cost

Fig. 9.

Simulation result of operational efficiency

ALOS

As shown in Fig. 5, ALOS in all four scenarios initially declines and then rises. Scenario 3, simulating rapid patient growth, leads to the highest ALOS over time, exceeding the baseline (scenario 0), suggesting demand-induced strain on hospital capacity. Scenario 1, with reduced staffing costs, maintains a lower ALOS than the baseline but trends upwards later, indicating potential service trade-offs. Scenario 2 achieves the lowest ALOS throughout, showing that increased fiscal support can sustain efficiency and ease inpatient pressure.

BOR

As shown in Fig. 6, BOR fluctuates over time but exhibits a general trend of stabilization after an initial period of variation. Scenario 0 (baseline) shows moderate fluctuations and a gradual decline over the long term. Scenario 3, which simulates accelerated patient demand, results in the lowest BOR among all scenarios, falling below the baseline. This suggests that the existing resource expansion in scenario 3 may not keep pace with the increasing demand, leading to lower bed utilization due to capacity bottlenecks or system strain. In contrast, both scenario 1 (reduced human resources [HR] expenditure) and scenario 2 (enhanced fiscal subsidies) result in consistently higher BOR levels compared with the baseline. Notably, scenario 2 yields the highest BOR throughout most of the simulation period, indicating that increased government investment may enable hospitals to better utilize inpatient resources by maintaining service continuity and operational stability.

Workload index

The workload index displays minor fluctuations in the early simulation period, followed by a gradual upwards trend across all scenarios. Among them, scenario 3 exhibits the highest workload index consistently, reflecting the intensified patient inflow and resulting strain on medical staff. Scenario 0, which represents the baseline, ranks second, indicating steady pressure under unchanged resource conditions. Scenario 1 shows a slightly lower workload due to staffing cost reductions, which may moderate personnel expansion. Scenario 2 yields the lowest workload index, suggesting that increased fiscal support enables hospitals to better match resource availability with service demand, thus alleviating staff burden over time (Fig. 7).

Average cost per visit

As shown in Fig. 8, average cost exhibits initial fluctuations followed by a gradual downwards trend across all scenarios. Scenario 3 results in the highest average cost throughout most of the simulation period, likely due to increased service demand and potential inefficiencies under resource strain. Scenario 0 follows, reflecting cost patterns under a static resource environment without policy intervention. Scenario 1, involving reductions in human resource expenditure, shows slightly lower average costs, potentially driven by constrained staffing-related expenses. Scenario 2 yields the lowest average cost, indicating that enhanced fiscal subsidies may help streamline hospital operations and reduce per-service expenditure through improved resource alignment and service delivery efficiency.

Operational efficiency

As shown in Fig. 9, operational efficiency initially fluctuates and then diverges across scenarios. Scenario 2 achieves the highest efficiency overall, reflecting the positive impact of increased fiscal subsidies. Scenarios 1 and 0 perform similarly, indicating that modest cost adjustments yield limited improvement over the baseline. In contrast, scenario 3 shows a clear decline in efficiency over time, consistent with the model structure where rising ALOS-driven by sustained patient inflow – reduces performance.

Discussion

This study aimed to evaluate the operational impacts of key hospital policy interventions through an SD model. In this section, we interpret the simulation outcomes across different scenarios, linking observed trends to underlying feedback structures and performance mechanisms. The findings are discussed in light of prior research and their practical implications for public hospital management.

In scenario 1 (reduction in human resources), efficiency initially improves slightly but soon stabilizes at a modest level. Although reduced staffing alleviates financial burden, it activates a reinforcing loop of declining care quality and staff satisfaction, as identified in previous SD models of hospital resilience [38]. This erosion in workforce stability ultimately limits throughput, as reflected in the modest performance gains relative to the baseline. Although cutting the human resources budget may initially seem to ease financial concerns, this approach generates long-term resistance, not only economically but also in terms of care quality and treatment quotas [62]. As quality deteriorates, pathways to success disappear, and a self-reinforcing negative cycle emerges, which becomes increasingly difficult to reverse once established [62].

Scenario 2 (increased financial subsidies) showed that enhanced financial resources could stabilize operations and facilitate improvements. Research by Songul and Cemal [63] noted that applying systematic management practices enhances financial performance and operational efficiency, which is consistent with the positive impacts of increased financial subsidies on hospital sustainability. Another study found that, despite instances of misuse or waste of government subsidies by many healthcare institutions in South Korea, direct subsidies (such as cash and medical equipment) and indirect subsidies (such as tax exemptions) remain effective in supporting the management performance of public hospitals [64]. When the government assumes a core role in both the financing and provision of public healthcare services, particularly when it constitutes a substantial portion of the revenue, it is crucial to consider the implementation of incentive mechanisms to drive efficiency and enhance overall performance [65, 66].

Scenario 3 (accelerated increase in patient visits) demonstrated the healthcare system’s ability to respond to rising demand, though initial operational challenges were evident. This underscores the need for flexible resource management, consistent with views emphasizing the role of adaptive performance and organizational support in addressing fluctuating demands in healthcare [67]. Similarly, the study by Holden et al. [68] on agile innovation in healthcare also stresses the importance of adaptive processes in handling complex, evolving patient care environments. These insights mirror the challenges faced by real-world public hospitals, where managing unpredictable patient volumes requires efficient allocation of limited resources.

Conclusions

By simulating the impacts of fiscal subsidies, staffing adjustments and patient inflow dynamics, the study offers a dynamic view of hospital behaviour over time – an area where traditional static methods fall short. Unlike prior SD applications that have focused primarily on infection control or patient flows, this study integrates financial, workforce and performance indicators into a unified efficiency framework.

However, the model was calibrated using data from a single large public general hospital in China, which may limit its contextual specificity and generalizability. Healthcare systems differ widely in their administrative structures, financing mechanisms and demand patterns; thus, direct application of the findings to other hospitals should be approached with caution. However, the modelling framework – particularly its integration of fiscal, workforce and performance dynamics – can be adapted to other public healthcare settings facing similar resource constraints or undergoing system-level reforms. Future studies could validate and refine the model using multi-site data, or tailor its structure to reflect institutional heterogeneity, thereby enhancing its external validity and policy relevance across diverse healthcare systems.

Implications for practice

This study offers a replicable modelling framework for public hospital administrators and policy-makers to evaluate operational strategies under resource constraints. The findings support evidence-based decisions aimed at sustaining efficiency, workforce balance and service quality in complex healthcare environments.

Abbreviations

SD

System dynamics

BOR

Bed occupancy rate

ALOS

Average length of stay

CPV

Cost per visit

PNR/PDR

Patient–nurse ratio/patient–doctor ratio

GMB

Group model building

AHP

Analytic hierarchy process

Author contributions

Each author significantly contributed to the conception and development of this research. W.L. conceived the study, designed the overall research framework and supervised the entire project. He contributed to the analysis and interpretation of data and was involved in drafting and revising the manuscript. H.X.H. developed the system dynamics model, contributed to the study design and provided input on result interpretation. She assisted in drafting and revising the manuscript. X.J.H. reviewed the model structure for accuracy and relevance, contributed to data analysis and provided critical feedback on the manuscript. H.C. was responsible for data acquisition and preprocessing, executed the model simulations and contributed to drafting the manuscript. W.W. supported the simulation process and assisted in manuscript drafting and revision. Z.H.Z. contributed to model refinement, coordinated expert consultation for the AHP component and assisted in formatting and final checks during the revision process. All authors reviewed the manuscript.

Funding

This research was supported by the following grants: National Natural Science Foundation of China, grant no. 72204069. National Natural Science Foundation of China, grant no. 72464012.

Academic Enhancement Support Program of Hainan Medical University, grant no. XSTS2025071.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable. This study was entirely an analysis of data from published secondary sources, and no specific human subjects were involved.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.