The PPRR (Prevention, Preparedness, Response, and Recovery) theory is a crucial theoretical framework in the field of emergency management (32). Originating in the United States, it has gained widespread adoption in research related to health and public health emergencies in recent years (33). This theoretical model encompasses four interconnected stages: Prevention, Preparedness, Response, and Recovery. These stages form a continuous loop through their interactions, effectively supporting the digital and intelligent processes involved in the prevention and management of public health emergencies.

Prevention: This stage involves implementing measures to prevent the occurrence of adverse events. The goal is to protect people’s lives, health, and property by reducing or eliminating risk factors before an event can take place.

Preparedness: The focus of this stage is ensuring that the resources, capabilities, and systems necessary to respond to public health incidents or disasters are in place and readily available. The objective is to enhance the efficiency and effectiveness of emergency responses and to minimize confusion and panic when incidents occur.

Response: In this stage, immediate actions are taken when a public health emergency arises to mitigate its adverse effects and protect lives and property. The purpose is to control the development of the incident and minimize its impact.

Recovery: This phase involves the process of restoring society, the economy, the environment, and public life to normal or improved conditions once the impact of the public health emergency has been largely controlled. The goal is to derive lessons from the emergency to improve the overall emergency management capabilities of the health system in the future.

The concept of the digital twin was first introduced by Professor Grieves for product lifecycle management (34, 35). It refers to the accurate virtual representation of physical objects, processes, or systems within a digital space (36). This high-level simulation integrates historical data and real-time sensor data, encompassing both the physical and functional characteristics of real-world models (37, 38). Digital twins serve as in silico copies of physical entities, utilizing multimodal data to build models that optimize decision-making and predict the effects of interventions on outcomes. This operable data connection system between the virtual and physical realms collects real-time data from physical objects and transmits it to their digital counterparts for processing and analysis, thereby providing precise analytical insights and simulating potential scenarios (39). The digital twin transforms the performance and condition of real-world objects in the physical environment. Initially, the three-dimensional model of a digital twin consisted of physical products, virtual representations, and their interconnections (40). As digital technologies advanced and the concept of digital twins gained traction among experts and scholars, the five-dimensional model of digital twins was proposed (41). This expanded model includes physical entities, virtual models, services, twin data, and the connections between these components. The core idea of this model is to establish real-time interactive feedback between physical entities and their digital counterparts through data transmission, enabling real-time monitoring, simulation analysis, and predictive optimization. This allows for the full-cycle monitoring, management, and optimization of an object or system.

Digital twins fulfill the need for real-time, two-way connections and interaction by integrating various information technologies such as the Internet of Things (IoT), big data, artificial intelligence (AI), and deep learning (42). IoT is a key technology to strengthen digital transformation, and it involves a broad system of interconnected “things.” IoT collect and link data collected by sensors and other devices, using edge computing to perform preliminary processing close to the site, and performing aggregation and recalculation in the cloud (43). This determines the freshness and scalability of data-to-model and is the technical basis for the real-time nature of digital twins. Blockchain is a decentralized and growing chain of records, which uses distributed ledgers to record the “who used it, when and why” of data, and has the characteristics of non-tampering and traceability (43, 44). These technologies enable both virtual and physical realms to interact seamlessly, creating a robust platform for real-time data exchange. System-level and complex system-level digital twins can dynamically address diverse computing, storage, and operational requirements through on-demand cloud computing services. Additionally, 5G communication technology, with its high speed, large capacity, and low latency, supports the data transmission and equipment interconnection essential for digital twins (45). This facilitates real-time data exchange between virtual models and physical entities, thereby accelerating the application of digital twin technology. Twin data integrates multi-source, multi-type, and multi-structured data, often encompassing vast volumes of information (46, 47). To manage big data effectively, a comprehensive database can be established to extract valuable insights that explain and predict the behaviors and outcomes of the physical world (48). Blockchain technology further strengthens the reliability and security of digital twins by ensuring data integrity and providing traceability throughout the process. The trust mechanism enabled by blockchain ensures that users can confidently access and utilize various services (49). Furthermore, artificial intelligence enhances the value of twin data by automatically conducting data analysis and deep learning, utilizing optimal algorithms to improve the responsiveness and accuracy of services.

Although digital twin is recognized as an advanced simulation architecture in the field of engineering technology, its development potential in the field of healthcare cannot be ignored. Due to the variety of ways to obtain patient data, the incomplete mechanism model of disease pathophysiology, and the interference of factors outside biological constraints, the “passive” data collection of traditional digital medicine is based on the data of a large number of retrospectively scanned models to predict future health conditions, while lacking “active” feedback from patients. Digital twins can and leverage continuous, real-time data streams to reactively update predictions, providing an opportunity to broaden the scope of medical digital twins.

In recent years, with the development of personalized and precision medicine, the application of digital twins in the healthcare sector has garnered increasing attention (23). Researchers can create a virtual replica of a patient or an organ in a virtual space, using various Internet technologies to construct a digital twin and perform planned operations on the replica, observing the real-time feedback to optimize the treatment process and select the most effective treatment methods. Figure 1 depicted the construction process and application diagram of digital twins in the medical field. The process of digital twin construction included data perception and collection, model construction, simulation exercises and optimization in the actual treatment process. Considering the characteristics of the digital twin itself, its construction process and application scenarios collect patient data such as diagnosis, treatment, clinical examination parameters, and clinical symptoms in physical space. After the calculation and optimization of the big data model, real-time optimization schemes and decision-making suggestions were provided for doctors’ clinical decisions, patients’ clinical pathways, basic research and mechanism model construction in digital space. Table 1 summarizes the typical applications of digital twin technology in the medical field in recent years.

The application of digital twins in cardiovascular system primarily involves disease prediction, personalized treatment for cardiovascular diseases, treatment visualization, and planning (50–52). Cardiovascular digital twins enable close monitoring of an individual’s cardiovascular health by integrating clinical data, genetic information, and continuously monitored physiological and biochemical data. By combining sensor data with AI algorithms, a digital model of the heart can be constructed. This model can simulate potential cardiovascular issues and medication effects by applying various interference factors. It allows for accurate tracking of both the physical heart and its virtual counterpart, evaluating the health status, and continuously monitoring and providing feedback continuously (53). Hwang et al. created a digital twin of a patient’s left atrium using data from computed tomography (CT) scans and electroanatomical mapping (EAM) (54). They developed a virtual left atrial model to induce and observe virtual atrial fibrillation, testing the effects of different concentrations of virtual amiodarone after ablation, aiming to advance personalized healthcare. Jin et al. the likelihood of diseases such as heart failure and myocardial infarction (55). For integrated patient CT and electrophysiological data to construct a high-fidelity atrial fibrillation model. They simulated catheter ablation in a human atrial fibrillation digital twin and compared the effects of treating atrial fibrillation under different pulmonary vein gap conditions. Digital twins can simulate the outcomes of various virtual interventions under the same, repeatable conditions, without ethical concerns. During this process, physicians can continuously adjust and optimize treatment plans. High-risk procedures can be tested on digital twin models before being performed in real life, ensuring lower risk during actual operations (56). Ložek et al. constructed a right ventricular model of tetralogy of Fallot to simulate the ability of personalized CRT to improve ventricular function and efficiency, and realized personalized treatment decisions driven by digital twins (57). Vanzella et al. adopted a digital twin model to meet the challenges of cardiac rehabilitation caused by the COVID-19 pandemic, improving patient compliance and quality of life (58).

Furthermore, as an innovative simulation and analysis tool, digital twins can provide decision support and optimization for the accurate diagnosis, surgical planning, and postoperative rehabilitation of orthopedic diseases (59–61). Similar to their application in cardiovascular diseases, digital twins can create a virtual surgical environment where doctors can simulate surgical plans and predict risks, thereby reducing the likelihood of surgical complications and improving patient outcomes (62). Swaitly et al. promoted the application of digital twins in personalized dental care. They used the U-Net architecture to automatically segment teeth directly from CBCT images, improving the accuracy of the generated 3D model and reducing the time required for segmentation (63). Guo et al. used digital twin technology to build a dynamic monitoring model of upper limb force to achieve real-time interactive operation between physical space and digital space. This model ensured the accurate solution of muscle force and promotes the application of digital twins in the fields of human rehabilitation medicine and sports training (64). Due to the versatility and high fidelity of digital twins, they can simulate biological representations from a single cell to a complete human body. This is achieved through advanced computational algorithms and artificial intelligence, enabling the acquisition of simulated experimental data. As a result, digital twins can play a crucial role throughout various stages of new drug development, including preclinical research and clinical trials, significantly accelerating the pace of drug development and the time it takes for drugs to reach the market (51). By modeling populations using the physiological and biochemical indicators collected from early-stage data and recruited subjects, virtual models that meet clinical trial requirements can be created (65). While individual differences mean that digital twins cannot fully replace real subjects, they provide immense value in drug development. This process allows for multiple drug trials, enabling the observation of the metabolic processes of drugs within the body (66). In the future, the further development and application of digital twins in the healthcare sector may gradually reduce reliance on real subjects. It can play an even greater role in the precise diagnosis of various systemic diseases, personalized treatments, and full-cycle rehabilitation, leading to more efficient and effective healthcare solutions (67).

In recent years, digital twins have become increasingly integrated into the construction of smart cities, giving rise to the concept of “digital twin cities.” By creating virtual replicas of urban environments and continuously evolving through learning, digital twins provide real-time insights into city planning through big data, machine learning, and artificial intelligence assistance (68). The application scenarios and models of digital twins in urban community development are illustrated in Figure 2. In the process of building and applying urban digital twins, IoT and sensor technology are used to collect data in urban physical space, transmit and preprocess it through the cloud, and build a multi-dimensional city model based on multi-source data. Collected data on infrastructure, traffic, weather, and other variables in digital space enable continuous monitoring of the city and predict its future development. This dynamic approach helps in optimizing urban planning, enhancing resource allocation, and improving the overall quality of life for city residents. Table 2 summarizes typical applications of digital twins in urban planning and development.

In recent years, the integration of AI technology with digital twin methodologies has significantly advanced the management and development of urban communities (69, 70). For instance, Gkontzis et al. combined AI with digital twin approaches to handle spatial issues in the interaction platform of Petros City citizens and urban data (71). Their framework quickly maps the spatial distribution of citizen requests related to urban problems based on specific coordinates. Additionally, the trained machine learning models in their framework predict the likelihood of recurring issues, thus enabling proactive urban management. In Los Angeles, the implementation of an intelligent traffic system leveraged digital twin simulations to test different traffic scenarios, optimizing signal lights and other infrastructure in real-time (72, 73). This system effectively addressed traffic congestion while enhancing the city’s mobility.

Wolf et al. developed a city digital twin using Microsoft’s Azure cloud computing technology (74). This model allows for the visualization of available resources, sends automatic updates, and integrates location-based real-time weather and traffic data. It serves as a foundational framework for the development and design of multi-agency event responses in future smart cities. Similarly, Barcelona has adopted an integrated smart city system with digital twin technology (75). By collecting real-time data on traffic conditions, available parking spaces, and bus usage, the system accurately monitors the status of the traffic network. Advanced analytics and AI algorithms applied to continuously updated digital twins help city workers precisely monitor the current traffic network and gain insights into traffic patterns, identify congestion hotspots, and forecast future mobility challenges. Similar applications have been successfully implemented in Chicago and Los Angeles, where the systems have been tested and validated (73, 76).

In Singapore, VIZZIO created the world’s largest digital twin—Virtual Singapore (77, 78). A wide array of sensors, cameras, and IoT devices are deployed throughout the city to collect real-time data from various urban systems. This data is analyzed using digital twin technology for accurate detection, intelligent decision-making, and improving the efficiency of public service departments. Austin et al. constructed a city digital twin to enhance data collection, event identification, and automated decision-making, emphasizing the importance of AI and machine learning in analyzing energy usage in Chicago (79). Beckett et al. explored an AI-driven, 3D modeling, visualization tool integrated with spatial cognition algorithms to provide a detailed understanding of evolving urban development patterns (80). Furthermore, Zvarikova et al. used different urban sensor data to drive machine learning models, creating a powerful and complex framework for enhancing the accuracy of spatiotemporal modeling in urban settings (81). Amsterdam has built the first 3D printed pedestrian bridge in history, and a digital twin has been created through sensors to track the performance of the structure under pressure in real time and predict the performance and life of the bridge (82).

It is noteworthy that digital twin cities can leverage data from past major crisis events to simulate emergency situations within a virtual city model. This simulation enhances the city’s ability to respond to sudden crises and increases its resilience. By preparing and planning for potential future challenges, digital twins provide a theoretical foundation and practical basis for the application of this technology in public health emergency management.

The comprehensive preparatory phase in emergency management functions as the “reserve warehouse” for the entire emergency process. During the prevention phase, big data analytics are employed to examine data within urban and public health management models, predicting potential public health risks and enabling proactive preventive measures. For instance, public health facility placement can be planned based on trends in urban population mobility and the patterns of disease transmission. Furthermore, the system facilitates the execution of routine simulated emergency drills and training for emergency management personnel.

During the preparation phase, digital twin technology is utilized to collect and analyze real-time data from public health emergencies, such as influenza outbreaks, including the geographic distribution of flu cases and infection rates within populations. By integrating data from multiple sources, the digital twin model is continuously updated. This enables the simulation of the resources required for public health events of varying severity, allowing for proactive resource allocation and preparedness. Additionally, the system enhances the surge capacity of the healthcare system, which refers to its ability to meet peak demand during emergencies. The digital twin platform facilitates the creation of a “seamless” surge capacity across institutions and regions, using intelligent technologies to improve surge recognition capabilities. It also mobilizes multiple stakeholders to respond collaboratively, thereby strengthening the emergency response capabilities during the preparation phase. Furthermore, detailed emergency response plans are developed and continuously optimized. This model can be used to monitor the progression of an epidemic and predict future trends, providing valuable data-driven insights for public health decision-making.

In the event of a public health emergency, the emergency response phase functions as the “frontline” of the entire process. The Digital Twin system facilitates comprehensive, real-time monitoring of public health emergency conditions, extending from urban to community levels, based on live data. This encompasses the tracking of the velocity and extent of disease transmission, in addition to the efficacy of community containment measures. Concurrently, the system employs monitoring outcomes to expeditiously simulate, optimize, and adjust response strategies within the Digital Twin platform, thereby enhancing resource allocation. The enforcement of proportionate measures during emergency responses is of paramount importance in preventing the further escalation and worsening of the public health crisis. The implementation of well-coordinated, efficient, and evidence-based interventions is imperative to mitigate the impact of the emergency.

The recovery phase functions as the “supply station” in the emergency management process. The proposed Digital Twin system in the recovery phase encompasses two dimensions: the recovery of patients’ health and the functional recovery of urban communities following a public health emergency. The utilization of the Digital Twin facilitates the evaluation of the progression and outcomes of the disease, the effectiveness of treatment methods, and the rationality of response measures. Furthermore, the system can predict the long-term impacts of public health emergency measures on urban society, the economy, and the environment, providing scientific evidence to guide the comprehensive recovery of cities.

When discussing the application of digital twin technology in public health emergency management, it is necessary to consider its technical and practical challenges in a comprehensive manner. At the technical level, the system relies on high computational power to process multidimensional heterogeneous data (e.g., population flow, virus transmission) in real time, but model complexity and network bandwidth constraints can cause computational delays, and regional digital divides (e.g., insufficient sensor coverage in underdeveloped areas) further undermine the accuracy of the twin model. In addition, idealized model assumptions (e.g., uniform population distribution) conflict with realistic spatio-temporal heterogeneity, and cross-sectoral data barriers hinder dynamic information integration. Limitations of the framework include sensitivity to data quality, high infrastructure costs, lack of scenario generalization, and system vulnerability to extreme events. In addition, although the conceptual framework proposed in this study provides a theoretical basis for follow-up research, it still needs to be combined with empirical data calculation and analysis in the future to provide a set of unified theoretical analysis templates and measurement paths for the field of health emergency.

In the future, hybrid modeling, resilient infrastructure design and cross-domain ethical governance are needed to facilitate the transformation of technology from theoretical validation to real-world resilience. Lastly, constructing a digital twin requires a large amount of data, but the accuracy of this data is crucial, as the reliability of the model can be easily affected by unreliable data. These limitations constrain the development of digital health emergencies. In future research, efforts can be made to establish unified operational standards, enhance data collection and processing technologies, improve the accuracy and effectiveness of data, and build more secure and reliable databases. These measures will help drive the development of digital twins in the broader public health field and promote the advancement of digital medicine.

It is worth noting that the DT framework based on PPRR theory has been influenced by different economic levels and different political systems from construction to application. Among them, the economic level mainly affects SF and MP through resource endowment and the construction of digital infrastructure. Higher financial and information investment usually improves multi-source coverage and data freshness, and supports higher frequency model calibration and cross-scale coupling. Therefore, the marginal effects and differences reflected in health emergency DT in regions with different economic levels are most significant in the prevention/preparation stage. In addition, the institutional environment mainly affects DC and GC. Rule stability and cross-departmental collaboration maturity determine the execution receipt rate and execution delay. The rule of law and privacy compliance system, cross-departmental collaboration mechanism, performance and accountability rules, etc. Clarification and improvement are more conducive to creating higher GC in the response/recovery stage. In the future, on the basis of improving the framework quantification and evaluation system, quantitative indicators such as API penetration rate, execution binding rate, audit closed-loop rate, and interoperability standard adoption can be used to implement heterogeneity testing and weight reallocation, so as to quantify the economic impact SF/The role of MP and institutional impact DC/GC is quantitatively embedded in DT-PPRR evaluation and comparison.

It is important to note that digital twin technology is not a future technology. The successful application of digital twins in health emergency management will depend on the development of accurate risk assessment and intervention mathematical models based on a deep understanding of the fundamental mechanisms of public health emergencies. These types of digital twin systems can provide accurate risk prediction models, real-time data collection for model support, and effective interventions based on model outputs. Therefore, digital twins in health emergency management should not be viewed as replacements for existing emergency management methods, but rather as a tool to effectively translate ongoing foundational research into practical emergency responses. For government macro-coordination departments, it is necessary to formulate phased data sharing and temporary authorization rules for different PPRR scenarios, and clarify the “minimum necessary data set,” “use purpose restrictions” and post-audit to improve GC. At the same time, a unified interoperability standard is established to improve DC. MP and DC are improved by specifying regular drills, and the number and distribution of sentinels for real-time data collection are optimized to improve MP. For grassroots executive departments such as communities and hospitals, a digital twin sensing and scheduling plug-in for ICU/outpatient/emergency/operating room should be established, and a syndrome monitoring and community follow-up data recovery mechanism should be established. Lightweight IoT should be used to synchronize with batch processing to improve SF and MP. For data/communication technology supply departments, standardized SDKs for edge-cloud collaboration, data quality monitoring and integration blocks should be provided to support privacy-protected computing and improve SF/GC. In addition, legislative supervision and network information departments should clarify the exemption of data use for emergencies to improve GC.

Public health emergencies are highly uncertain and timely, so policy trade-offs need to be considered between the stability of the framework and emergencies to ensure the controlling of the system under when uncertainty surges. In the prevention/preparation stage, it is necessary to balance the early warning advance with the false alarm rate caused by early panic. In order to pursue a longer early warning advance, the risk of false alarm and induce emotional spillover may increase. A graded early warning and expert confirmation model should be established to avoid directly amplifying social media noise into public panic.

The response phase should balance the potential contradiction between digital twin-driven automated scheduling and audit/authorization by relevant departments. Rapid automated scheduling must be limited by event scope, use restrictions, temporary authorization whitelist and full-link audit, to achieve traceable, fast and legal calls between speed and compliance. At the same time, the response stage should also balance the efficiency of resource scheduling with the protection of vulnerable groups, improve the efficiency of medical resource scheduling in emergency events, and avoid allocation bias caused by a single efficiency orientation.

However, data security and privacy concerns are significant issues. In the public health field, digital twin technology involves large amounts of sensitive public health, medical, and personal privacy data. If this data were to be stolen by malicious actors, it could severely infringe on citizens’ right to life, health, and privacy. Issues related to data access and protection, regulatory approval, integrating data from different populations, and building trust with public health officials and communities during the decision-making process will turn the transformation of digital twins in health emergency management into a long-term battle, rather than a short-term sprint.

In this study, an integrated anonymization scheme is proposed under the framework: Under the principle of purpose limitation and minimum necessity, data classification and minimization are completed first. The direct identifiers such as name and age are removed and retention strategies such as purpose limitation and expiration destruction are set. Then, through the multi-layer technology superposition of table anonymization and differential privacy, privacy protection calculation and synthetic data, named entity recognition and desensitization processing are implemented on the data. At the governance level, establish pre-processing data protection impact assessment and audit, and clarify resharing prohibition and return/destruction clauses. The prevention and preparation phases prioritize aggregate release using minimal data sets, quasi-identifier generalization, and differential privacy. In the response phase, privacy computing is enabled under the constraints of the restrictions to support minute-level linkage. Minimize the risk of re-identification, and play a safe support role in DT-PPRR modeling, early warning and decision-making collaboration in sudden and normal scenarios.