1. Introduction
The rate of population aging is rapidly advancing, and contemporary life is beset with mounting pressures as a result of this. Notably, the incidence of risk factors associated with cardiovascular diseases is evident. Cardiac arrest (CA) has emerged as a prominent cardiovascular emergency, with grave effects on life and well-being due to its abrupt onset symptoms and its dismal survival rate [
1]. When contrasted with instances of in-hospital cardiac arrest (IHCA), the majority of cardiac arrests (ranging from 70% to 80%) take place in non-hospital environments, exhibiting a stochastic and abrupt spatiotemporal distribution. This circumstance exacerbates the already diminished survival prospects for individuals experiencing out-of-hospital cardiac arrest (OHCA). In the event of a patient undergoing cardiac arrest, the timing of rescue becomes paramount. With each minute of delay in intervention, the likelihood of survival diminishes from 7% to 10%. Timely interventions, including the application of automated external defibrillation (AED), prove indispensable in augmenting the survival prospects of OHCA patients [
2]. Empirical research has demonstrated that administering AED defibrillation within a 3 to 5 min timeframe can yield a survival rate ranging from 50% to 70% for individuals experiencing OHCA. Consequently, the timely application of defibrillation to OHCA patients is crucial in increasing their likelihood of survival.
At present, automated external defibrillator (AED) installations predominantly comprise conventional, static fixtures located within public spaces, such as subway stations and shopping centers. These installations exhibit deficiencies encompassing unequal spatial dispersion, challenges in detection, restricted maneuverability, and incongruence in sites where out-of-hospital cardiac arrest (OHCA) emergencies occur [
3]. Consequently, there are difficulties in providing swift responses to medical emergencies, leading to a suboptimal utilization rate of AEDs [
4]. To increase patient survival rates and achieve the judicious refinement of AED distribution, scholars have harnessed mathematical modeling approaches, such as maximum coverage site selection models and two-step mobile search techniques, to optimize deployment strategies. Nevertheless, in China, investigations regarding the use of AEDs are relatively infrequent and primarily consist of qualitative examinations [
5]. These examinations undertake a comparative analysis of AED deployment standards and research advancements in various nations, the identification of shortcomings in China’s current AED distribution, and the formulation of recommendations for enhancing AED implementation [
6]. Unmanned aerial vehicles (UAVs), commonly referred to as drones, present advantages such as exceptional mobility, convenience, efficiency, delivery speeds akin to those of ground-based emergency medical service (EMS) vehicles, and immunity to ground traffic impediments. Consequently, scholars have proposed the integration of AED-equipped drones to establish a comprehensive specialized medical drone system, enabling swift and effective responses to a wide spectrum of emergency medical demands [
7]. Nonetheless, the establishment of suitable take-off and landing sites (vertiports) is of paramount importance for drone usage. Selecting optimal locations for medical drone vertiports could expand their coverage range, improve accessibility to healthcare, reduce environmental impacts and health inequalities, and promote health system reforms, which could then help to guide policy and healthcare resource allocation, promote sustainable urban development, and improve people’s quality of life, thus contributing to the realization of sustainable health development goals (SHDGs). Hence, the location of these medical drone vertiports has become a pressing concern that demands immediate attention. These efforts are crucial for the construction of contemporary, efficient emergency medical service infrastructure, in keeping with the United Nations Sustainable Development Goals and the realization of sustainable healthcare development.
Currently, research on medical drones equipped with AED facilities is in its initial stages. Therefore, when researching the location of vertiports for medical drones, it is advisable to draw insights from the prior development of logistics drones as well as relevant studies in emergency rescue operations. Scholars have conducted a number of studies on decision-making models for the selection of emergency medical facilities, aiming to determine the number and location of emergency facilities and the scope of their service under different qualifying conditions. While the four basic models of facility selection have been applied in varying degrees to the emergency medical facility location problem, in an emergency out-of-hospital cardiac arrest scenario, out of the spirit of humanitarian rescue, emergency medical facilities need to cover as much as possible of an area in demand (or its entirety) at the fastest speed possible; therefore, the coverage model [
8] is highly suitable for addressing the location problem of medical drone vertiports.
In logistics-focused drone research, scholars have considered three main aspects: meeting customer demands, operational costs, and drone performance. Shavarani et al. considered drone flight endurance and speed, aiming to minimize the overall construction cost and the number of customers out of reach of the drones. They developed models for the number and location of drones and their distribution [
9]. Chauhan et al. considered drone energy consumption and range limitations, with the goal of maximizing demand coverage. They established models for the strategic placement of logistics drone facilities [
10]. Ren et al. integrated time-sensitive requirements to create a fully automated vertiport location model with the shortest pickup time as the objective, using queuing theory to determine the optimal number of vertiports [
11]. Zhang et al. considered various levels of logistics drones operating in urban areas and constructed a vertiport location planning model with the objectives of minimizing total costs and maximizing customer satisfaction [
12]. In the field of emergency rescue research, primary objectives include adapting to the temporal variations in emergency needs and service areas, minimizing dispatch time or tardiness, and maximizing service coverage. Golabi et al. assessed the survival probability of individual streets based on the distance from the earthquake epicenter and the risk level of the corresponding area; they established a disaster relief siting model with the objective of minimizing the total number of trips [
13]. Wang et al. combined the characteristics of suddenness and urgency in civil aviation emergency rescue work and used the Analytic Hierarchy Process (AHP) and p-median models to analyze the factors influencing the layout of civil aviation emergency regional support centers. They selected appropriate layout points from existing alternative airports, contributing to a scientifically planned network layout for regional support centers [
14].
Reviewing the studies about the integration of AED with drones, Chan et al. employed a maximum coverage location model to improve the coverage of AED for OHCA accidents. Through comparative analysis, they demonstrated that the maximum coverage location model is superior to population-based AED location methods [
15]. Baumgarten et al. assessed the feasibility of drone systems equipped with AED facilities to improve patient survival rates in rural areas of northeastern Germany [
16]. Schierbeck et al. identified areas with high incidences of OHCA nationwide for the strategic placement of drone-equipped AED facilities. They quantified the number of drones required to reach 50%, 80%, 90%, and 100% of the target population within 8 minutes, proving that a minimal number of drone EMS systems significantly enhances the national OHCA coverage [
17]. Wankmüller et al. optimized the allocation of defibrillator drone stations within a designated geographic area by developing an integer linear programming model that minimizes the number of drones used and the average travel time for drones responding to patients [
18]. Claesson et al. considered the differences in response times of emergency medical facilities in urban and rural areas. They used GIS to design an appropriate layout for drone systems, significantly reducing the response time of drone-equipped AED facilities [
19]. Pulver et al. utilized a maximum coverage location model to optimize the positioning of drone-equipped AED facilities, aiming to reduce emergency response times for OHCA patients. The results showed that positions determined by the maximum coverage location model could meet 90.3% of emergency demands and save costs [
20].
Upon reviewing the relevant literature, it is evident that there is limited research on drone-equipped AED facilities; previous models for locating emergency medical facilities have typically focused on traditional single-objective functions, such as “minimizing response time”, “maximizing coverage”, or “minimizing cost”, with minimal consideration for the patient. In this paper, we address this gap by introducing the survival rate function and the concept of “golden 4-minute” rescue coverage from the patient’s perspective. Simultaneously, we consider the performance constraints related to the operation of medical drones in various phases. And we construct a multi-objective medical drone vertiport location model that addresses the emergency needs of sudden cardiac arrest, considering both patient survival rates and system operational costs. The immune optimization algorithm is applied to solve this model. Using Tianjin’s Jinnan District as an example, this paper obtains a layout of medical drone vertiports. These results can offer decision-making support for the construction of sustainable and healthy municipalities.
3. Model Building
3.1. Medical Drone Vertiport Modeling
3.1.1. Objective function
- (1)
Survival rate function
In an OHCA emergency scenario, prioritizing the fundamental rescue principle of “life first” is crucial. Therefore, on the basis of the location set covering problem and the maximum coverage location model, this paper adopts a patient-centric approach. It introduces the survival rate function and designates it as the primary optimization goal, specifically focusing on maximizing the average survival rate of OHCA patients:
is the survival rate of OHCA patients, and the survival function is a concave function of time with reference to the survival function of OHCA patients proposed in the literature [
21] and adjusted in this paper.
i is the serial number of the alternative medical drone vertiports.
are the coordinates of the vertiports;
I is the aggregation of vertiports for alternative medical drones;
= 1 or 0 indicates whether or not to establish a medical drone vertiport at the point
i;
j is the serial number for OHCA accident points;
are the coordinates of the accident points;
J is the OHCA accident point set;
= 1 or 0 indicates whether the OHCA accident point is within the coverage area of the medical drone vertiport;
is the horizontal cruising altitude for medical drones;
is the AED delivery time, i.e., the time it takes to fly from the drone vertiport to the OHCA accident point;
is the pre-scheduling and lead time (the AED can initiate external defibrillation once it arrives, with a negligible time to save the patient); and
,
are the vertical ascent phase speed, horizontal cruise speed, and vertical descent speed of the medical drone.
- (2)
Total cost function
In addition to the prioritization of the patient survival rate in the rescue, the system operating costs also need to be taken into account. Relevant researchers and scholars have measured the cost-effectiveness of medical drones’ modes of operation, so this paper incorporates the system operating costs as a secondary objective. The goal is to minimize the total cost of the medical drone operation, achieving an optimal equilibrium state where both patient survival rate and system operation costs are appropriately balanced, that is:
TC is the total cost of operating a medical drone, consisting of fixed cost FC and variable cost VC, of which represents the construction and operation of medical drone landing sites measured in CNY, is the flight path costs for AED delivery of medical drones measured in CNY per meter, including maintenance costs per unit distance , depreciation costs per unit distance , and battery maintenance costs per unit distance ; is the one-way distance for medical drone flights from a landing site to an OHCA accident site.
- (3)
Overall objective function
In this paper, we comprehensively consider two objective functions: the average survival rate of OHCA patients and the total cost of operation. Maximizing the average survival rate of patients serves as the primary optimization objective, while minimizing the cost of medical drone operation acts as the secondary objective. Due to the different magnitudes of these objectives, direct weighting is not feasible. Therefore, this paper normalizes them separately to eliminate discrepancies in magnitudes.
MT represents the total objective function considering multiple objectives, and is the weighting factor for the average patient survival after normalization. The range of values for is [0, 1]. Considering that maximizing the average survival rate of patients is the primary goal and following the principle of “life is paramount” in relief efforts, the value of should generally be greater than 0.5 and converge to 1. This paper aims to maximize the average patient survival rate, making it a positive objective in Equation (8). The larger the average patient survival rate, the better the result. Conversely, in Equation (9), the total cost of medical drone operation is the objective, and the smaller the total cost, the better the result. This represents an inverse objective.
3.1.2. Constraint
Based on the problem description, assumptions, and the characteristics of medical drone phased operation time, energy consumption, and coverage area, the medical drone vertiport location model constraints are as follows:
- (1)
Constraint of the medical drone range
The round-trip distance of a medical drone for a single AED delivery rescue mission cannot exceed its maximum range. The constraint is expressed as follows:
is the maximum range for medical drones in Equation (10);
- (2)
Constraints of rescue matching relation
For any OHCA accident point
j, there will be only one medical drone vertiport responsible for AED delivery, and there is no scenario in which the accident is not responded to or multiple points are responded to. The constraint is expressed as follows:
- (3)
Constraint of emergency response relationship
Only when a medical drone vertiport is established at point
i can the emergency needs of the OHCA accident site at point
j be responded to by a medical drone at point
i to provide AED delivery services. The constraint is expressed as follows:
- (4)
Constraints on the number of landing sites
The total number of locations for medical drone vertiports cannot exceed the number of alternative vertiports. The constraint is expressed as follows:
N is the number of alternative medical drone vertiports;
- (5)
Energy consumption constraints for medical drones
Medical drones do not exceed their maximum energy consumption during the phased operation of AED delivery. The constraint is expressed as follows:
W is the self-weighting for drones, m is the weight of AED facilities carried, g is the acceleration due to gravity (the general value is 9.8 ), is the drone propeller power transfer efficiency, is the drone flight lift-to-drag ratio, e is the energy consumption of electronic components of drones, and is the maximum energy consumption of medical drones;
- (6)
Constraint of survival rate threshold
We must ensure that the survival rate of OHCA patients rescued by medical drones does not fall below a set survival rate threshold. The constraint is expressed as follows:
is the survival of OHCA patients receiving AED defibrillation within the “golden 4-minute”.
3.2. Algorithm Design
The medical drone vertiport location problem is a classical NP problem. Existing research often tackles NP problems using optimized intelligent algorithms, which can reduce the difficulty in solving problems and simplify the solution process. The genetic algorithm is an efficient adaptive evolutionary search algorithm that realizes the improvement of each individual’s adaptability through the mechanisms of natural selection, heredity, and mutation. However, numerous practices and studies have shown that the standard genetic algorithm has defects of poor local search ability and “early maturity”. In response to these challenges, the immune optimization algorithm emerges as a new type of intelligent search algorithm inspired by the biological immune system. This algorithm exhibits strong robustness and finds wide applications in emergency facility site selection. Therefore, this paper applies the immune optimization algorithm to solve the model. The flowchart of the algorithm is shown in
Figure 4 and the specific flow of the algorithm is as follows:
Step 1: Recognition of antigens
Define stock information as a structure;
Step 2: Generate initial antibody population
The initial antibody population is randomly generated in the feasible solution space using a simple coding approach. Each medical drone vertiport scheme forms an antibody of a specific length, where each antibody represents the sequence of OHCA accident points selected as medical drone vertiports. For example, the antibody [2 7 14 31 52] represents the accident points 2, 7, 14, 31, 52 selected as medical drone vertiports;
Step 3: Evaluation of the diversity of solutions
After generating the initial antibody population, the diversity of solutions is evaluated using the expected reproduction probability as the criterion. The higher the expected reproduction probability of the antibody, the stronger the adaptive ability of the antibody. Throughout the algorithmic search process, antibodies with strong adaptive abilities are retained in the memory bank. Under the effect of immune balance, both individuals with high adaptive abilities and those with higher concentrations are promoted and suppressed. This mechanism enhances the diversity of individuals in the population.
- (1)
Antibody–antigen affinity
The affinity between an antibody and an antigen is used to indicate the degree of recognition of the antigen by the antibody. The affinity function is designed based on the medical drone vertiport siting model:
is the affinity function and is the overall objective function;
- (2)
Antibody–antibody affinity
The affinity between antibodies reflects the degree of similarity between them. In this context, the R-site continuum method proposed by Yang et al. [
22] is borrowed to calculate the affinity between antibodies. The formula is as follows:
represents the identical parts in antibody and antibody and L is the length of the antibody.
- (3)
Antibody concentration
The concentration of antibodies is the proportion of similar antibodies in the population:
n is the total number of antibodies and T is a pre-set threshold.
- (4)
Expected probability of reproduction
The expected probability of reproduction is determined by the combination of the affinity between the antibody and antigen and the concentration of the antibody. It is calculated using the following formula:
is a constant in mathematics. At the same time, an elite retention strategy is adopted, in which a number of individuals with the highest affinity to the antigen are deposited into the memory bank each time the memory bank is updated. Then, the best individuals from the remaining samples are deposited into the memory bank according to the desired probability of reproduction.
Step 4: Operations of immunization
Selection: Operates according to the roulette wheel selection mechanism, where the probability that an individual is selected is the expected probability of reproduction;
Crossover: This paper employs a single-point crossover method for crossover operations;
Mutations: Mutations are performed using the commonly used mutation method, i.e., randomly selected mutation sites.
5. Discussion
We focus on out-of-hospital cardiac arrest scenarios, considering the application of medical drones equipped with AED facilities for efficient and rapid emergency response, endeavoring to address diverse emergency medical needs comprehensively. This paper formulates a collaborative and refined operational model for medical drones based on the nature of accidents, operational characteristics of medical drones, and characteristics of out-of-hospital cardiac arrest. Differing from conventional objectives that aim to minimize time [
11] or cost [
12] in emergency facility location, driven by the humanitarian rescue spirit and from the patient’s perspective, we integrate the patient survival rate into the model as the primary optimization objective while also considering system operating costs. Additionally, informed by real emergency data, we introduce the innovative concept of achieving full “golden 4-minute” coverage in rescue, ensuring a balance between patient survival rates and equitable accessibility to healthcare services.
Through a review and synthesis of existing works in the literature, it is evident that drones hold significant potential and broad prospects in emergency medical rescue [
1]. Previous studies have effectively validated the feasibility of medical drones equipped with AED facilities to improve patient survival rates [
16]. Building upon this foundation, we integrate the goal of maximizing patient survival rates with the imperative of achieving full “golden 4-minute” coverage. Utilizing the coverage location model, we reduce the rescue time from the previously documented 8–10 minutes [
17] to an efficient 4 minutes. Additionally, we elevate the coverage rate of emergency needs from 90.3% [
20] to 100%, thereby significantly enhancing both patient survival rates and rescue coverage. The utilization of medical drones for emergency response emerges as a transformative measure, distinctly elevating patient survival rates compared to existing ground-based emergency services.
In our effort to maximize patient survival rates, we take a comprehensive approach by concurrently assessing the overall operating costs of the system. Unlike prior studies on site selection, which predominantly focus on costs associated with the site itself, encompassing construction and maintenance costs [
9], we extend this to include the construction and operation costs of medical drone vertiports [
12] and incorporate the flight path costs of phased medical drone operation, including repair costs per unit distance, depreciation costs, and battery maintenance costs. The results not only prioritize patient survival but also contribute to a substantial reduction in mortality and long-term healthcare costs, demonstrating the financial feasibility of the model and its results.
The current study primarily addresses the location of medical drone vertiports in normal operation scenarios, without factoring in potential obstacles [
23], adverse weather conditions, complex environments, or unforeseen factors. There is a need for further exploration into public acceptance of drones, associated operational risks [
24], and considerations of policy and regulations pertinent to medical drone applications. At the same time, economic feasibility aspects, such as the specific operating costs for medical drones, require more detailed examination [
25]. The impact of accident first-witness behavior on response times and patient survival rates, along with the skill levels of operators and their emergency response capabilities, introduce a certain level of error between the study’s results and real-world situations. Further studies will strive for increased relevance and reduced errors.
We strategically deploy medical drone infrastructure, including charging stations and vertiports, based on population data and the distribution characteristics of the elderly population, ensuring effective logistical support [
20]. Future work can extend to evaluating the layout and operational costs of medical drone vertiports under real demand distributions in specific areas and across various operational scenarios. Additionally, exploring innovative operational modes that integrate seamlessly with ground-based emergency service facilities will be crucial. Furthermore, in the proposed medical drone operational model, there is a focus on enhancing coordination between bystanders and healthcare personnel, ensuring the feasibility of personnel and the efficiency of emergency responses. At the same time, public awareness of emergency care should be strengthened, first aid knowledge popularized, and collaborative efforts with community and neighborhood committees should be fostered to elevate the standard of automated external defibrillator configurations. This approach can contribute to promoting health system reform for sustainable and healthy development. Future advancements in technology and innovation hold the potential to further enhance emergency response by mitigating existing risks and challenges.
6. Conclusions
This paper focuses on scenarios involving out-of-hospital cardiac arrests and explores the application of medical drones equipped with automated external defibrillators for emergency response. Addressing the challenge of locating medical drones vertiports, we introduce the survival rate function and frame the optimization objectives as maximizing the average survival rate of patients while minimizing the total system cost. Simultaneously, we take into account the constraints related to staged operation time, energy consumption, and drone coverage. The model for medical drone vertiport location in cardiac arrest emergencies is solved by applying the improved immune algorithm and verified through numerical examples. The results show that:
The demand for out-of-hospital cardiac arrests in Jinnan District, Tianjin City, based on population, age distribution, and morbidity rate, can be effectively met by deploying 24 medical drone vertiports to achieve full “golden 4-minute” rescue coverage;
By assigning a weight of 0.9 to the average patient survival rate, we optimize the system operation to achieve the highest average patient survival rate while minimizing the total cost. This approach results in an impressive increase in the average patient survival rate to 64.06%;
Comparatively, the application of medical drones for out-of-hospital cardiac arrest treatment surpasses ground ambulances in reducing response time. The average survival rate of patients increases by 41.96%, making a significant improvement over the current situation characterized by low survival rates for out-of-hospital cardiac arrest patients.
The application of medical drones not only significantly improves OHCA emergency services, but it also holds profound implications for various dimensions of sustainable health development. It contributes to improving survival rates, mitigating healthcare disparities, enhancing accessibility to health services, reducing environmental impacts, and facilitating the reform of health systems, thus significantly supporting the fulfilment of sustainable health development goals.