Description

STUDY PROTOCOL: Drone Delivery of Automated External Defibrillators to Lay Users (DAEDALUS): A proof of concept study

1. BACKGROUND Out-of-hospital cardiac arrest (OHCA) is a medical emergency where the heart suddenly stops beating, causing a cessation of blood flow to vital organs. Many cardiac arrests are triggered by the heart entering an abnormal rhythm, known as an arrhythmia, which disrupts its ability to pump blood effectively. Defibrillation, the process of delivering an electric shock to the heart using an Automated External Defibrillator (AED), can correct this abnormal rhythm and restore normal heart function, significantly improving the chances of survival when administered promptly (1, 2). AEDs are designed to be simple to use, with prompts guiding untrained bystanders through applying the device and performing CPR. Quick access to an AED can increase the chances of survival by 50-70% (3), and each minute without defibrillation reduces survival rates by approximately 10%. In the UK, over 30,000 OHCAs occur each year, but survival rates remain below 10%, largely due to delays in administering defibrillation (4).

Public access AEDs are placed in busy public areas to assist bystanders during a cardiac arrest before ambulance services arrive, significantly increasing the chances of survival (5). However, their effectiveness is limited by the accessibility and speed at which they can be retrieved and used, particularly in residential areas where 80% of cardiac arrests occur (6). AED usage remains low (\<5%) because most cardiac arrests happen at home, where AEDs are rarely available (7). This highlights the critical need for solutions that can rapidly deliver AEDs to residential locations. Only 2% of out-of-hospital cardiac arrests (OHCAs) in residential settings result in successful resuscitation (8). The disparity in survival rates between public and private settings underscores the need for solutions to ensure AED deployment, including in residential areas. Only 48% of 999 calls for cardiac arrests meet the target response time of an ambulance arriving with an AED in 7 minutes (9). A new solution is therefore much needed, and this project is timely. 2. RATIONALE Drones offer a promising solution by enabling rapid AED delivery. However, while drone-based AED delivery systems have shown potential in other countries like Sweden, the UK has unique regulatory, logistical, and cultural factors that have yet to be explored or addressed.

Our study fills this gap by focusing specifically on integrating drones into the UK's emergency response system, accounting for UK airspace regulations, NHS ambulance service protocols, and public perception within the UK cultural context. This research aims to ensure that drone-delivered AEDs are integrated into real-world practice, ultimately improving the rate of defibrillation, decreasing the time to defibrillation and improving survival.

A key aspect of what sets our study apart from any previous study is the collaboration with the UK Civil Aviation Authority (CAA), Kent Surrey Sussex Air Ambulance (KSS), and Everdrone. The CAA is the UK's aviation regulator, responsible for providing oversight of all air traffic and drone operations within UK airspace, and importantly, we have CAA approval for this project. KSS, the air ambulance service for the region, will be the organisation running the drone service in this study, ensuring it is integrated with NHS emergency response systems. Everdrone, a Swedish company with extensive experience in medical drone technology, will pilot the drones and provide technical support.

Flying drones Beyond Visual Line of Sight (BVLOS)-where the drone operates outside the range that the operator can physically see-has been restricted. To overcome these challenges, there needs to be robust systems in place for detecting and avoiding obstacles, managing air traffic, and ensuring communication between the drone and the operator. Until these issues are fully resolved, the use of drones for regular operations, especially in complex airspace, has remained limited.

The CAA's BVLOS Sandbox is a controlled environment designed to address these challenges. It allows organisations to test the safety of BVLOS drone operations. By participating in the sandbox, KSS can trial drones in real-world scenarios under carefully monitored conditions, while the CAA gathers data to develop regulations that will eventually allow BVLOS operations to become a regular part of UK airspace. As indicated above, importantly, we have CAA approval for the use of BVLOS for this study, as part of the Sandbox.

KSS and Everdrone are key partners of Project LifeLine, a specific initiative within the BVLOS Sandbox that focuses on using drones for emergency medical deliveries, such as delivering Automated External Defibrillators (AEDs), EpiPens, and other life-saving equipment.

Our collaboration with KSS, Everdrone, and the CAA places this study at the forefront of medical drone technology, which has not been possible until this point. Everdrone has already successfully deployed AED drones in Sweden, showing that this technology can save lives. However, the UK's airspace is more complex, especially in urban areas like those near London Gatwick Airport. The CAA BVLOS Sandbox is crucial in helping us test and refine drone operations to ensure they are safe and well-integrated into the UK's regulated airspace. This study will generate the data needed to help pave the way for BVLOS drone operations to become a routine part of emergency medical services and help generate CAA approval beyond the CAA Sandbox.

Everdrone's system has already been assessed by the CAA as effective and safe, with a pilot project scheduled by KSS to start in late 2025. The aim of our study is to ensure that the integration and protocols developed through our work will optimise drone operations for this pilot. If our study does not go ahead, the scheduled pilot may proceed without the benefit of refined, UK-specific protocols, potentially limiting its effectiveness. 3. RESEARCH QUESTION/AIM(S)

The main research questions for this study are:

1. What are the communication, operational, and logistical challenges of using drones to deliver Automated External Defibrillators (AEDs) as part of the UK 999 emergency response system?
2. What are the public perceptions and acceptability challenges associated with the use of drones for delivering AEDs in emergency situations?

The overarching aim of the DAEDALUS study is to conduct a proof-of-concept study to develop, test and refine an integrated system to enable

3.1 Objectives

The study has the following objectives:

i. Iteratively developing and testing protocols for the integration of drone-delivered AEDs into the 999-emergency response system ii. Evaluating public and responder perceptions, including barriers and acceptability, to ensure lay responders can effectively use drone-delivered AEDs.

iii. Identifying and addressing operational challenges to ensure AEDs are deployed and applied quickly.

3.2 Outcome The primary output of this research will be tested protocols for drone-based AED delivery in the UK. These protocols will be adaptable for other emergency medical equipment, such as stop the bleed kits, medications, and blood products. We will publish all findings in open-access formats, ensuring wide dissemination to support other Air Ambulance and NHS services in augmenting healthcare logistics. This project will also help inform future medical drone initiatives, improving the delivery of life-saving devices across the country.

4 STUDY DESIGN and METHODS of DATA COLLECTION AND DATA ANALYSIS This mixed-methods study uses a concurrent triangulation design to gather complementary quantitative and qualitative data (22). The study is divided into two work packages (WP): WP1 involves developing an integrated system for drone-delivered AEDs and testing this system through simulated trials. WP1 is split into two phases: protocol development and simulation trials to refine the system. WP2 is a qualitative interview study with individuals who have lived experience of OHCA or their family/carers, as well as members of the public with no experience of OHCA. Both WPs will run concurrently, with findings from each informing the other through a continuous feedback mechanism, enhancing the overall study design and implementation (see section on integration of WP1 and WP2).

The flowchart presented in Figure 1 illustrates an example of the end-to-end process of drone-delivered AEDs as part of the UK's emergency medical response system, and the focus of our study in designing robust and tested protocols for this pathway.

Figure 1 End-to-End Process Flow of Drone-Delivered AED Integration into the UK Emergency Response System

WP 1 Phase 1 - Development stage A key challenge in drone-delivered medical devices is integrating 999 callers, ambulance dispatch, ATC, and drone operators into a rapid, coordinated system. Phase 1 focuses on creating and refining protocols for seamless drone use. Clear communication is essential, ensuring the caller knows exactly where the AED is delivered (e.g., "on the back door mat") to minimise any time without CPR, especially if only one lay responder is present.

To develop this, experts, public contributors, the research team, Everdrone, SECAmb, ATC, and KSS-will collaborate to map out the process, design the communication pathway, and establish criteria for AED drone deployment. The current plan involves the HEMS dispatcher in SECAmb's Emergency Operations Centre auto-allocating drones to Category 1 incidents (e.g., presumed cardiac arrest) to prioritise early defibrillation and improve survival outcomes.

Monthly online meetings over four months will guide the protocol development (see Gantt chart). Between meetings, the research team will refine protocols based on discussions. Findings from WP2-gathering public and responder perceptions-will inform the ongoing development.

Phase 2 - Simulation process

In this simulation, we replicate a real-world out-of-hospital cardiac arrest (OHCA) scenario as follows:

i. A lay responder, positioned with a medical training manikin representing a patient experiencing OHCA at Redhill Aerodrome, begins the scenario by placing a simulated 999 call. This call reaches the study Emergency Medical Advisor (EMA) (999 call handler) stationed at SECAmb's Emergency Operations Centre (EOC).

ii. The EMA triages the 999-call using the NHS Pathways system (https://digital.nhs.uk/services/nhs-pathways) (as per standard protocol within SECAmb), determining the patient is in cardiac arrest. The lay responder is guided through basic life support instructions, mirroring real-life emergency response protocols for OHCA.

iii. Simultaneously, a Helicopter Emergency Medical Service (HEMS) (air ambulance) dispatcher, using the computer-aided dispatch system (CAD), assesses the incident based on pre-established criteria from Phase 1. If the scenario meets the criteria for drone-delivered AED intervention, the dispatcher initiates the deployment process.

iv. Upon receiving the dispatcher's signal, an Everdrone pilot in Sweden, connected through a real-time web communication link, prepares for launch. A local ground pilot is stationed at Redhill Aerodrome, in accordance with CAA regulations requiring a visual line of sight for drone operations. The Everdrone pilot launches the AED-equipped drone from the Skybase (drone hangar) at the KSS base, aiming for a launch time of less than 90 seconds from the initial 999 call.

v.: The drone is navigated to the simulation site, where it uses a winch and spool system to lower the AED to the lay responder's location. The HEMS dispatcher, observing via the drone's video link, communicates with the EMA through the CAD system to precisely direct the lay responder to the AED's drop point.

vi. The EMA instructs the lay responder on retrieving the AED, which is housed in a quick-release case with clear visual instructions. These instructions are designed for ease of use, allowing the lay responder to quickly access the AED without additional drone interaction. Once retrieved, the drone releases the spool, and the AED's built-in visual and audio prompts guide the responder through the defibrillation process. These instructions are designed to be intuitive, ensuring that even untrained individuals can use the AED effectively.

vii. The simulation ends when the lay responder successfully delivers the first shock to the manikin.

Study size WP1 will consist of 4 simulation days over 4 months, allowing for seasonal variation and protocol amendments.

Each day will feature 4 simulated 999 emergency calls, resulting in 16 simulations in total.

Data sources/measurement

During the simulation phase of the study, the following data sources will be collected to enable a comprehensive evaluation of communication, human factors, and operational processes:

1. Audio recording of simulated 999 calls
   • All calls between the simulated bystander and the Emergency Medical Advisor (EMA) will be recorded using secure call-recording software.
   • The recordings will capture the entire interaction, including recognition of out-of-hospital cardiac arrest, delivery of CPR and AED instructions, and bystander responses.
   • Audio files will be transcribed verbatim, anonymised, and quality-checked. Following transcription, all audio recordings will be permanently deleted. Only de-identified transcripts will be retained for analysis.
2. Video recording of the simulated bystander
   • Fixed, wide-angle cameras will record the bystander participant throughout the simulation.
   • The video will capture physical actions, usability of the AED, compliance with EMA instructions, and interaction with the environment.
   • Video files will be pseudonymised for analysis. Once the analysis is complete, the video files will be permanently destroyed. They will not be shared outside the study team, uploaded to any database, or retained for teaching or dissemination purposes.
   • Still images may be extracted for analysis, but only if fully de-identified.
3. Computer Aided Dispatch (CAD) data
   • Data will be captured from SECAmb's CAD system for each simulation call. This will include:
   • Electronic timestamps (e.g. call connect, event creation, dispatch updates).
   • Dispatch data and system event logs relevant to the simulation scenario.
   • Written researcher notes will also be made during simulations to document CAD screen outputs, workflows, and any notable system behaviours.
4. Field observations
   • Trained researchers will observe each simulation using free-text notes.
   • Observations will focus on human factors (task sequence, slips/lapses, error recovery, teamwork), bystander performance, and the clarity and timing of EMA instructions.
   • Observation data will be recorded on paper forms or on University laptop computers and subsequently transcribed into secure electronic formats for analysis.

All data will be pseudonymised at source using unique study IDs. Incidental identifiers (e.g. a participant's name spoken during a call, or identifiable features in video footage) will be removed or redacted in research copies.

We will measure key time intervals, including:

• Time from 999 call to drone take-off.
• Time from drone take-off to arrival at the AED delivery point.
• Time from drone arriving at delivery point to AED being on the ground.
• Time from AED being on the ground to lay responder holding the AED (capturing the time it takes for the lay responder to remove the AED from the quick-release case it is stored in).
• Time from lay responder stopping CPR (in order to retrieve and apply the AED to the manikin) to the delivery of the first shock by the AED (hands-off CPR time).

These time intervals will be measured using a combination of electronically time-stamped events on the different computer-based systems (e.g., SECAmb CAD, Everdrone system) and manual measurements. The research team will use a high-precision stopwatch or a dedicated timing application to capture intervals that cannot be automatically recorded by the systems, ensuring all time-based outcomes are accurately measured.

Data analysis Analysis will be both quantitative and qualitative, drawing on the multiple data streams collected during the simulation phase.

Audio recordings of 999 calls will be transcribed verbatim and anonymised prior to analysis. Transcripts will be coded thematically to examine recognition of cardiac arrest, delivery and clarity of EMA instructions, and bystander responses. Interactional features (e.g. hesitations, misunderstandings, confirmation checks) will be identified, and recurring communication patterns will be mapped. Following quality checks, the original audio files will be deleted and only the de-identified transcripts retained.

Video recordings of bystander performance will be analysed using a structured coding scheme. Key events (e.g. retrieving the AED, opening the case, pad placement, compliance with safety prompts) will be timestamped, and usability issues or errors will be coded. Human factors frameworks such as SEIPS (10) will guide analysis of interactions between the person, tasks, tools, and environment. Quantitative outputs will include task completion times and error frequencies, while qualitative notes will capture contextual factors and adaptation strategies. All video files will be pseudonymised for analysis and permanently destroyed once the analysis is complete.

CAD data will provide objective timestamps (e.g. call connect, event creation, dispatch updates) which will be aligned with audio and video event markers. This triangulation will allow calculation of critical intervals (e.g. time to AED application, time to first shock-ready). Notes input to the CAD by the call handler and HEMS dispatcher during the call will be coded thematically.

Field observations will be analysed thematically, focusing on human factors such as slips, lapses, error recovery, teamwork, and the clarity and timing of EMA instructions. Observation notes will be digitised, pseudonymised, and coded alongside transcripts and video data.

Integration across these datasets will allow us to build a comprehensive picture of communication, bystander behaviour, system performance, and contextual influences. This mixed-methods approach will enable both fine-grained measurement of key performance outcomes and a richer understanding of the human and organisational factors shaping them.

Statistical methods Data will be recorded on custom case report forms. Time intervals will be reported using descriptive statistics: normally distributed data as mean ± standard deviation, non-normal data as medians \[interquartile range\]. All statistical analyses will be performed using Microsoft Excel.

Focus Groups Following each simulation, those involved will be asked to participate in a focus group to provide feedback, discuss challenges faced, lessons learned and opportunities to refine and improve the process. The focus groups will be facilitated by one of the research team with expertise in focus groups, with a second member of the research team taking notes. The focus group will be a hybrid of online with Microsoft Teams and face-to-face, as the different roles in the simulation will be at separate locations. The focus groups will be recorded using Microsoft Teams. Audio files will be transcribed, anonymised, and quality checked. Following transcription, all audio recordings will be permanently deleted. Only de-identified transcripts will be retained for analysis. Audio recordings of the focus groups will be transcribed and anonymised prior to analysis. The findings from these, alongside the data from the process times, will be collated by the research team and used to make changes to the next simulation day. To ensure the application of lessons learned, these simulation days will be spaced apart (see detailed timetable). This approach allows time for iterative modifications to be made based on insights gained from each preceding simulation and will allow for some seasonal variation.

The focus groups from WP1 and interviews from WP2 (below) are grounded in the Technology Acceptance Model (TAM)(11), a theory that offers insights into the factors influencing the acceptance or rejection of technologies or systems. The theory states that two external factors influence attitudes towards using technology (A): one is the belief that technology will enhance performance, referred to as "perceived usefulness (U)", and the other is the belief that incorporating technology into tasks will require additional effort, known as "the ease of use" (E)(11).

Technical data The modified DJI M600 Pro hexacopter used in this study includes several Everdrone safety features: dual stereo vision cameras, a parachute recovery system, and a winching device for AED deployment. It also has LTE communication, an anti-collision light, and an ADS-B IN system to track nearby aircraft. The AED payload is tested for safety, even from 120 meters, and the total weight is 12.5 kg. KSS owns two Everdrone E1 drones and a Skybase at Redhill Aerodrome, minimising study costs by eliminating the need for drone funding. These drones deploy a spool-winch system to lower AEDs to the ground.

WP2 WP2 aims to explore public opinions and the acceptability of drones delivering AEDs to OHCAs. Participants will include individuals with no OHCA experience, those who have assisted in a cardiac arrest, OHCA survivors, and their family members or carers.

Sample size For our qualitative interviews in Work Package 2 (WP2), we plan to conduct approximately 15-20 interviews, though this number is flexible. Instead of strictly aiming for data saturation, we will follow the concept of "information power," focusing on collecting rich, detailed data relevant to our research (12). As Braun \& Clarke (12) argue, saturation is not always a useful goal for qualitative work. We will stop data collection when the research team agree that we have sufficient depth to address our research questions, prioritising the quality of insights over a fixed sample size.

Data collection Semi-structured interviews will be conducted online or in person, lasting 45 minutes to 1 hour. The interview guide, informed by the Technology Acceptance Model (TAM) and WP1 findings, will be iteratively refined. Field notes will capture key impressions and ideas. The interviews will be recorded using secure call-recording software. Audio files will be transcribed verbatim, anonymised, and quality-checked. Following transcription, all audio recordings will be permanently deleted. Only de-identified transcripts will be retained for analysis.

Data analysis Audio recordings of the interviews will be transcribed verbatim and anonymised prior to analysis. Data will be thematically analysed using the Framework method (30), which allows for both inductive and deductive coding. Two researchers will independently code the data, resolving discrepancies through discussion. NVivo software will be used to organise and manage the data for theme comparison and analysis.

WP3 A cost-effectiveness/cost-utility model will be constructed to explore the effect that introducing drone-delivered AEDs into the UK emergency response system could have on patient health outcomes and NHS costs (and potential wider costs), incorporating data from this study, where possible. This early modelling exercise will collate information from pragmatic literature reviews to relate response times to out-of-hospital cardiac arrest outcomes, compared to the current response system. Costs of drone deployment and AED delivery will be estimated, and downstream costs based on generalisable sources. Sensitivity analyses will explore uncertainties around possible reductions in response time and inform future data collection.

Definition of End of Study The end of the DAEDALUS study will be defined as the date of the last data collection activity involving the last participant, which includes the completion of any follow-up interviews or focus groups and the final collation of simulation and qualitative data. This marks the conclusion of all participant involvement in the research.

6 SAMPLE AND RECRUITMENT 6.1 Eligibility Criteria 6.1.1 Inclusion criteria WP1

• Adults aged 18 years and older
• Able to understand verbal explanations given in English
• Physically able to perform CPR and apply a defibrillator to a training manikin

WP 2

• Purposeful sampling will recruit participants from diverse backgrounds, including those with and without OHCA experience.
• Able to understand verbal explanations or written information given in English

6.1.2 Exclusion criteria WP1

• Under 18 years of age
• Unable to perform CPR
• Severe cognitive impairments
• Pregnant individuals
• Healthcare professionals
• Unable to understand verbal English sufficiently

WP2

• Individuals under 18.
• Unable to provide informed consent.
• Experiencing severe psychological distress triggered by events surrounding cardiac arrest
• Unable to understand or speak verbal or written information given in English

6.2 Recruitment WP1 We will recruit 16 lay responders, each participating in a single scenario, to ensure new perspectives for each simulation. Participants will be purposefully sampled for diversity, including an even gender split, a range of ages, ethnicities, and socio-economic backgrounds. Recruitment will be supported by our PPI group and advertised across social media and professional networks.

WP2 Purposeful sampling will recruit participants from diverse backgrounds, including those with and without OHCA experience. We will collaborate with community and charity groups (e.g., Sudden Cardiac Arrest UK, Action for Carers) and use social media for recruitment. Additionally, we will invite individuals from our pre-grant PPIE group and the KSS Patient and Family Aftercare Group for potential involvement.

6.2.1 Participant identification For WP1 Our lead for PPIE and the PPIE group will assist in developing a recruitment strategy for WP1 and WP2, and opportunities for involvement will be advertised across social media.

For WP2, we will collaborate with community and charity groups (e.g., Sudden Cardiac Arrest UK, Action for Carers) and use social media for recruitment. Additionally, we will invite individuals from our pre-grant PPIE group and the KSS Patient and Family Aftercare Group for potential involvement.

The identification of potential participants will not involve reviewing or screening identifiable personal information of patients, service users or any other person.

6.2.2 Consent

Informed consent for the DAEDALUS study will be obtained from adult participants through a structured and thorough process designed to ensure that participants are fully informed, understand the study, and are making a voluntary decision to participate. This process will involve several steps carried out by the research team.

Potential participants will be initially approached directly by the research team. The research team members will provide the Participant Information Sheet (PIS) and explain the basic details of the study, ensuring that the participant understands the purpose, procedures, potential risks, and benefits.

Researchers will make sure to communicate clearly and respectfully, avoiding any technical jargon to ensure participants fully understand the information.

The research team will conduct detailed information sessions with interested participants. These sessions can be conducted in person or virtually, depending on the participant's preference and convenience.

During these sessions, a trained researcher will provide an in-depth explanation of the study, including the specific procedures involved, any potential risks and benefits, and what participation entails. The researcher will use layperson-friendly language, ensuring a comprehensive understanding.

Participants will be given ample time to consider their involvement, discuss it with family or friends, and seek further clarification if needed before making a decision.

Participants will be encouraged to ask any questions they may have about the study. The researcher will address these questions comprehensively, ensuring that all concerns are adequately resolved.

The researcher will also explain that participants can withdraw from the study at any time without any consequences.

The consent form will be reviewed with the participant in detail. The researcher will go through each section of the form, ensuring that the participant understands the content and implications.

The researcher will highlight the voluntary nature of participation and reiterate that there are no penalties for choosing not to participate or for withdrawing from the study later.

Once the participant confirms that they understand the study and agrees to participate, they will be asked to sign the consent form. The participant will be provided with a copy of the signed form for their records.

The original signed consent form will be securely stored in accordance with data protection regulations.

Participant Information Sheet (PIS):

The PIS will be designed to be clear, concise, and easy to understand. It will include information about the study's purpose, procedures, risks, benefits, and the rights of participants.

The PIS will be designed and approved alongside our PPIE group to ensure it is fit for purpose and understandable.

Researchers will not solely rely on the PIS. They will provide verbal explanations and ensure that participants have a thorough understanding of the study.

Researchers will be trained to assess the participant's understanding and capacity to consent, ensuring ethical and valid consent.

Throughout the study, participants will have access to the research team to ask questions or raise concerns.

All researchers involved in obtaining consent will have completed Good Clinical Practice (GCP) training and will receive comprehensive training on the ethical principles underpinning informed consent, including:

• Understanding the purpose and nature of the research.
• Effectively communicating with participants in a clear and respectful manner.
• Assessing participants' capacity to consent.
• Managing any concerns or issues that arise during the consent process.

6.2.2 Withdrawal of Consent Participants have the right to withdraw from the study without giving a reason and without any consequence to their access to services or care. If a participant chooses to withdraw consent, they will be asked whether they are happy for the research team to retain and use any data already collected up to the point of withdrawal. No further data will be collected from that participant after withdrawal.