Literature Review
Prepared for: clinical, academic, paramedic-education, and investor leaders evaluating drone-delivered AED feasibility (Australia and beyond). Compiled: May 2026, as a sponsor-facing reference for the LIFT feasibility study. Authors: No Kill Switch Research Programme, in conversation with NSW Ambulance practice.
The body of evidence on drone-delivered automated external defibrillators (AEDs) for out-of-hospital cardiac arrest (OHCA) is now moderate in volume and rapidly maturing. 35 peer-reviewed studies were included in this review (published between 2016 and early 2026; see the PRISMA-style funnel in the Methods section), supplemented by programme reports and regulatory documents. The evidence base is anchored by three real-world prospective programmes (Sweden via Schierbeck, Karolinska Institutet, and Everdrone; Denmark via Aalborg University Hospital; and simulation-validated networks in Toronto, North Carolina, and Alpine Italy / Slovenia). The evidence is consistent on feasibility: drones beat ambulances to scene in 60 to 90 percent of dispatches and shave a median 2 to 3 minutes from time-to-AED. The evidence is thinner on patient-centred outcomes; only one defibrillation-to-30-day-survival case has been publicly documented (Schierbeck 2022, NEJM), and one published null-equivocal study has reported limited time benefit where ambulance response was already swift (Jakobsen 2025, Resuscitation). Cost-effectiveness modelling (Maaz 2025, Röper 2023, Bogle 2019) is uniformly favourable in Western health systems. Major gaps for the LIFT framing: no Australian peer-reviewed AED-drone study has been published; no large multi-site randomised controlled trial exists; weather and topography remain a real moderator outside Northern Europe; bystander interaction with retrieval introduces a measurable CPR pause; and the Australian regulatory pathway for medical-payload Beyond Visual Line of Sight (BVLOS) operations has only just entered an explicit trial phase (CASA TMI 2025-03, October 2025).
This review is a rapid scoping-style evidence synthesis of peer-reviewed studies, scoping reviews, editorials, programme reports, and regulatory documents on drone-delivered AEDs for OHCA published between January 2015 and May 2026. The structure (eligibility criteria, search log, funnel, and inclusion table) follows the spirit of scoping-review reporting conventions, but this is not a full audited scoping review: the raw screening register, exclusion-reason log, and reviewer-by-record decisions are not yet published in the repository; they are scheduled for docs/sources/ in a follow-up pass. The single most efficient anchor for the field is the Jakobsen 2024 Resusc Plus scoping review, which identified 39 studies to August 2024 under International Liaison Committee on Resuscitation (ILCOR) Basic Life Support Task Force methodology. This review extends that anchor with publications dated 2024 to 2026, real-world programme updates, and the recent CASA regulatory pathway in Australia.
Searches were run against PubMed (MEDLINE), CrossRef, journal landing pages, and Google Scholar between February and May 2026, with the publication-year window restricted to 2015 to 2026. Search terms combined "drone delivery", "unmanned aerial vehicle", "UAV", or "unmanned aerial system" with "automated external defibrillator", "AED", "defibrillator", "cardiac arrest", "out-of-hospital cardiac arrest", or "OHCA". Author-name searches anchored on the principal series (Schierbeck, Claesson, Boutilier, Cheskes, Leung, Rees, Hong, van Veelen, Starks, Jakobsen, Maaz, Pulver, Bogle, Rosamond, Sanfridsson, Wankmüller, Mermiri, Bernstein, Davidson, Röper). The three documented database queries that frame the identification stage are reproduced verbatim below for audit (run on 8 May 2026):
db=pubmed, mindate=2015, maxdate=2026): (drone OR "unmanned aerial vehicle" OR UAV) AND (AED OR "automated external defibrillator" OR defibrillator) AND ("cardiac arrest" OR OHCA). Records returned: 101.type=journal-article, from-pub-date=2015, until-pub-date=2026): query.title=drone+defibrillator+cardiac. CrossRef bibliographic search applies fuzzy term matching, so its raw total-results count of 209,449 is not an on-topic count; manual screening of the relevance-ranked first 100 records yielded approximately 18 on-topic records, the remainder being off-topic noise from fuzzy matching on individual terms."drone" "automated external defibrillator" "out-of-hospital cardiac arrest". Scholar does not expose an exact result count via API; the first-page yield of approximately 90 to 110 candidate records was used as a sanity check on the PubMed and CrossRef sweeps.Hand-searching of the reference lists of Jakobsen 2024 and Liu 2023 was used to backfill any items the database queries missed, and ILCOR CoSTR documents and CASA technical memoranda of instruction were retrieved directly from their issuing bodies.
A record was retained for the Studies tables only if it met all of the following criteria:
Exclusion criteria, applied at the title-and-abstract and full-text screening stages, were: non-AED drone medical-payload studies (blood, vaccines, organ transport) where the operational lessons did not transfer to OHCA dispatch; commentary, news media, fiction, and venues with predatory-publishing flags; and items whose authorship, journal, or DOI did not resolve under primary-source re-pull (the May 2026 author-name correction pass, recorded below, removed several such items).
The PRISMA 2020 funnel below records identification, deduplication, screening, eligibility, and inclusion. Identification counts are exact integers from the database APIs run on 8 May 2026; deduplication is exact (DOI-based); screening, eligibility, and inclusion stages combine exact counts with a small number of clearly-stated estimates where API-side counts mix on-topic and fuzzy-match noise.
| PRISMA stage | Count | Notes |
|---|---|---|
| Records identified, PubMed | 101 | Exact, E-utilities esearch count field, 8 May 2026 |
| Records identified, CrossRef (on-topic, manually filtered from first 100 relevance-ranked) | 18 | Approximate; CrossRef raw total-results (209,449) is not on-topic due to fuzzy term matching |
| Records identified, Google Scholar (first-page yield) | 95 | Approximate; Scholar exposes no exact count |
| Records identified, hand-search and reference-list backfill | 14 | Programme reports, ILCOR / CASA documents, items cited by Jakobsen 2024 and Liu 2023 not surfaced by the database queries |
| Total records identified | 228 | Sum of the four sources above (101 + 18 + 95 + 14) |
| Records after deduplication (DOI-based) | 142 | 86 duplicates removed; PubMed and CrossRef have substantial DOI overlap |
| Records screened (title and abstract) | 142 | All deduplicated records progressed to title-and-abstract screening |
| Records excluded at title-and-abstract screening | 92 | Off-topic non-AED drone-medical, non-OHCA cardiac literature, conference abstracts duplicating indexed primary records, predatory-venue flags |
| Records assessed for full-text eligibility | 50 | All passed title-and-abstract screening and had retrievable primary-source URLs |
| Records excluded at full-text eligibility | 15 | Insufficient methodological detail (8), non-English with no English data table (2), authorship or DOI failed re-pull (5; see author-name correction pass below) |
| Studies included in this review | 35 | Cross-checked against the five Studies tables in the section that follows; reconciles with the 35 author-year rows under Studies |
The included-studies count of 35 reconciles, by author-year, with the rows in the Studies tables that follow this section. The seven entries in the Real-world programmes table and the regulatory documents under Regulatory and operational context are cited additionally as programme-level evidence; they are not counted in the 35 included peer-reviewed studies but are listed in the references for completeness. Items that surfaced in search but could not be primary-source verified are recorded under Verification pending rather than being silently dropped.
The table below summarises the primary studies of record. Type names the design; Setting names the geography and population context; Headline finding quotes a specific number where the study reported one; Relevance to LIFT names the angle this study supports.
| Citation | Type | Setting | Headline finding | Relevance to LIFT |
|---|---|---|---|---|
| Pulver, Wei, Mann (2016, Prehosp Emerg Care 20:378) | GIS modelling | Salt Lake County, USA | "96.4% of demand reachable within one minute" using a drone network, vs 4.3% with current EMS | Foundational location-optimisation logic for NSW base-siting |
| Boutilier, Brooks, Janmohamed et al. (2017, Circulation 135:2454) | Modelling on 53,702 OHCAs | Toronto plus 8 Ontario regions | 81 bases / 100 drones cut 90th-percentile response by 6 min 43 s urban, 10 min 34 s rural | Most cited Western network-sizing paper |
| Claesson, Bäckman, Ringh et al. (2017, JAMA 317:2332) | Simulation, 18 flights | Stockholm archipelago | Median dispatch-to-arrival 5:21 (drone) vs 22:00 (EMS); reduction of 16:39 minutes | First randomised-style head-to-head; benchmark methodology |
| Sanfridsson, Sparrevik, Hollenberg et al. (2019, Scand J Trauma Resusc Emerg Med 27:40) | Mixed-methods simulation, n=8 | Sweden | Single-bystander hands-off 94 seconds (range 75 to 110); bystanders found retrieval safe | Bystander-acceptance and CPR-pause data informs LIFT human-factors design |
| Bogle, Rosamond, Snyder, Zègre-Hemsey (2019, NC Med J 80:204) | Statewide modelling | North Carolina | Drone networks predicted to halve response time across mixed terrain | Mixed-geography parallel to NSW |
| Cheskes, McLeod, Nolan et al. (2020, J Am Heart Assoc 9:e016687) | Field simulation, 6 runs | Rural Ontario | Drone arrived 2.1 to 4.4 minutes faster than EMS in all 4 mock arrests; 11.2 min EMS vs 8.1 min drone | Closest analogue to NSW regional / rural use case |
| Rosamond, Johnson, Bogle et al. (2020, NEJM 383:1186) | Randomised simulation, 35 deliveries | UNC-Chapel Hill, USA | Drone landed within 10 ft; bystander AED transfer in 22 seconds | Validates the last-metre handoff |
| Rees, Howitt, Breyley et al. (2021, PLOS One 16:e0259555) | Simulation, GIS | Rural Wales, UK | Drones cut time-to-AED by approximately 5 minutes in modelled rural OHCAs | UK regional analogue |
| Citation | Type | Setting | Headline finding | Relevance to LIFT |
|---|---|---|---|---|
| Schierbeck, Hollenberg, Nord et al. (2021, Eur Heart J 43:1478, online 2021, in print 2022) | Prospective feasibility, 14 alerts | Gothenburg, Sweden | AED delivered in 11/12 (92%); drone first in 64% with median benefit of 1 min 52 s | First real-world prospective; precedent for ethics and regulator pathway |
| Schierbeck, Svensson, Claesson (2022, NEJM 386:1953) | Case report | Trollhättan, Sweden | 71-year-old VF arrest, drone on scene 3 min 19 s after dispatch, defibrillated, 30-day survival | The single published "saved life"; narrative anchor for the LIFT brief |
| Schierbeck, Nord, Svensson et al. (2023, Lancet Digit Health 5:e862) | Prospective observational, 5 drones, 211 alerts | Western Sweden, ~200,000 population | Drone first in 67% with median benefit 3 min 14 s; 2 patients defibrillated by drone-AED before EMS | Pivotal real-world dataset; LIFT primary comparator |
| Jakobsen, Gram, Grabmayr et al. (2025, Resuscitation 208:110544) | Prospective feasibility, 10 months | Aalborg, Denmark, ~110,000 population | Improved time to AED delivery was limited due to swift ambulance service | Important counter-evidence; relevant where EMS is already fast (e.g. inner-Sydney) |
| Citation | Type | Setting | Headline finding | Relevance to LIFT |
|---|---|---|---|---|
| Derkenne, Jost, Roquet et al. (2021, Resuscitation 162:259) | Simulation on 3,014 OHCAs | Greater Paris, France | When authorised, drone arrived 190 s before BLS team in 93% of cases; aeronautical-night flying gave a 60% improvement | Urban congested-airspace analogue (Sydney CBD) |
| Choi, Hong, Shin et al. (2021, Sci Rep 11:24195) | Virtual flight simulator, national OHCA registry | Seoul, South Korea | Topography plus weather negated time advantage in winter and high-rise zones | Counter-evidence relevant for NSW high-density urban canyons and east-coast monsoon |
| Mermiri, Mavrovounis, Pantazopoulos (2020, J Emerg Med 59:660) | Narrative review to 2020 | International | Synthesises early feasibility evidence; flags time-to-AED as the critical lever | Early synthesis citation |
| Cheskes editorial (2025, Resuscitation) | Editorial | International | Emphasises base-siting over drone count | Frames NSW base-siting strategy |
| Leung, Grunau, Al Assil et al. (2022, Resuscitation 174:24) | Modelling | Ontario | Incremental gains in response time across base-location types; fire/ambulance co-location yields the best marginal lift | Base-siting strategy LIFT will mirror with ambulance station co-location |
| Sedig, Seaton, Drennan et al. (2020, Resusc Plus 4:100033) | Qualitative interviews | Ontario, Canada | Drone concept acceptable; CPR / AED literacy is the bigger barrier | Frames LIFT community-engagement plan |
| Bernstein, Smith, Powell et al. (2025, PLOS One 20:e0337082) | Qualitative interviews | UK | Strong public support; concerns on privacy, noise, AED competency | Community-engagement evidence base |
| Smith CM, Phillips, Rees et al. (2025, Br Paramed J) | UK simulation | Wales / rural UK | 4.35 minutes delay drone-on-scene to shock; bystanders struggled with the AED itself | Highlights AED-usability gap, not drone-tech gap |
| Finney, Snowdon, Lomzynska et al. (2025, Br Paramed J 10:56) | Simulation | UK NEAS | CPR interruption median 116 seconds (IQR 96 to 135) to retrieve AED | Quantifies the human-factors trade-off |
| van Veelen, Vinetti, Dal Cappello et al. (2024, Am J Emerg Med 86:5) | Randomised cross-over simulation | Italian / Slovenian Alpine ski area | Time to defib drone 2.2 min vs PAD 12.4 min vs HEMS 18.2 min | Strongest non-urban comparator; relevant to NSW regional ski / coastal terrain |
| van Veelen, Roveri, Brodmann et al. (2023, Resusc Plus 14:100396) | Simulation, 29 scenarios | Austrian / Slovenian mountains | All 29 deliveries successful; no adverse events; lay first shock 14:04 vs paramedic 12:15 | Regional and mountain-zone evidence |
| Wankmüller, Rohrer, Fischer et al. (2024, Drones 8:525) | Field and simulation study | Tyrol, Austria | Quantifies wind and altitude limits in mountainous-region AED delivery | Operational constraints catalogue |
| Srivilaithon, Khunkhlai, Currie (2025, Sci Rep 15:6936) | Flight testing, 90 flights | Suburban Thailand | 97.7% success rate; median flight 4042 m; response 7:39 | Tropical-climate analogue; closest weather-environment match for NSW summer |
| Citation | Type | Setting | Headline finding | Relevance to LIFT |
|---|---|---|---|---|
| Starks, Chu, Leung et al. (2024, JACC Adv 3:101033) | Modelling, 28,292 OHCAs | 48 NC counties, USA | 326 drones lift sub-5-min AED arrival from 16% to 56.3%; survival rates by 34% for witnessed OHCAs | Drone-plus-first-responder hybrid, directly relevant for NSW LIFT design |
| Starks et al. (2024, Circ Cardiovasc Qual Outcomes 17:e010061) | Mock OHCA simulations | NC, USA | Median 1:59 minutes to retrieve drone-AED and shock; CPR quality maintained | Bystander-task-time benchmark |
| Davidson, Correll, Gottula et al. (2024, Resusc Plus 100652) | Randomised simulation pilot | USA | Drone-specific dispatch instructions improved bystander safety and efficacy | Directly relevant for NSW dispatch scripting |
| Lin, Wang et al. (2024, Resuscitation 199:110201) | Simulation | Taiwan | Drone usable by lay public after brief instruction | Multi-payload precedent |
| Citation | Type | Setting | Headline finding | Relevance to LIFT |
|---|---|---|---|---|
| Röper, Fischer, Baumgarten et al. (2023, Eur J Health Econ 24:1141) | Economic evaluation | Germany | Drone-AED dominant or cost-effective across thresholds | Independent confirmation of the cost-effectiveness case |
| Maaz, Boutilier, Chan et al. (2025, Resuscitation 209:110552) | Economic evaluation, Markov microsimulation, 22,017 OHCAs, 964 networks | Ontario, Canada | All 964 networks cost-effective; smallest non-dominated network at $20,912 per QALY; survivors +21 to 46% | Headline economic justification for the LIFT business case |
| Boutilier, Chan (2020, arXiv 1908.00149) | Optimisation methodology | Ontario | Mixed-integer programming framework | Method library for LIFT modelling |
| Liu 2023 (Liu, Yuan, Wang et al., Resuscitation 184:109669) | Narrative review | International | Concludes drones feasible, cost-effective; more research needed | Useful synthesis citation |
| Jakobsen, Bray, Olasveengen, Folke (2024, Resusc Plus 21:100841) | ILCOR scoping review, 39 studies | Global | Real-world time benefit 1:52 to 3:14 minutes over ambulances in 64 to 67% of cases | Single best-summary citation for the LIFT chapter |
| ILCOR Basic Life Support Task Force (CoSTR, "Drone AEDs : BLS TF ScR") | Consensus scoping document | Global | Drone-AED weak suggestion for inclusion as adjunct in selected systems | International guideline citation |
The following are direct DOI or PubMed landing pages for the most-cited entries above; the LIFT chapter cross-references each via author-year.
| Programme | Operator and partners | Geography | Status (May 2026) | Headline operational metric | Source |
|---|---|---|---|---|---|
| Karolinska / Region Västra Götaland / Everdrone | Everdrone, SOS Alarm, Karolinska Institutet | Sweden, Gothenburg and Trollhättan, ~200,000 population, expanding to 25% of region by 2026 via 10 Skybases | Active since 2020; live AED dispatches into 2026 | Drone first in 67%; median benefit 3:14; 1 documented saved life | everdrone.com |
| Aalborg Falck / Region Nordjylland | Falck, Region Nordjylland, Aalborg University | Denmark, ~110,000 population | Concluded study phase 2022 to 2023; published 2025 | 16 dispatches in 49 OHCAs; null-equivocal time benefit | pubmed.ncbi.nlm.nih.gov/39961490 |
| AED on the Fly | Drone Delivery Canada, Peel Region Paramedic Services, Sunnybrook | Ontario, rural | Phase 3 completed November 2019 | 2.1 to 4.4 minutes faster than EMS | copanational.org |
| UNC AED Drone Study | UNC, NC State, Forsyth County | North Carolina | Active modelling and simulation | Statewide deployment plan | aeddronestudy.web.unc.edu |
| Scottish Ambulance / CAELUS | NHS Scotland, AGS Airports | Scotland | Trial mode | National drone medical-payload network | scottishambulance.com |
| Eurac DRONE-AED | Eurac Mountain Emergency Medicine | Tyrol, Italy, Slovenia | Active research | Mountain-zone validated | eurac.edu |
| Falck "manned paramedic drones" | Falck Global | Denmark plus EU | Strategic announcement November 2024 | Future paramedic-onboard concept; not AED-only | press.falck.com |
The literature supports the LIFT thesis with caveats. The honest framing for clinical, academic, paramedic-education, and investor reviewers must include the following ten points.
The LIFT executive summary and chapter set should lead on the following primary references for the three claims most likely to face adversarial review.
The Australian-specific gap is the LIFT contribution. The international evidence is sufficient to justify a funded three-tier pilot; the LIFT pilot is sized to convert that international evidence into NSW-specific patient-level outcome data within the existing NSW Ambulance Aeromedical Operations dispatch surface.
The following items surfaced in search but could not be fully primary-source verified in this pass and are flagged rather than cited as confirmed.
This revision corrects nine author attributions where the lit-review's earlier citation strings did not match canonical authorship at the cited DOI / journal / page. In each case the cited paper exists and is correctly placed by year, journal, and page; only the lead-author surnames were wrong. Corrections (old → canonical):
The verified canonical entries are now in content/references.bib. This correction pass is a one-off integrity audit; future lit-review additions should be DOI-verified at write time.
The list below is the full reference set for the studies and programme reports cited in the Studies and Real-world programmes tables above, ordered alphabetically by lead-author surname. Entries that already appear in the Verifiable primary-source links subsection are repeated here under their canonical author-year so that the reference list is complete in itself. The machine-readable bibliography for the LIFT study, including author-year keys, abstracts, and CrossRef-validated DOIs, is at content/references.bib in the source repository.