Preface

I’ll never forget the first time I was alerted as a GoodSAM responder. During my final year of Paramedicine study, I was due to go on shift with NSW Ambulance at 11am during my penultimate placement.

The alert came through at 07:15 on a Thursday morning and I was on scene three minutes later at a house on the corner of a main road, opposite a service station. The patient, an elderly grandmother, had collapsed in a narrow passage between a downstairs bedroom and the toilet where she had collapsed. Her granddaughter was doing CPR, but the narrow passage made it ineffective. The family were distressed and CPR was being directed by the 000 call taker on speaker. I took over the scene, confirmed arrest, and with help moved the patient to get proper access. I administered an airway, and asked the teenage grandson to check for an AED at the service station despite one not on the GoodSam AED map. He came back empty-handed.

There were five AEDs within 1.5 kilometres of us: one at a church, at a GP surgery, another church, a primary school, and a community centre but none were accessible. None of the organisations were due to open for at least 40 minutes. Even if they were open, or the AEDs accesible, the round trip time especially in morning traffic, was too long to make a difference.

So we worked the arrest with what we had; CPR, an airway adjunct, a bag-valve mask, and untrained bystanders, until the ambulance arrived 15 minutes after me. We never achieved ROSC.

What struck me wasn’t the clinical care. It was the absence of the one intervention most likely to change the outcome: defibrillation.

A few months later, I was alerted by GoodSam again. This time at 1am. This patient lived about 600 metres from my house and I arrived just before the ambulance. Again, there were AEDs nearby and again, none were accessible at that time of night.

This patient, however, had a shockable rhythm, and despite rapid response across empty streets, we didn’t get a defibrillator onto him until 15 minutes after his arrest. Even so, we achieved ROSC, but ultimately he died in the ED.

These experiences changed how I think about cardiac arrest. Despite my clinical training, despite my passion in cardiology and research in all the amazing clinical interventions we wield, BLS, ALS, retrieval and eCPR, it strikes me that this is a system design problem.

We have trained responders, established protocols and proven technology. We simply don’t have alignment between where cardiac arrests occur and where defibrillators are available. Surely the biggest impact we can make to survivability of OHCA is tweaks to what happens after the ambulance arrives, but how quickly we can defibrillate and get high quality CPR happening?

For every minute that passes without defibrillation, the chance of survival from cardiac arrest decreases by 7 to 10 percent.

American Heart Association, Early Defibrillation Advisory Statement [1]

A note on data labels

Every quantitative claim in this report carries one of four provenance labels so the reader can see exactly what kind of signal is being used. The table below names the four labels and the treatment each implies.

Label	Meaning
OBSERVED	Figure taken directly from a published or curated source.
ESTIMATED	Planning-grade scenario output; bounded, explicit assumptions.
SYNTHETIC	Demographic proxy; pending verification against NSW Ambulance data.
INFERRED	Candidate geometry derived from SA2 centroids; not dispatch-grade.

Chapter 02

Executive Summary

---

Every year in Greater Sydney, roughly 4,800 people suffer out-of-hospital cardiac arrest [ESTIMATED]. The majority collapse inside private residences, beyond the reach of public AEDs mounted in train stations and shopping centres. About 22% present with shockable rhythms — the subset where rapid defibrillation changes the outcome [2].

A five-base drone network delivering AEDs across suburban Sydney could save approximately 69 additional lives per year within its coverage zone, at a five-year cost of ~$1.5M and a cost per QALY well below published thresholds for public-access defibrillation programmes [ESTIMATED].

For every minute that passes without defibrillation, the chance of survival from cardiac arrest decreases by 7 to 10 percent.

American Heart Association, Early Defibrillation Advisory Statement [1]

The arithmetic is direct. Current median ambulance response in Greater Sydney sits at 8.5 minutes [3]. At that interval, survival probability for shockable ventricular fibrillation is roughly 29% [4]. A drone arriving in 4 minutes shifts that probability to approximately 45%. The gap between those two numbers, applied to the shockable-rhythm population within drone range, produces the incremental survival estimate.

The modelled network uses five bases, each covering a 6 km service radius. That configuration reaches about 40% of the synthetic demand surface [ESTIMATED]. It prioritises areas with older populations, high detached housing, and longer ambulance travel times. Bases are sited using demand-weighted optimisation — but the demand itself is modelled from demographic proxies, not real dispatch records [SYNTHETIC].

A word on what this is and what it isn’t. This is a feasibility model. It estimates plausible impact under a set of stated assumptions. It is not an operational plan. The demand distribution is synthetic. The cost figures are estimates drawn from published drone AED programmes in Sweden and operational logistics networks in Rwanda [5,6]. Drone flight times assume unobstructed corridors and regulatory permissions that do not yet exist in Australia. Bystander retrieval rates remain untested in Australian conditions.

None of these caveats invalidate the core finding. The survival-time relationship in cardiac arrest is among the most replicated in emergency medicine. Faster defibrillation saves lives. The question is whether drones can reliably deliver that speed advantage in practice.

This report presents the model, its limits, and a specific next step.

Chapter 03

Spatial distribution of cardiac arrest in Greater Sydney

---

A defibrillator-carrying drone network is, at heart, a spatial optimisation problem: where the events occur dictates where the rooftop cradles must be, and how far a drone must fly to reach each new collapse. This chapter characterises the spatial distribution of out-of-hospital cardiac arrest (OHCA) across 37 SA2s of Greater Sydney using the synthetic cohort carried in data/derived/synthetic_ohca_hotspots.csv. Two complementary views are shown below: a rate choropleth that normalises for population (task 3.3), and a proportional-symbol map that reports absolute event volume (task 3.4). Read together, they reveal a pattern that is not obvious from either map alone.

Synthetic-cohort caveat. The figures and rankings below descend from a proxy dataset that combines the ABS 2021 Greater Sydney SA2 population and age structure with the ABS 2021 SEIFA Index of Relative Socio-economic Disadvantage (IRSD), anchored to a Hasselqvist-Ax 2015 incidence base rate. The derivation is documented in full in Appendix M — Methodology: synthetic OHCA proxy with SEIFA IRSD weighting. Treat every absolute count on this page as pending verification against NSW Ambulance incident data; relative ordering between SA2s is expected to hold under replacement with the real cohort.

Where rates concentrate

The SA2 polygons sit on a muted Sydney base map (CartoDB Positron, © OpenStreetMap contributors) showing the coastline around Sydney Harbour, Botany Bay, the Parramatta and Georges rivers, and the principal road network — so readers can locate the hotspots against a recognisable geography. The data layer uses a ColorBrewer 5-class sequential-reds palette. Dark shading clusters across Sydney’s south-west corridor (Fairfield, Cabramatta, Bankstown, Campbelltown); the palest shading covers the northern shore and the outer growth LGAs such as Kellyville and Rouse Hill where the IRSD decile is high and the age-65+ share is low. A signal-red callout ring marks the highest-rate SA2. A quantile legend, scale bar, north arrow, base-map attribution line, and data-provenance footnote sit adjacent to the map.

Key insight. Rate concentration is a joint geography of age and disadvantage. Under the SEIFA-IRSD-weighted proxy (Appendix M), the highest synthetic rates sit in Sydney’s south-western corridor — Fairfield, Cabramatta–Lansvale, Bankstown, Campbelltown-South and Campbelltown-North all land in the top quantile. Age structure still matters, but the IRSD weighting pulls the rate surface decisively away from the leafy northern-shore band the earlier age-only proxy highlighted. The policy implication is inverted from that earlier reading: a “wealthy boomers” heuristic would mis-site base placements away from the SA2s the combined age-plus-deprivation signal actually prioritises. The choropleth also carries the full MAUP caveat: re-aggregating the same cohort to SA3 or LGA boundaries would smooth the south-west cluster into a single mid-range polygon, and pooling across LGA would re-introduce the wealthy-suburb age-signal artefact.

Where the volume sits

The 37 centroid circles sit on a muted Sydney base map (CartoDB Positron, © OpenStreetMap contributors) showing coastline and the principal road network so each symbol reads against a recognisable Sydney geography. Circle sizes range from small (tens of synthetic cases in wealthier northern SA2s) to large (more than 200 synthetic cases in the south-west cluster led by Fairfield, Bankstown and Cabramatta–Lansvale). Three signal-red callout rings label the top-ranked hotspots. A graduated-circle legend shows minimum, midpoint, and maximum case counts, with a Viridis class key underneath. Scale bar, north arrow, base-map attribution line, and a data-provenance footnote sit adjacent to the map.

Key insight. Volume and rate now point at the same part of the city — the disadvantaged south-west. The top-three SA2s by synthetic volume — Fairfield, Bankstown, and Cabramatta–Lansvale — between them carry approximately 16% of synthetic cases, and the top-five extends this cluster south to Campbelltown-South and Campbelltown-North. Every SA2 in the top-five sits in SEIFA IRSD decile 1 or 2 (most disadvantaged), confirming that the combined age-plus-IRSD proxy is the logical first-pass driver of drone-base placement. Note the deliberate encoding: a choropleth would misrepresent this view because event counts (as opposed to rates) are an extensive variable and must not be mapped by area fill. The proportional symbol map is the cartographically honest form for the question “where are the biggest hotspots, in absolute terms?”

Caveats and next steps

Both maps draw on the synthetic cohort only; neither represents observed NSW Ambulance incident data. From Sprint 26 onwards the synthetic proxy is SEIFA-IRSD-weighted; the derivation, the Hasselqvist-Ax 2015 base-rate anchor, and the pending-verification labelling convention are documented in Appendix M — Methodology: synthetic OHCA proxy with SEIFA IRSD weighting. The synthetic counts remain above the n < 5 suppression threshold for every SA2, so no cells are suppressed, but rates are not age-standardised and will need Empirical Bayes smoothing when the real cohort is slotted in. The nearest-centroid raster proxy used in Figure 3.1 is a presentational stand-in for ABS SA2 polygons and is explicitly labelled as such on the map itself. The final report build will swap in the true SA2 geometry and re-run the quantile classification; the class breaks will shift slightly but the south-west cluster is expected to survive the substitution because the ranks are carried by the underlying rates, not by the polygon shapes.

Out-of-hospital cardiac arrest kills roughly 30,000 Australians each year [2]. That is more than breast cancer. More than prostate cancer. More than road trauma. In Greater Sydney alone, the model estimates approximately 4,800 events per year [ESTIMATED], concentrated not in public spaces but inside private homes. Over 70% of OHCA occurs in residential settings [2].

The majority of out-of-hospital cardiac arrests in Australia occur in private residences, where publicly accessible AEDs are rarely available.

AIHW, OHCA Annual Report [2]

The clinical science is settled. For every minute without defibrillation, survival from witnessed ventricular fibrillation drops 7-10% [4,8]. The American Heart Association recommends defibrillation within 3-5 minutes of collapse [1]. Bystander defibrillation before EMS arrival is associated with 53% 30-day survival for shockable rhythms. Without it: 16% [9]. That is the difference between a patient who walks out of hospital and a patient who does not leave.

And yet.

In at least one regional NSW community, the aspiration has been to place an AED within 200 metres of every resident. The ambition is sound: saturate the landscape with defibrillators so that no cardiac arrest occurs beyond arm’s reach of the one device proven to change outcomes. On its face, this is the obvious solution. Put enough devices in enough places and the problem disappears.

It does not disappear. It reveals a paradox with three dimensions.

Scale and cost

What is achievable in a compact regional town does not transfer to a metropolitan area of five million people spread across 12,000 square kilometres. Placing an AED within 200 metres of every resident in a regional centre requires a manageable number of devices in a concentrated geography. Replicating that density across Greater Sydney — from Penrith to Cronulla, from Campbelltown to the Northern Beaches — would demand tens of thousands of units, each requiring purchase, installation, signage, registration, and ongoing maintenance including battery replacement and pad expiry [ESTIMATED]. The capital outlay alone would run into the tens of millions. The logistics of maintaining currency across that fleet — ensuring every device is charged, in-date, and functional at the moment it is needed — would constitute a programme of work in its own right.

Scale is not merely a larger version of the same problem. It is a different problem. The coordination cost grows non-linearly. The maintenance burden compounds. And the marginal return diminishes: the last 10% of coverage — the cul-de-sacs, the acreage blocks, the isolated pockets between commercial strips — costs disproportionately more than the first 90%.

Awareness, access, and use

Even where AEDs exist, three barriers stand between the device and the patient.

Awareness. The bystander who witnesses a cardiac arrest must know that an AED is nearby. Most do not. Public-access AED registries exist, but cardiac arrest does not wait for someone to consult a database. The 000 call-taker may direct a bystander to the nearest registered device, but the bystander is already managing CPR, managing panic, managing a dying person on the floor. Knowledge of AED locations is not widely held in the community, and under the cognitive load of a cardiac arrest, even known information becomes difficult to retrieve.

Access. The AED must be physically reachable. An AED mounted inside a surf club is not reachable at 06:42 on a Tuesday morning. An AED in a school is not reachable before 08:30 or after 15:30 or during holidays. An AED in a GP surgery is not reachable on weekends. An AED in a community centre 400 metres away requires a round trip of five to eight minutes on foot — and that assumes the bystander leaves the patient to go looking [OBSERVED]. Five AEDs within 1.5 kilometres, none accessible. That is not a hypothetical. It is the pattern described in the preface of this report, replicated across suburban Sydney every week.

Willingness and competence to use. Even when a bystander locates and reaches an AED, they must be willing to apply it to a human being in extremis. Public anxiety about using AEDs remains high despite their design for lay use [4]. Fear of causing harm, uncertainty about legal liability, unfamiliarity with pad placement — these are documented barriers to bystander defibrillation in the literature. The device is forgiving. The psychology is not.

Measurable outcomes

The most difficult dimension. Does saturating an area with AEDs produce measurable improvements in survival? Proximity alone does not complete the chain of survival. A defibrillator that is purchased, installed, registered, maintained, located by a bystander, accessed within minutes, and correctly applied can save a life. Remove any link and the chain fails. The question is not whether AEDs work — they do, unambiguously [9]. The question is whether a fixed-installation strategy translates proximity into actual use, and actual use into population-level survival gains.

The evidence is mixed. Public-access defibrillation programmes in high-traffic locations — airports, casinos, sports arenas — have demonstrated clear survival benefit [4]. These are environments with trained staff, high footfall, and rapid recognition. Residential settings share none of these characteristics. The patient collapses at home, often in the early morning, often alone or with a single household member who may themselves be elderly. The nearest AED is behind a locked door. The chain of survival breaks at the same link every time: the device does not reach the patient.

The paradox, then, is this: the aspiration to place an AED within 200 metres of every resident is clinically sound but operationally insufficient. It cannot scale to a metropolitan area at reasonable cost. Where it does exist, awareness, access, and willingness erode the benefit. And even where every barrier is cleared, the evidence for population-level survival gains from fixed residential AED installations remains inconclusive.

This is a system design problem. The devices exist. The evidence for early defibrillation is unambiguous. The clinical protocols are established. What is missing is a delivery mechanism that matches the geography and timing of the events themselves: residential, dispersed, and time-critical. The following section examines how current emergency response systems perform against that requirement — and where they fall short.

Campbelltown South is an SA2 in Sydney’s outer southwest. Disadvantage decile 1. Population 11,200. Detached housing: 82%. Residents aged 65 and over: 14.5% [10]. Streets of single-storey brick and tile homes set back from the kerb. Driveways with two cars. Low front fences. Quiet before seven.

06:42 — A 71-year-old man steps out of the bedroom and collapses in the hallway [SYNTHETIC]. His wife hears the fall from the kitchen. She finds him unresponsive on the carpet, not breathing normally. She does not know what cardiac arrest looks like. She knows something is wrong.

06:43 — She calls 000. The dispatcher recognises the pattern within 40 seconds: unresponsive, abnormal breathing. The call is coded as suspected cardiac arrest. Two things happen simultaneously. The dispatcher begins telephone CPR instructions, talking the wife through chest compressions on the hallway floor. And the dispatch system triggers a drone from the nearest candidate base, approximately 5 kilometres away [SYNTHETIC].

06:44 — The drone is airborne. Cruise speed: 100 km/h [ESTIMATED]. It flies a direct path, unimpeded by traffic lights, school zones, or the roundabout on Narellan Road. The wife is on her knees doing compressions. They are shallow and fast. The dispatcher corrects her tempo.

06:47 — The drone arrives overhead. Audio and visual alerts activate. The AED descends to the front yard. Three minutes from launch.

06:48 — The wife retrieves the AED. She opens it. The device talks to her. Peel the pads. Place them on his bare chest. Analysing. Shock advised. She presses the button. First shock delivered at approximately six minutes from collapse [ESTIMATED].

At six minutes, the survival probability for a shockable rhythm is roughly 36% [ESTIMATED, derived from [4]; P = 0.67 x exp(-0.10 x t), t = 6].

06:53 — The ambulance arrives. Paramedics take over with advanced life support. The AED has recorded two shocks and six minutes of CPR data. The patient has a pulse.

Now remove the drone.

06:42 — The same collapse. The same hallway.

06:43 — The same 000 call. The same telephone CPR. No drone dispatch.

06:53 — The ambulance arrives carrying the first defibrillator. Eleven minutes from collapse. The wife has been doing compressions for ten minutes. She is exhausted. Compression depth has dropped.

At eleven minutes, survival probability is roughly 22% [ESTIMATED, same model, t = 11].

The difference: 14 percentage points. In a population where roughly one in five attended OHCA presents with a shockable rhythm [2], that difference is not abstract.

The drone did not replace the ambulance. The paramedics still arrived at 06:53. The clinical handover still happened. What changed was the six minutes between collapse and first shock. Those six minutes are the entire intervention.

This scenario is synthetic. The SA2 is real. The survival curve is evidence-based. The question it raises is not.

The preface to this report is a first-person account of two GoodSAM activations in Sydney, both without a reachable defibrillator in time. That chapter is the lived experience. This one is the programme evidence that sits underneath it. Trained-bystander-responder systems are already operational in NSW and Victoria. They work. They also reveal, with some precision, the gap that an AED-delivery layer is meant to close.

What GoodSAM is

GoodSAM — Good Smartphone Activated Medics — is a smartphone application that notifies registered off-duty responders (paramedics, nurses, doctors, CPR-trained laypeople) when a suspected cardiac arrest is reported to the emergency-call service within a defined radius. The responder can accept, decline, or ignore the alert. On acceptance, the app routes them to the patient and, where relevant, to the nearest registered public-access AED. Dispatch of the ambulance is parallel, not contingent on the alert. The app originated in the United Kingdom and has since been adopted by several Australian state ambulance services, as well as services in Europe and North America.

Where it is deployed

In Victoria, Ambulance Victoria launched GoodSAM in 2018 alongside a public AED registry [11]. In New South Wales, NSW Ambulance launched the programme on 23 November 2023; at the 12-month review it had 8,372 registered responders with 1,745 recorded on-scene arrivals [12]. By early 2026 the NSW roll had grown past 13,500 registered responders and the service had publicly credited the programme with 100 lives saved [13]. Comparable systems run in London, the English East Midlands, Stockholm, Copenhagen, and parts of the Netherlands and Germany.

What the evidence says about clinical impact

Three peer-reviewed signals are worth citing directly. A Stockholm randomised trial of mobile-phone dispatch found that bystander-initiated CPR rose from 48% in the control arm to 62% in the intervention arm — a 14-point absolute increase from a single software layer [14]. A 2022 analysis of GoodSAM across London and the English East Midlands (5,237 confirmed OHCAs) reported that when a responder accepted an alert, the adjusted odds of survival to hospital discharge were 3.15× in London (95% CI 1.19–8.36) and 3.19× in the East Midlands (95% CI 1.17–8.73) [15]. The most recent Victorian evidence, a 2018–23 cohort of 9,196 OHCAs, found that when a smartphone-activated volunteer arrived before EMS, the patient was 8× more likely to receive bystander CPR, 16× more likely to receive bystander defibrillation, and had 37% higher odds of survival to hospital discharge [16]. Across the same two decades in Victoria, the share of OHCAs receiving bystander CPR rose from 40.3% in 2003–04 to 72.2% in 2021–22; Utstein-comparator survival to discharge roughly tripled [11]. Earlier Victorian work on the AED-registry layer sits alongside these findings as a precedent for public-access defibrillation in the same jurisdiction [17].

When a responder accepted an alert, the adjusted odds ratio for survival to hospital discharge was 3.15 in London and 3.19 in the East Midlands.

Smith et al., European Heart Journal — Acute Cardiovascular Care [15]

The Sydney gap this paper addresses

Two points matter for the rest of the argument. The first: the responder layer is built. A trained person can be in the hallway inside three to five minutes of an alert in a well-populated part of Sydney, as Chapter 01 records. The second: what that responder cannot do, in more than seven out of ten Sydney residential arrests, is arrive with a defibrillator in their hand. Public-access AEDs cluster in the places where GoodSAM responders are least likely to need them — commercial strips, transport interchanges, sports venues — and are behind locked doors at night and in the early morning, which is when a large share of residential events occur [2]. The outcome gap is not awareness or willingness. It is payload delivery.

How drone-AED delivery complements, rather than replaces

GoodSAM gets the responder to the scene. A drone-AED layer gets the defibrillator to the scene. The two vectors run in parallel and arrive independently; neither one replaces the ambulance, and neither one replaces the other. International evidence has already shown that airborne AED delivery can beat EMS to the patient in more than two-thirds of suspected OHCAs in a live operational setting [18], and at least one documented save now exists in which the drone-delivered device was used by a bystander before EMS arrival [5]. The question this paper poses is narrower and more local: given that Sydney already has a trained-responder layer and an AED registry, what would it take to add a dispatched-defibrillator layer, and what would the incremental survival benefit be in the residential geography where GoodSAM is already active?

That is the question the chapters that follow address.

Out-of-hospital cardiac arrest does not strike uniformly. It clusters — driven by age, housing type, and social disadvantage. To estimate where burden falls across Greater Sydney, we built a synthetic demand surface across 37 SA2 areas, selected for geographic spread and demographic diversity [2,10].

NSW records approximately 4,800 OHCA events per year in Greater Sydney [ESTIMATED from state-level reporting]. No publicly available dataset geocodes these events to suburb level. We allocated the annual total across SA2 areas using a composite proxy score built from five demographic indicators drawn from ABS 2021 Census data: proportion of residents aged 65 and over, private dwelling count, population density, Index of Relative Socio-Economic Disadvantage decile, and percentage of detached housing [10]. Each indicator was normalised and weighted. The resulting event counts are synthetic — they reflect demographic plausibility, not verified incident locations [SYNTHETIC].

The five highest-burden SA2 areas under this model:

A pattern emerges. The areas carrying the highest estimated burden share three traits: older population profiles, a high proportion of detached housing, and moderate-to-high socioeconomic disadvantage. Detached housing matters because it increases the distance between a collapsed patient and any arriving resource. Older populations carry elevated cardiac arrest risk. Disadvantage correlates with higher rates of chronic disease and, in many areas, longer ambulance response times.

These top five SA2s alone account for roughly 814 of the 4,800 annual events — about 17% of the total demand concentrated in 14% of the study areas [SYNTHETIC].

A transparency note on one result. Wahroonga-Warrawee ranks third with 162 synthetic events, yet it is one of Sydney’s most affluent suburbs — SEIFA IRSAD score of approximately 1,095, placing it in the 9th decile nationally [OBSERVED]. Its presence near the top of the list may appear counterintuitive, but it is a direct and defensible consequence of how the proxy model is weighted. In the base scenario, socioeconomic disadvantage carries a 15% weight. The remaining 85% is driven by age structure (35%), dwelling count (25%), population density (15%), and detached housing proportion (10%). Wahroonga has a notably older resident population, a high count of private dwellings, and a suburban fabric dominated by detached houses on larger lots — all of which are genuine, well-documented OHCA risk factors regardless of household income [2,10]. Cardiac arrest does not discriminate by affluence. An older person who collapses in a detached home in Wahroonga faces the same time-to-defibrillation challenge as one in Campbelltown. The model reflects this. We acknowledge that this ranking is untested against real incident data, and the model may overestimate burden in areas where protective socioeconomic factors — access to preventive cardiology, lower smoking rates — partially offset demographic risk. Validation against de-identified NSW Ambulance records would resolve this question. We flag Wahroonga explicitly not as a weakness in the model but as evidence that it is capturing clinical risk factors rather than functioning as a proxy for disadvantage alone.

This demand surface is not dispatch-grade. It cannot tell us which street will see the next cardiac arrest. What it can do is identify the zones where drone positioning would intercept the greatest number of events, given demographic realities. For feasibility scoping — deciding where to place a small number of bases and estimating population-level coverage — this resolution is sufficient. Operational deployment would require geocoded incident data from NSW Ambulance, a step beyond the boundary of this analysis.

The surface is a planning tool. It tells us roughly where to look, not precisely where to land.

A drone on a launchpad covers nothing. Coverage depends on where bases are placed and how far each drone can fly within the critical response window. We modelled base placement using a greedy set-cover algorithm: place the first base where it captures the most OHCA events, then place each subsequent base to maximise the incremental gain [ESTIMATED].

Using a 6 km service radius as the base scenario:

The full coverage-by-base series across conservative / base / optimistic scenarios is carried in outputs/tables/coverage_vs_base_count_table.csv.

The first two bases deliver 21.3% coverage. Doubling to five bases does not double coverage — it reaches 40.0%. By ten bases, each additional site adds only two to three percentage points. Diminishing returns set in after five.

We tested three radius scenarios. Under a conservative 4 km radius — reflecting headwinds, airspace restrictions, or heavier payloads — coverage contracts at every fleet size. Under an optimistic 8 km radius, coverage expands, but the shape of the diminishing-returns curve stays the same. Across all scenarios, service radius is the dominant parameter. Whether a drone can reach 6 km or only 4 km matters more than whether the network has seven bases or ten. This finding directs attention toward aircraft performance and airspace access as the binding constraints on real-world effectiveness.

All radii are geometric — great-circle distances from base to SA2 centroid. They do not account for controlled airspace boundaries, terrain obstacles, weather, or flight-path routing. Operational coverage will be smaller than these estimates.

Staged rollout

The coverage model supports a staged deployment — not because caution is a virtue in itself, but because it is the fastest way to generate usable evidence.

Stage 1 positions two bases in the highest-burden SA2 clusters. Two bases cover an estimated 21.3% of annual events [ESTIMATED] — enough to generate statistically meaningful response-time data within 12 to 18 months. This stage validates flight operations, dispatch integration, and real-world delivery times against the geometric estimates.

Stage 2 extends to five bases, guided by Stage 1 operational data. If actual service radii prove shorter than 6 km, the algorithm re-optimises placement. If certain areas show access constraints — controlled airspace near Kingsford Smith, for instance — the model reallocates those bases. Five bases at validated parameters are worth more than ten placed on unchecked assumptions.

Stage 3 scales toward ten bases only if Stage 2 demonstrates worthwhile incremental gains at confirmed operational radii.

This is risk reduction through sequencing. Each stage produces data that sharpens the next.

Cardiac arrest survival is a race measured in minutes. The relationship is not linear. It is exponential decay.

The model follows the form P = 0.67 x exp(-0.10 x t), where P is survival probability and t is minutes from collapse to first defibrillation. This draws on Valenzuela et al. 1997 [4] and Larsen et al. 1993 [19], both establishing the steep, time-dependent decline in survival for ventricular fibrillation and pulseless ventricular tachycardia. It is one of the most replicated findings in resuscitation science.

The numbers at specific response times:

At 4 minutes — the modelled drone delivery target — survival probability sits at roughly 45%. At 8.5 minutes, the median ambulance response in the base scenario, it drops to approximately 29%. By 11 minutes, a plausible outer bound for suburban response, the figure falls to around 22%. Every additional minute costs roughly 7-10 percentage points. The curve does not negotiate. [ESTIMATED]

Bystander defibrillation before EMS arrival was associated with a 30-day survival rate of 53 percent, compared with 16 percent when defibrillation was delayed until EMS arrival.

Hasselqvist-Ax et al., NEJM [9]

Not all cardiac arrests benefit equally from rapid defibrillation. Only about 22% of OHCA patients present with shockable rhythms (VF/VT) on first assessment [2]. The remaining cases require CPR and advanced life support but will not convert with a shock alone. The lives-saved estimates below apply only to that shockable subset.

Three response-time scenarios produce these annualised estimates within the five-base coverage zone (approximately 40% of Greater Sydney’s modelled demand surface) [ESTIMATED]:

• Low scenario (drone at 5.5 min, ambulance at 10 min): 59 additional lives per year
• Base scenario (drone at 4 min, ambulance at 8.5 min): 69 additional lives per year
• High scenario (drone at 3 min, ambulance at 7 min): 69 additional lives per year

These figures apply the survival uplift only to the shockable-rhythm cases within the network’s coverage zone — not to the total Greater Sydney population. With full metropolitan coverage (requiring more bases), the theoretical maximum would be approximately 172 lives per year, but that configuration is beyond the scope of a staged pilot.

These are modelled estimates, not observed outcomes. The survival function is well established, but the decay rate parameter (0.10 per minute) is derived from international studies, not calibrated to NSW conditions. Varying that parameter by +/-20% would shift the base-case lives-saved estimate by approximately +/-25%, producing a range of roughly 52 to 86 lives per year. The response times are also assumptions, built from published platform specifications and ambulance performance data.

Bystander retrieval sensitivity. The figures above assume that every drone delivery results in successful bystander retrieval and AED application — a 100% retrieval rate. This has been demonstrated once in a real-world case [5] but has not been tested at scale. If bystander retrieval succeeds in only 50% of deliveries, the base-case lives-saved estimate falls to approximately 35. At 30% retrieval, it falls to approximately 21. Even at pessimistic retrieval rates, the intervention produces meaningful impact — but the assumed retrieval rate is the single largest source of uncertainty in the survival model [ESTIMATED].

Actual benefit depends on operational performance: whether the drone launches, whether the AED is retrieved, whether a bystander applies it correctly.

External evidence supports the direction. A Swedish registry study of 30,000+ cases found that bystander defibrillation before EMS arrival produced 53% 30-day survival, compared with 16% when no bystander defibrillation occurred [9]. In 2021, the Karolinska/Everdrone AED delivery pilot recorded the first documented case of a drone-delivered AED saving a life [5]. That single case does not prove scalability. It confirms the mechanism works outside a simulation.

The gap between 29% and 45% is the window this proposal targets.

The indicative cost for a five-base network over five years is approximately $1.5 million [ESTIMATED]. This covers platform acquisition, AED units, ground infrastructure, maintenance cycles, and estimated operating expenditure. It does not include regulatory compliance costs, dispatch system integration, community engagement, or operational overhead that a funded pilot would budget separately.

At the base scenario of 69 lives saved per year within the coverage zone, this produces a cost per life saved of roughly $4,500 and a cost per quality-adjusted life year of approximately $450 [ESTIMATED].

Those numbers look favourable. They deserve scrutiny rather than celebration.

For context: published evaluations of public-access defibrillation programmes report costs in the range of $30,000 to $55,000 per QALY [20]. The drone model appears to outperform those benchmarks substantially. That gap should prompt caution. It likely reflects what the model does not yet capture.

The hardware cost inputs are drawn from published platform specifications and manufacturer pricing where available. They are feasibility-grade numbers, not procurement quotes. Opex assumptions come from operational analogues, not contractual rates. A funded pilot would produce different numbers once real variables — staffing requirements, insurance, maintenance contracts, airspace compliance — are factored in.

Several cost categories sit outside the current model entirely. CASA regulatory engagement for BVLOS approvals requires specialist aviation consulting. Dispatch integration with Triple Zero and state ambulance services involves technical and contractual work that cannot be costed from published data. Community engagement to support bystander retrieval is a programme cost, not a hardware cost.

The full base-by-base cost-versus-coverage series is carried in outputs/tables/cost_vs_coverage_table.csv, with lifecycle breakdown in outputs/tables/lifecycle_cost_table.csv and sensitivity envelope in outputs/tables/sensitivity_analysis.csv.

The honest framing: the model suggests the intervention is likely cost-effective relative to existing defibrillation access programmes, and the margin appears large enough to absorb significant cost escalation before the value proposition collapses. But the current figures are a planning envelope, not a budget. They should inform the decision to invest in a pilot. They should not be quoted as operational costs.

Three drone platforms and two AED units form the shortlist from the device evaluation conducted for this study. Each was scored against a weighted heuristic covering the operational factors most relevant to suburban AED delivery.

Drone Candidate Evaluation

Candidates were scored across six dimensions: payload capacity (25%), range (20%), regulatory readiness (20%), operational maturity (15%), maintainability (10%), and cost (10%). Each dimension was scored 1–5 and weighted to produce a composite score. The weighting reflects a feasibility-stage priority: the platform must carry the AED (payload), reach the patient within the service radius (range), and operate legally in Australian airspace (regulatory readiness).

DJI FlyCart 30. A heavy-lift multirotor designed for industrial cargo operations. The FlyCart 30 carries up to 30 kg — comfortably accommodating any AED with payload margin — with a practical range of 16 km and a cruise speed of 72 km/h. It has the strongest operational maturity record in industrial settings and benefits from DJI’s established maintenance and support ecosystem. Its limitation is range: at 16 km, it covers a smaller service radius than fixed-wing alternatives. Estimated unit cost: US$90,000. [OBSERVED]

Wingcopter 198. A fixed-wing VTOL platform purpose-built for medical logistics. The Wingcopter 198 carries 6 kg to 75 km range at 100 km/h cruise speed, offering the widest coverage radius from a single base of any candidate. It holds medical logistics certifications in multiple international markets. The trade-off is cost (approximately US$250,000 per unit) and a specialist maintenance footprint that requires dedicated technical staff. [OBSERVED]

Swoop Aero KITE. Designed for sustained healthcare logistics in distributed networks, particularly remote and low-infrastructure corridors. The KITE carries 4 kg to a maximum range of 175 km at 120 km/h. It has operational history through UNICEF partnerships in Malawi and the Pacific [21]. Its regulatory readiness in Australian suburban airspace is less established than the other candidates. Estimated unit cost: US$150,000. [OBSERVED]

The full scored platform comparison, including per-criterion scores and the composite weighting, is carried in outputs/tables/drone_comparison_table.csv (with the final shortlist summary in outputs/tables/device_shortlist_final.csv).

All three platforms are VTOL-capable, meaning they can launch and land without a runway — a requirement for suburban base operations. None has been evaluated against CASA Part 101 BVLOS requirements for Australian suburban airspace. The cost figures are pre-procurement estimates and should be treated as indicative only.

AED Candidate Evaluation

AED candidates were scored across five dimensions: bystander usability (30%), regulatory readiness (20%), maintainability (20%), cost (15%), and recall history (15%). The heavy weighting on usability reflects the operational reality: the AED will be used by untrained bystanders under extreme stress, not by clinicians.

ZOLL AED 3. Weighs 2.5 kg with IP55 environmental protection. Rated for untrained bystander use with real-time CPR feedback via voice and visual prompts. Integrated paediatric mode reduces the number of electrode configurations that need to be stocked. Five-year battery and electrode shelf life minimises maintenance burden. Estimated unit cost: US$2,300. [OBSERVED]

Stryker LIFEPAK CR2. Weighs 2.0 kg with IP55 rating. Provides adaptive audio guidance that adjusts instruction pacing based on user interaction. Bystander-rated with strong global distribution and service network. Four-year battery and electrode life. Higher unit cost at approximately US$2,500, but the strongest vendor support infrastructure of any candidate. [OBSERVED]

The full AED candidate comparison is carried in outputs/tables/aed_comparison_table.csv.

Both AEDs are within the payload envelope of all three drone candidates. Both are TGA-listed or equivalent-family listed for the Australian market. The selection between them is a procurement decision that should be deferred to the pilot design phase, where hands-on bystander testing can inform the choice.

Operational Precedents

Two programmes provide the closest analogues to what is proposed here.

The AED was delivered by drone to the scene before the ambulance arrived, and was used by a bystander to deliver a shock that restored a normal heart rhythm.

Schierbeck et al., New England Journal of Medicine [5] — describing the Gothenburg incident

The Everdrone/Karolinska AED delivery pilot in Gothenburg, Sweden, integrated drone AED dispatch with the national 112 emergency system. Drones were dispatched simultaneously with ambulances to suspected OHCA. In at least one documented case, the drone arrived before the ambulance and a bystander retrieved and applied the AED, resulting in survival [5,22]. The median drone arrival was three minutes ahead of ambulance in pilot areas.

Zipline’s national medical drone logistics network in Rwanda has operated continuously since 2016, completing hundreds of thousands of deliveries of blood products and medical supplies [6]. Rwanda’s airspace regulatory environment differs significantly from Sydney’s, but the programme demonstrates that sustained autonomous medical drone logistics is operationally feasible at national scale.

Regulatory Context

CASA Part 101 provides for BVLOS approvals but requires case-by-case assessment against specific operational risk profiles [23]. No blanket pathway exists. Sydney’s proximity to Kingsford Smith Airport and established helicopter routes creates a regulatory context with no direct international precedent. International approvals may inform but will not determine Australian outcomes.

The estimates above depend on assumptions untested in Australian suburban conditions. Some carry enough uncertainty to invalidate the modelled benefits entirely. Acknowledging this is not a weakness. It is the reason a funded pilot, rather than direct deployment, is the appropriate next step.

CASA BVLOS approval (high risk). The concept requires beyond-visual-line-of-sight drone operations in suburban airspace. CASA Part 101 permits BVLOS under case-by-case assessment, but the Swedish and Rwandan precedents may not transfer to Greater Sydney’s controlled airspace [23]. Sydney’s proximity to Kingsford Smith Airport, helicopter routes, and dense suburban terrain creates a regulatory context with no direct precedent. Mitigation: early engagement with CASA, beginning with a scoped pre-application discussion for a geographically bounded trial corridor.

Bystander retrieval (high risk). The survival model assumes someone at the scene collects the AED from the drone and applies it. This has been demonstrated once in a documented real-world case [5]. It has not been tested in Australian settings, with Australian bystander populations, under Australian dispatcher protocols. If the AED arrives but nobody applies it, the intervention produces no benefit. Mitigation: dispatcher-assisted retrieval protocols, where the 000 call-taker coaches the bystander through collection and pad placement in real time. Community awareness programmes alongside any pilot.

Weather and environmental reliability (medium risk). Wind, rain, and temperature extremes reduce drone range, stability, and battery performance. Published specifications describe optimal-condition envelopes. Suburban Sydney weather is generally mild but includes storm events and high winds that may ground flights. Mitigation: define a weather operating envelope with clear stand-down thresholds. Accept that the network will not achieve 100% availability and model expected downtime.

Synthetic demand model (medium risk). The OHCA demand surface is a proxy model built from demographic indicators. It does not use real incident locations. Coverage claims are geometric, not flight-path validated [SYNTHETIC]. Mitigation: validate against de-identified NSW Ambulance cardiac arrest data before operational commitment.

Cost estimates (medium risk). All cost figures are pre-procurement. Real costs will differ, and the current model likely understates total programme cost. Mitigation: treat the cost envelope as a planning range, not a budget. Conduct detailed costing during pilot design.

Each risk has a corresponding validation gate. CASA engagement resolves regulatory feasibility. A controlled trial resolves retrieval assumptions. Weather logging resolves availability. Registry data access resolves the demand model. Vendor engagement resolves cost.

The appropriate response to these uncertainties is not to wait for perfect data. It is to design a pilot that resolves them.

The clinical science of out-of-hospital cardiac arrest is clear about where the survival gains sit: the first few minutes, before the ambulance arrives. Research funding does not reflect that geometry. Dollar for dollar and paper for paper, the money flows to the hospital end of the chain — to mechanical CPR, extracorporeal life support, drug protocols, advanced airway management, and post-ROSC neuroprotection — while the defibrillation link, the single biggest survival lever, receives a fraction of the attention.

This is not an attack on in-hospital research. Those advances matter. The argument is about marginal returns: the same dollar, spent earlier in the chain, buys more survival.

Post-arrival research dominates the portfolio

A decade-long analysis of NIH-funded cardiac arrest grants found that cardiac arrest research received roughly $91 per annual US death, against $2,100 per death for heart disease and $2,200 for stroke — a twenty-fold disparity at the headline level [24]. Within that already-small envelope, the allocation is skewed again. A structured audit mapping NIH cardiac arrest grants against the American Heart Association’s top ten science gaps found that “the majority of grants have focused on optimization of post-cardiac arrest care (78 grants; 45.3%), prediction of patients at risk of cardiac arrest (12; 7%), and developing tools for early neuroprognostication (7; 4.1%),” while dispatcher-directed CPR — a gap with clear survival upside — received no funding at all [25]. A follow-up study using the chain-of-survival framework reached the same conclusion: the later links dominate the spend [26].

The pipeline analysis that followed showed why the imbalance persists. Stroke research, a comparable time-critical cardiovascular condition, attracts roughly ten-fold more federal grants and six-fold more mentored K awards than cardiac arrest research, thinning the pool of early-career investigators able to push the pre-arrival agenda [27]. The Australian picture echoes the US one. An analysis of NHMRC grants (2013–2023) and Heart Foundation grants (2020–2024) found NHMRC allocated roughly 60% of its cardiac arrest envelope to prevention, with the Heart Foundation directing 43% of its smaller pool to pre-hospital care — Australia still spends about AUD $887 per cardiac-arrest death on research, against AUD $4,673 per breast-cancer death [28].

The annual NIH investment in cardiac arrest research is low relative to other leading causes of death in the United States and has declined over the past decade.

Coute et al., J Am Heart Assoc (2017) [24]

Pre-arrival interventions remain under-studied

Publication volume tells the same story as grant allocation. A PubMed query run for this report (2026-04-21) returned roughly 1,606 papers indexed under “extracorporeal cardiopulmonary resuscitation” and 1,299 papers on “targeted temperature management” in cardiac arrest. The same database returned 387 papers on “public access defibrillation” and 136 on “dispatcher-assisted CPR.” Bystander-applied defibrillation — the intervention that most directly determines whether a patient in a shockable rhythm survives — returned 79 papers. The ratio of post-arrival to pre-arrival research attention runs at roughly 10:1 to 20:1 depending on the keyword pair.

The clinical consequence of this imbalance is measurable. Each minute of delay to defibrillation in witnessed ventricular fibrillation costs 7–10% of absolute survival [4,19]. Bystander defibrillation before EMS arrival is associated with 53% 30-day survival for shockable rhythms; without it, 16% [9]. Public-access defibrillation programmes are cost-effective when AED density matches incident density — and demonstrably ineffective when retrieval distance exceeds a few hundred metres [20,29]. The 2015 Institute of Medicine review drew the conclusion explicitly: community response capacity is the under-invested link in the chain of survival [30].

The marginal-return case for correcting the skew

The marginal-return argument is straightforward. An extra dollar spent on eCPR protocols reaches patients who are already deep in a low-probability state; the effect size is small, and the downstream costs — ECMO, ICU days, neurorehabilitation — are large. An extra dollar spent on getting a working defibrillator to the patient within three minutes of collapse reaches patients whose survival odds are still high and drops the downstream cost curve at the same time. The 2015 IOM review put the same logic in policy terms: national survival will not move without investment in the pre-arrival links [30]; Callaway’s 2017 editorial called for targeted grant mechanisms to correct the skew rather than simply enlarging the pot [31].

Where drone-AED delivery fits

Drone-delivered defibrillation is best read as a corrective investment in this context. It does not displace in-hospital research; it addresses the specific link — time-to-defibrillation in residential settings — where the survival gradient is steepest and the research spend has been thinnest. A pilot of the scale proposed here (two bases, AUD $2–3M) sits well inside the noise floor of the Australian cardiac-arrest research budget and targets the single intervention with the largest evidenced per-minute survival effect. The ask is not a new priority but a rebalancing toward the part of the chain the literature has, for structural reasons, left under-studied.

We propose a funded 12-month pilot to test whether the modelled performance holds under real operational conditions. Two bases. Two high-burden areas. One clear question: do drones arrive faster than ambulances, and can bystanders use the AEDs they deliver?

The pilot sites should be Campbelltown and Fairfield. Both SA2 areas rank high on socioeconomic disadvantage, have elevated OHCA incidence estimates, and contain large concentrations of detached housing — the residential type where public AED access is lowest and ambulance navigation times are longest [10]. These are not cherry-picked for easy wins. They represent the conditions where drone delivery faces its real test: suburban streets, backyard landings, residents who have never handled a defibrillator.

The pilot requires integration with NSW Ambulance dispatch. Drones must launch on the same 000 call that triggers paramedic response, not on a parallel research protocol. Without dispatch integration, arrival-time comparisons are meaningless. This is a non-negotiable design requirement and the single largest coordination challenge.

Estimated budget: $2-3M [ESTIMATED]. That covers drone hardware and maintenance for two bases, CASA regulatory approval and ongoing compliance, operational staffing, and independent clinical evaluation. The evaluation component is not optional — it must be designed and run by researchers with no stake in the programme’s continuation.

The pilot would validate three things. First, actual drone arrival time versus ambulance arrival time across a statistically meaningful number of dispatches. Second, bystander retrieval success rate — what proportion of recipients collect the AED and attach it. Third, preliminary clinical outcomes, recognising that a 12-month sample from two bases will lack power for definitive survival analysis but can establish operational viability.

If the pilot fails to demonstrate a meaningful time advantage, or if bystander retrieval rates fall below usable thresholds, the programme stops. If it succeeds, the data supports a case for scaled deployment built on evidence from Australian conditions rather than extrapolation from international trials.

We are available for a working session to scope the pilot design with NSW Ambulance, CASA, and clinical stakeholders.

This chapter documents the derivation of the synthetic out-of-hospital cardiac arrest (OHCA) counts carried in data/derived/synthetic_ohca_hotspots.csv. The file is a proxy dataset that stands in for real NSW Ambulance incident data until that feed lands in the project; every number downstream of this chapter — the Chapter 3 spatial maps, the Sprint 20 choropleth and proportional-symbol figures, the base-placement shortlist inputs — descends from the formula set out here. The output is therefore explicitly pending verification against NSW Ambulance data access; this status is propagated through the CSV’s data_label column (value synthetic) and through the run log.

Motivation

An earlier version of the synthetic cohort (Sprint 01 checkpoint, later retained through Sprints 16–20) ranked SA2s by an age-dominated composite proxy. That proxy ordered three wealthy northern-shore SA2s — Chatswood (East)–Artarmon, Hurstville, and Wahroonga–Warrawee — at the top of the list. Clinically, that ordering is implausible: OHCA incidence is driven by age and socio-economic disadvantage, not by age alone. Wealthy suburbs carry older age structures but markedly better cardiovascular health profiles (higher preventive-care uptake, lower smoking prevalence, better access to primary care) than the disadvantaged south-west corridor — Fairfield, Cabramatta, Liverpool, Bankstown, Mount Druitt, Campbelltown — where absolute OHCA counts should be higher. Sprint 26 replaces the old composite with an explicit SEIFA-weighted formula so the proxy reflects this.

Formula

For each SA2 i:

OHCA_i ∝ population_i × age65+ share_i × IRSD weight_i

where the IRSD weight is a monotonically decreasing function of the ABS 2021 SEIFA Index of Relative Socio-economic Disadvantage (IRSD) decile [32]:

$$\text{IRSD weight}_i = \frac{11 - \text{IRSD decile}_i}{10}$$

Decile 1 (most disadvantaged) maps to a weight of 1.0; decile 10 (least disadvantaged) maps to 0.1. A linear decile weighting is chosen for v1 because it is interpretable, monotone, and does not require committing to a specific published effect-size estimate — the published OHCA-by-deprivation gradient literature reports relative risks in the 1.5×–3× range between most- and least-advantaged quintiles (Australian and international cohorts), and a linear decile weighting sits conservatively inside that envelope. A log-linear or quintile-step weighting can be substituted when the NSW Ambulance cohort is joined in and the empirical gradient can be measured directly.

The product population × age65+ share gives the expected count of at-risk person-years; the IRSD weight then modulates that by cardiovascular-risk profile. Formally the proxy is a multiplicative hazard on person-years, not an age-standardised rate — the latter requires the NSW Ambulance event-level data.

Scaling and base-rate anchor

The per-SA2 raw weight is turned into an integer event count by linear scaling:

$$\text{scale} = \frac{T}{\sum_i \text{pop}_i \times \text{age65+}_i \times \text{IRSD}_i} \qquad \text{OHCA}_i = \text{round}\left(\text{raw}_i \times \text{scale}\right)$$

The target total T = 4 200 cases per year is chosen so the sum across the 37 project SA2s lands inside a [3 500, 5 000] envelope consistent with:

• Hasselqvist-Ax et al., 2015 [9] — the Swedish national registry paper reports a crude OHCA incidence of approximately 100 cases per 100 000 person-years, the closest well-characterised incidence we trust without a bespoke Australian fit.
• NSW Ambulance Annual Report 2022–23 [3] — reports roughly 4 800 adult-attended OHCA events per year across the Greater Sydney footprint. The project 37-SA2 subset covers a population substantially smaller than the full metro, so a 4 200 target is the upper-mid point of the envelope after the ABS 2021 Greater Sydney denominator [10] is allowed for.

Tessa’s Sprint 26 test test_csv_total_count_is_within_nsw_ambulance_envelope fails the build if the emitted total falls outside the [3 500, 5 000] window — the invariant is enforced, not just documented.

Inputs and provenance

All inputs live in data/curated/sa2_greater_sydney.csv and are labelled observed+estimated in that file’s data_label column. The “estimated” component reflects that the SEIFA decile values carried in the curated CSV are the project’s best reconciliation of the ABS IRSD SA2 datapack pending a full re-fetch; values flagged pending-verification in the curated table would cascade into the synthetic CSV as zero-weighted rows rather than hand-filled numbers.

Top-5 SA2s after re-ranking

The re-ranked top of the table shifts decisively into Sydney’s south-west — the pattern Roger anticipated in the Sprint 26 brief when he observed that “there are lots of wealthy boomers in the northern suburbs, but unhealthy people in the west / south west.”

All five SA2s sit in IRSD decile 1 or 2 (most disadvantaged). Wahroonga–Warrawee (IRSD decile 9) falls from rank 3 to rank 33, Chatswood (East)–Artarmon (decile 8) falls from rank 1 to outside the top-10, and the absolute counts redistribute mass toward the south-west — visible in the Chapter 3 proportional-symbol map once it is regenerated from this CSV.

Limitations and pending verification

1. Ecological fallacy. A monotone IRSD weight captures the average deprivation gradient across an SA2 but cannot resolve within-SA2 variation. Empirical Bayes smoothing at SA1 granularity is the right refinement, and is deferred to the sprint that consumes the NSW Ambulance incident data.
2. Age mid-point only. The proxy uses the 65+ share as a single breakpoint. The real OHCA age curve is steeper in the 75+ band; once the NSW Ambulance data is joined, per-year age bins should replace the 65+ share.
3. No sex stratification. Male OHCA incidence is roughly 1.7× female in the comparable age bands [9], but the ABS SA2 sex split is close to 50/50 in the 37 project SA2s so the aggregate ranking is insensitive to this. The model should still be re-fit once a stratified NSW Ambulance series is available.
4. Linear decile weighting is a placeholder. The v1 linear weighting is deliberately conservative and has no empirical fit. When the NSW Ambulance cohort lands, re-fit f(IRSD) directly on the joined dataset.
5. Output remains a proxy. Every downstream figure, table, and shortlist entry must carry the “synthetic cohort, pending NSW Ambulance data access” caveat until the real feed replaces this CSV. The Chapter 3 spatial-analysis caveat paragraph points readers back to this chapter.

Reproduction

The CSV is regenerable from the curated inputs with a single-file Python routine:

count_i = round(
    population_i * (age_65_plus_pct_i / 100) *
    ((11 - irsd_decile_i) / 10) * scale
)

where scale is calibrated so the integer totals sum to the target T = 4 200. Tessa’s Sprint 26 test test_counts_are_consistent_with_formula inverts this — it re-derives the scale from the dataset’s own totals and fails the build if any row deviates from formula prediction by more than one count.

Image Credits

This appendix records every image that appears in LIFT Study 01, its licence, and its source. It is generated from content/assets/images/manifest/image-manifest.json by scripts/build-image-credits.js; manual edits here will be overwritten on the next build. Corrections flow through the manifest.

Every image also carries alt text in the manifest; that alt text is wired into the published HTML and paged.js PDF via Paige’s templates so that screen-reader users receive a specific, clinical-register description of each figure’s content.