Location-resolved performance of public-access defibrillation in an urban EMS system: a population-based study in Yokohama, Japan

Highlights

• •PAD shock was significantly associated with favourable neurological outcomes.
• •PAD shock rates varied substantially across different urban environments.
• •Favourable outcomes require comprehensive bystander performance, not just PAD.
• •A location-resolved framework can guide targeted urban resuscitation strategies.

Abstract

Background

Survival rates after out-of-hospital cardiac arrest (OHCA) in Japan have plateaued despite widespread automated external defibrillator (AED) deployment, suggesting a gap between availability and real-world effectiveness. We aimed to examine the association between public access defibrillation (PAD) shock and neurological outcomes and evaluate location-specific patterns of PAD shock rates.

Methods

We retrospectively analysed data from the Yokohama City OHCA registry from 2021 to 2024. Arrest locations were categorised. PAD shock rates and a composite bystander performance score were calculated to characterise location-level resuscitation performance. The primary outcome was favourable neurological outcome (Cerebral Performance Category score, 1–2) at 30 days. Propensity score matching was performed using age, sex, witnessed arrest, time of arrest, and dispatcher-assisted cardiopulmonary resuscitation to balance baseline characteristics between patients who did and did not receive a PAD shock. Logistic regression models evaluated the association between PAD shock and neurological outcomes before and after matching.

Results

Among 15,345 patients, 8828 (57.5%) were male and 272 (1.8%) received a PAD shock. After propensity score matching, PAD shock was significantly associated with favourable neurological outcome (OR 3.3, 95% CI: 1.9–5.7) and 30-day survival (OR 3.7, 95% CI: 2.2–6.1). PAD shock rates varied markedly across arrest locations, ranging from 34.0% (18/53) in sports facilities and 28.0% (30/107) at railway stations to 0.06% (6/10,426) in private residences. Crucially, favourable neurological outcomes across locations were not merely a function of PAD shock rates but were strongly aligned with high composite bystander performance scores.

Conclusions

PAD shock is independently associated with favourable neurological outcomes after OHCA. However, its real-world effectiveness is shaped by urban structural contexts and relies on high-quality, comprehensive bystander interventions. A location-resolved performance framework may serve as a practical tool for guiding targeted resuscitation strategies beyond uniform AED deployment.

Introduction

Public access defibrillation (PAD) programmes encouraging laypeople to use automated external defibrillators (AEDs) can substantially improve outcomes after out-of-hospital cardiac arrest (OHCA).¹ Despite widespread dissemination of AEDs and strong evidence supporting PAD effectiveness, improvements in population-level outcomes have been inconsistent. In Japan, nationwide 30-day survival following witnessed OHCA has plateaued at approximately 12% over the past decade (2011–2021), despite increasing AED deployment.² This discrepancy suggests that increasing device availability alone may be insufficient to translate PAD efficacy into real-world effectiveness. Indeed, the rate of PAD shock remains low in many settings, indicating that barriers to access and use play a critical role.³

PAD effectiveness depends significantly on the structural and behavioural context of cardiac arrest occurrence, being more frequent and effective in structured, high-occupancy public environments than in private residences.3, 4 Despite the inherent heterogeneity of urban environments—with variation in pedestrian flow, bystander composition, organisational readiness, and AED accessibility—limited research has examined how PAD shock rates vary across specific urban settings or whether certain environments systematically facilitate or constrain bystander response. This knowledge gap prevents municipalities from identifying where PAD strategies succeed and where targeted interventions are needed. Yokohama City, with its coordinated EMS-hospital system, provides an ideal setting to address these questions. Given the likely persistence of substantial variation in PAD shock across arrest locations within the city, identifying where targeted interventions could yield the greatest improvement in OHCA outcomes remains critical.

Therefore, an analytical framework that captures location-specific variation in resuscitation performance is required. Without such an approach, heterogeneous environments are effectively averaged, potentially obscuring both high-performing settings and those with unrealised potential. Accordingly, this study adopted a two-stage analytical approach within Yokohama City's unified EMS system. The primary aim was to establish a local performance baseline by examining the association between PAD shock and neurological outcomes after OHCA—a necessary reference anchor against which location-specific deviations can be meaningfully evaluated. The secondary aim was to characterise location-specific variation in PAD shock occurrence and bystander performance using a location-resolved framework. We hypothesised that while PAD shock would demonstrate strong overall effectiveness, its real-world performance would vary substantially by arrest location due to structural and contextual factors, enabling identification of priority environments for targeted resuscitation strategies.

Methods

Study design

This was a retrospective cohort study designed to examine the association between PAD shock and neurological outcomes after OHCA, and to evaluate location-specific patterns of PAD shock across urban environments.

Setting

This study was conducted in Yokohama City, a metropolitan area of approximately 3.7 million residents in Japan. We analysed the Yokohama City EMS OHCA Surveillance Database, which prospectively records all EMS-treated OHCA cases between January 2021 and December 2024. EMS respond to all OHCA incidents within the city, ensuring comprehensive case capture. This study was approved by the Institutional Review Board of Yokohama City University Medical Centre (approval number: G2023014; date of approval: 2023/10/13). The requirement for informed consent was waived because of the retrospective use of anonymised data. During the study period, emergency dispatchers routinely provided dispatcher-assisted CPR instructions according to standardised protocols, and callers were encouraged to identify and retrieve a nearby AED when appropriate. However, no smartphone-based first-responder activation system or dispatcher-linked AED location platform was in operation in Yokohama City during this period.

Participants

All patients with OHCA who were transported to hospital between January 2021 and December 2024 were screened for inclusion. This included patients in whom resuscitation was attempted by EMS personnel, as well as those who achieved return of spontaneous circulation following PAD shock prior to EMS arrival and were subsequently transported to hospital. Patients were excluded if they had missing patient identifiers, documented do-not-resuscitate orders, were not treated by EMS (including cases in which death was evident at the scene or resuscitation was not attempted), had resuscitation terminated under medical direction prior to hospital transport, declined hospital transport, or had unknown final outcomes.

Variables

The database included patient demographic characteristics (age, sex), arrest characteristics (witnessed status, time of occurrence, initial cardiac rhythm, and presumed cause of arrest), bystander interventions (CPR and PAD shock), arrest location, and prehospital return of spontaneous circulation (ROSC), recorded according to the Utstein style. Arrest location was categorised into predefined groups based on EMS classifications and further consolidated for analytical purposes. Patients were categorised into two groups based on prehospital defibrillation status. The PAD shock group included cases in which a bystander delivered at least one PAD shock prior to EMS arrival. The non-PAD shock group included all other patients, including those in whom an AED was applied but no shock was advised. The registry does not capture AED pad application without shock delivery; therefore, these cases cannot be distinguished from those without AED contact.

Outcome measures

The primary outcome was favourable neurological outcome at 30 days after OHCA, assessed using the Glasgow-Pittsburgh Cerebral Performance Category (CPC) score of 1–2, whereas those with scores of 3–5 were considered unfavourable, according to Utstein reporting guidelines.⁵ Secondary outcomes were 30-day survival, defined as discharge home or transfer to another healthcare facility, and prehospital ROSC, defined as restoration of circulation prior to hospital arrival achieved by either bystander or EMS intervention.

Covariates

The following variables were selected a priori as potential confounders for propensity score estimation: age, sex, witnessed arrest, time of occurrence, and dispatcher-assisted CPR. These variables were chosen based on prior literature and causal reasoning, as they are associated with both PAD shock occurrence and clinical outcomes. Initial cardiac rhythm was excluded because it is deterministically embedded within the exposure definition itself: AEDs deliver shocks only for shockable rhythms, meaning all patients in the PAD shock group were necessarily in a shockable rhythm at the time of shock delivery. Including initial rhythm would therefore introduce perfect collinearity with the exposure variable. EMS response time was excluded because it represents a process measure rather than a baseline patient characteristic; following standard propensity score methodology, we restricted covariates to pre-exposure baseline variables.

Missing data

Cases with missing data in the primary exposure (PAD shock status) or primary outcome (CPC score) were excluded from analyses. Cases with missing values in covariates used for propensity score estimation were excluded from propensity score matching (PSM) analyses only but retained in unmatched analyses. Missing values in other variables, including secondary outcomes, were handled using complete-case analysis in regression models. Given the low proportion of missing data, complete-case analysis was considered unlikely to introduce substantial bias.

Statistical analyses

Continuous variables were presented as mean ± standard deviation or median (interquartile range), as appropriate, and categorical variables as counts with percentages. Between-group comparisons in the unmatched cohort were performed using Student's t- or the Mann–Whitney U test for continuous variables and chi-square or Fisher's exact test for categorical variables, as appropriate.

To evaluate the association between PAD shock and outcomes, unadjusted analyses were first conducted. Propensity scores were estimated using logistic regression with PAD status as the dependent variable and age, sex, witnessed arrest, time of occurrence, and DA-CPR as covariates.1, 3, 6 One-to-one nearest-neighbour matching without replacement was performed using a calliper width of 0.2 of the standard deviation of the logit of the propensity score.

Covariate balance before and after matching was assessed using standardised mean differences (SMDs), with values <0.1 considered acceptable. Logistic regression analyses were subsequently performed in unmatched and matched cohorts to evaluate the association between PAD shock and outcomes. In the matched cohort, standard logistic regression was applied without additional covariate adjustment, as adequate balance was achieved through matching. Subgroup analyses were conducted for witnessed versus unwitnessed OHCA using the same logistic regression framework as the main analysis. Formal statistical tests for interaction between PAD shock and subgroup variables were not performed due to limited statistical power in individual subgroups. All statistical tests were two-sided, with p < 0.05 considered statistically significant. Analyses were performed using JMP version 18 (SAS Institute Inc., Cary, NC, USA).

Location-resolved performance framework

To complement outcome-based analyses, a location-resolved performance framework was applied to evaluate variation in PAD shock and bystander response across urban environments. A composite bystander performance score was calculated for each arrest location category by standardising (Z-score normalisation) and averaging three components: witnessed arrest rate, bystander CPR rate, and DA-CPR rate. Arrest locations were initially classified into 53 categories according to the fire department system and consolidated into 13 analytical categories (Supplementary Table 1). As no established thresholds exist for defining high or low performance, outcome rates and composite scores were dichotomised using cohort medians. A bubble chart was constructed with 30-day favourable neurological outcome rate on the x-axis and composite performance score on the y-axis, with bubble size representing PAD shock rate. Favourable neurological outcome was selected as the primary metric to reflect patient-centred outcomes.

Results

Participant selection

During the study period, 28,730 patients with OHCA were recorded in the registry. After excluding 13,385 patients in total (13,093 not transported to hospital, 202 who declined resuscitation, and 90 with missing data), 15,345 were included in the analysis. Among these, 272 patients received PAD shock and 15,073 did not. The PSM analysis included 264 patients each in the PAD shock and non-PAD shock groups (Fig. 1).

Fig. 1.

Flowchart of the inclusion and exclusion of OHCA cases in Yokohama between 2021 and 2024.

A total of 28,730 patients with OHCA were screened. After excluding patients based on the criteria shown, 15,345 patients were included in the final analysis.

Abbreviations: OHCA, out-of-hospital cardiac arrest; PAD, public access defibrillation.

Participant characteristics

Baseline patient characteristics are presented in Table 1. Of the 15,345 included patients, 5996 (39.1%) and 9349 (60.9%) had witnessed and unwitnessed arrests, respectively. In the unmatched analysis, the PAD shock group showed higher rates of cardiac arrest during daytime hours (08:00–18:00), witnessed events, DA-CPR, and arrests in public locations, versus the non-PAD shock group. Regarding bystander characteristics, laypeople and healthcare professionals were more frequently present in the PAD shock group, whereas family members were the predominant bystanders in the non-PAD shock group. After PSM, age, sex, time of occurrence, witnessed status, and DA-CPR were well-balanced between groups; however, differences in arrest location and bystander characteristics remained.

Table 1.

Baseline characteristics of patients with OHCA with and without PAD shock in the unadjusted cohort and the propensity score-matched cohort.

	Total (n = 15,345)		Unadjusted				Propensity score-matched
			Public-access defibrillation (n = 272)		No public-access defibrillation (n = 15073)		Public-access defibrillation (n = 264)		No public-access defibrillation (n = 264)		Post-matching SMD
Age, 0–20 years	201	(1.3)	6	(2.2)	195	(1.3)	6	(2.3)	5	(1.9)	0.026
Age, 21–70 years	3663	(23.9)	123	(45.2)	3540	(23.5)	123	(46.6)	117	(44.3)	0.046
Age, 71 years∼	10,796	(70.4)	136	(50.0)	10,660	(70.7)	136	(51.5)	143	(54.2)	0.052
Age, unknown	685	(4.5)	0	(0.0)	0	(0.0)	0	(0.0)	0	(0.0)	−
Men	8828	(57.5)	215	(79.0)	8613	(57.1)	212	(80.3)	211	(79.9)	0.010
Daytime, 8–18	7440	(48.5)	205	(75.4)	7235	(48.0)	200	(75.8)	194	(73.5)	0.052
Witnessed	5996	(39.1)	196	(72.1)	5800	(38.5)	191	(72.3)	188	(71.2)	0.025
Bystander CPR	7857	(51.2)	272	(100.0)	7585	(50.3)	264	(100.0)	264	(100.0)	0.00
Dispatcher assisted CPR	9351	(60.9)	187	(68.8)	9164	(60.8)	182	(68.9)	184	(69.7)	0.016
Type of bystander
Citizen	731	(6)	81	(30)	650	(5.4)	81	(31)	39	(15)
Family	5445	(45)	13	(4.8)	5432	(45)	12	(4.5)	94	(36)
Officer	77	(0.6)	8	(2.9)	69	(0.6)	8	(3.0)	6	(2.3)
Medical staff	1490	(13)	106	(39)	1384	(12)	100	(38)	85	(32)
Type of location
Residence	10,426	(67.9)	6	(2.2)	10,420	(69.1)	5	(1.9)	44	(16.7)
Public places	3671	(23.9)	194	(71.3)	3478	(23.1)	187	(70.8)	151	(57.2)
Road	830	(5.4)	35	(12.9)	795	(5.3)	35	(13.3)	38	(14.4)
Workplace	239	(1.6)	27	(9.9)	212	(1.4)	27	(10.2)	20	(7.6)
Others	179	(1.2)	10	(3.7)	168	(1.1)	10	(3.8)	11	(4.2)

OHCA, out-of-hospital cardiac arrest; PAD, public access defibrillation; CPR, cardiopulmonary resuscitation. Percentages for baseline characteristics are calculated based on valid denominators excluding unknown/missing value.

Favourable neurological outcome

Overall, 682 of 15,345 patients (4.4%) achieved a favourable neurological outcome; among those with witnessed OHCA, 550 of 5996 patients (9.2%) achieved a favourable neurological outcome, compared with 132 of 9349 patients with unwitnessed OHCA (1.4%). In the unadjusted analysis, PAD shock was significantly associated with favourable neurological outcome (OR: 17.8, 95% CI: 13.8–23.0). After PSM, PAD shock remained significantly associated with favourable neurological outcome (OR: 3.3, 95% CI: 1.9–5.7; Table 2).

Table 2.

Association between prehospital interventions, time intervals, and 30-day favourable neurological outcomes before and after propensity score matching.

Variable	Unadjusted population (n = 15,345)						Propensity score-matched population (n = 528)
	Total (n)	Unfavourable outcome at 30 days n (%)		Favourable outcome at 30 days n (%)		OR (95% CI)	Total (n)	Unfavourable outcome at 30 days n (%)		Favourable outcome at 30 days n (%)		OR (95% CI)
Public-access defibrillation
Yes	272	160	58.8%	112	41.2%	17.8 (13.8–23.0)	264	154	58.3%	110	41.7%	3.3 (1.9–5.7)
No	15,073	14,503	96.2%	570	3.8%	1.0 (Ref.)	264	214	81.1%	50	18.9%	1.0 (Ref.)
Dispatcher-assisted CPR
Yes	9351	9023	96.5%	328	3.5%	0.58 (0.50–0.68)	368	272	73.9%	96	26.1%	−
No	5994	5640	94.1%	354	5.9%	1.0 (Ref.)	160	96	60.0%	64	40.0%	−
Defibrillation by EMS
Yes	1317	1114	84.6%	203	15.4%	5.2 (4.3–6.1)	122	80	65.6%	42	34.4%	−
No	14,028	13,549	96.6%	479	3.4%	1.0 (Ref.)	406	288	70.9%	118	29.1%	−
Adrenaline administration
Yes	5623	5564	99.0%	59	1.0%	0.15 (0.11–0.22)	210	201	95.7%	9	4.3%	−
No	9722	9099	93.6%	623	6.4%	1.0 (Ref.)	318	167	52.5%	151	47.5%	−
Advanced airway
Yes	7559	7481	99.0%	78	1.0%	0.12 (0.10–0.16)	157	147	93.6%	10	6.4%	−
No	7786	7182	92.2%	604	7.8%	1.0 (Ref.)	371	221	59.6%	150	40.4%	−
Time intervals (continuous, mean ± SD)
Time between EMS call and EMS patient contact		12 ± 4.3		11 ± 3.9		0.94 (0.92–0.97)		12 ± 3.8		11 ± 4.1		−
Time between EMS patient contact and hospital arrival		22 ± 8.6		27 ± 13.2		1.03 (1.03–1.04)		21 ± 7.3		22 ± 8.5		−

Values are presented as n (row %) for categorical variables and mean ± standard deviation for continuous variables. Row percentages represent the proportion of outcomes within each intervention subgroup (e.g., the percentage of patients with a favourable outcome among those who received PAD). The primary outcome was a favourable neurological outcome at 30 days, defined as a Cerebral Performance Category (CPC) score of 1 or 2.

Propensity score matching was performed using a 1:1 ratio to balance baseline characteristics, including age, sex, witnessed status, time of arrest, and dispatcher-assisted CPR. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using logistic regression analysis, with the “No intervention” or “Unfavourable outcome” group as the reference (Ref.) where appropriate.

Abbreviations: CI, confidence interval; CPC, Cerebral Performance Category; CPR, cardiopulmonary resuscitation; EMS, emergency medical services; OR, odds ratio; PAD, public access defibrillation

Thirty-day survival

Overall, 1003 of 15,345 patients (6.5%) achieved 30-day survival, including 776 of 5996 (12.9%) in the witnessed OHCA subgroup. In the unadjusted analysis, PAD shock was significantly associated with 30-day survival (OR: 12.1, 95% CI: 8.7–16.7). After PSM, PAD shock remained significantly associated with 30-day survival (OR: 3.7, 95% CI: 2.2–6.1; Supplementary Table 2).

Prehospital ROSC

Overall, prehospital ROSC occurred in 1438 of 15,345 patients (9.4%), including 1029 of 5996 (17.2%) in the witnessed OHCA subgroup. In the unadjusted analysis, PAD shock was significantly associated with prehospital ROSC (OR: 11.6, 95% CI: 9.1–14.8). After PSM, PAD shock remained significantly associated with prehospital ROSC (OR: 3.1, 95% CI: 1.9–5.0; Supplementary Table 3).

PAD shock by location and structural clustering

Fig. 2 shows the distribution of PAD shock rates and resuscitation performance across arrest locations using a bubble chart framework. Distinct patterns were observed across location types. Specifically, the frequency of PAD shock among all OHCA cases was notable in sports facilities (18/53, 34.0%), railway stations (30/107, 28.0%), medical facilities (25/218, 11.5%), schools (7/27, 25.9%), and government offices (9/26, 34.6%). In contrast, PAD shocks were extremely rare in private residences (6/10,426, 0.06%) and hotels (1/114, 0.9%), despite these locations accounting for a substantial proportion of OHCA events. Nursing homes (66/2628, 2.5%) demonstrated higher PAD shock rates and prehospital ROSC than did private residences; however, favourable neurological outcomes remained poor in both settings. Commercial facilities (25/288, 8.7%), workplaces (21/239, 8.8%), roads (39/830, 4.7%), and public spaces (22/210, 10.5%) formed an intermediate cluster characterised by moderate composite performance despite relatively low PAD shock rates. Furthermore, the bubble chart visually demonstrated that even among cases receiving PAD shocks, the rates of favourable neurological outcomes varied substantially across locations, aligning with their composite bystander performance scores. A supplementary analysis using prehospital ROSC as the outcome is presented in Supplementary Fig. 1.

Fig. 2.

Location-resolved framework integrating PAD shock, bystander performance, and neurological outcomes.

The bubble chart displays the relationship between location-specific bystander performance and patient-centred outcomes across major urban settings. The x-axis represents the rate of 30-day favourable neurological outcome (CPC score, 1–2) among all OHCA patients within each location category and y-axis represents the composite bystander performance score, calculated by standardising and averaging the witnessed arrest rate, bystander CPR rate, and dispatcher-assisted CPR rate for each location category. Bubble size indicates the PAD shock rate within each location. Each bubble corresponds to a predefined location category.

Abbreviations: PAD, public access defibrillation; CPC, Cerebral Performance Category; CPR, cardiopulmonary resuscitation.

Discussion

This population-based study provides a location-resolved evaluation of PAD performance within a large, unified urban EMS system. By integrating individual-level outcome analysis with a location-level performance framework, we identified substantial heterogeneity in PAD shock rates and effectiveness across different urban environments. Furthermore, PAD shock demonstrated consistent associations across multiple outcome measures, remaining significantly associated not only with favourable neurological outcome but also with 30-day survival and prehospital ROSC after PSM, providing an essential benchmark for interpreting location-specific variation while acknowledging the inherent selection related to shockable rhythms.

First, PAD shock was significantly associated with favourable neurological outcome after OHCA, even after PSM. This is consistent with prior registry-based studies demonstrating the importance of early defibrillation in improving neurological survival.1, 7 However, individual-level effect estimates alone do not explain why PAD is effective only in some environments.

Second, the location-resolved analysis revealed marked heterogeneity in PAD shock rates and outcomes across arrest locations. Importantly, our location-level framework (Fig. 2) demonstrates that favourable outcomes across locations are not merely a function of PAD shock rates. Rather, locations achieving high survival rates consistently exhibited high composite bystander performance scores, highlighting that public access defibrillation alone is insufficient; it must be coupled with high-quality, comprehensive bystander interventions prior to defibrillation to optimise outcomes. High-performing public spaces typically share characteristics such as high pedestrian flow, defined on-site responsibility, and predictable AED accessibility, which facilitate this effective bystander response. Since bystanders cannot assess the underlying cardiac rhythm prior to AED application, the success in these environments likely reflects genuine environmental facilitators rather than mere differences in initial rhythm distribution.

Interestingly, within these structured environments, sports facilities and government offices demonstrated higher composite bystander performance scores than medical facilities. Although the reasons cannot be definitively determined, this paradox may reflect the heterogeneity of the 'medical facility' category (e.g., including small clinics), variations in patient comorbidities or end-of-life status, and psychological barriers such as the diffusion of responsibility among bystanders expecting immediate professional intervention.

In stark contrast to these public settings, private residences accounted for the majority (67.9%) of OHCA events yet exhibited extremely low PAD shock rates and poor outcomes, consistent with previous studies showing that conventional PAD programmes have limited reach in home settings.8, 9, 10, 11 The absence of readily accessible AEDs, delayed recognition of cardiac arrest, and reliance on untrained bystanders likely limited early defibrillation in this environment, highlighting private residences as a key target for alternative strategies.

Nursing homes demonstrated a distinct pattern. Although PAD shock and prehospital ROSC were more common than in private residences, favourable neurological outcome was not improved.¹² This discrepancy may reflect the clinical characteristics of residents, including advanced age, multimorbidity, and reduced physiological reserve, which may limit the potential for neurological recovery even when early resuscitation is achieved. Nursing homes may represent settings in which improvement in neurological outcomes is limited despite relatively favourable process metrics, whereas private residences may represent key targets for strategies aimed at enhancing early defibrillation.

Beyond residential settings, substantial heterogeneity was observed within public environments. Commercial facilities, workplaces, roads, and other multipurpose spaces formed an intermediate group characterised by moderate bystander performance but relatively low PAD shock rate. In these environments, increasing AED density alone may not automatically yield proportional improvements in outcomes, as we cannot distinguish whether the primary barrier is the physical absence of devices or a failure to retrieve available AEDs.13, 14, 15 Structural factors, such as dispersed layouts, variable bystander density, and delayed recognition of cardiac arrest, may reduce the likelihood of timely AED use. In such settings, improvement may depend on system-level approaches that extend beyond AED placement alone, including enhanced bystander activation and dispatcher-assisted support.16, 17, 18 In spatially diffuse environments, such as roads, limitations of fixed AED deployment are more pronounced, as distance and visibility can hinder rapid access. Alternative approaches, including dispatcher-guided AED retrieval or mobile AED deployment, may help address these challenges.¹⁹

Taken together, these findings indicate that PAD efficacy is determined not solely by AED availability but also by the structural compatibility between urban environments and the prehospital survival chain. In this regard, Yokohama City provides a particularly suitable context for implementing and evaluating a comprehensive, multi-tiered urban resuscitation strategy. Yokohama’s extensive public facilities and railway network, densely distributed population, and unified municipal fire–EMS–hospital coordination system create favourable preconditions for coordinated system-level interventions.20, 21 Compared with metropolitan regions spanning multiple administrative jurisdictions (e.g., the Greater Tokyo area), Yokohama’s single municipal governance structure may further facilitate streamlined policy implementation, data collection, and iterative quality improvement cycles.20, 21

From a policy translation perspective, these location-resolved patterns provide a foundation for future research to investigate the implementation of targeted urban resuscitation strategies, such as optimising AED distribution or exploring enhanced bystander mobilisation through advanced dispatcher support in low- and intermediate-coverage zones. If such interventions lead to improvements in bystander CPR, PAD shock, or neurologically favourable survival, this location-resolved framework may be transferable to other high-density cities with similar governance structures, both in Japan and internationally.

Limitations

This study has several limitations.

First, as a retrospective observational study, residual confounding cannot be excluded despite use of PSM. This registry records PAD shock delivery rather than AED pad application. Since bystanders cannot assess cardiac rhythm before AED contact, the decision to attempt defibrillation is determined by environmental factors (AED accessibility, bystander density, organisational readiness) rather than the underlying rhythm. However, because AEDs deliver shocks only for shockable rhythms, the PAD shock group inherently excludes patients with non-shockable rhythms who have fundamentally poorer prognoses. The observed association between PAD shock and favourable outcomes therefore reflects both the intervention effect and the inherent survival advantage of shockable rhythms. Cases with missing covariate data were excluded from the PSM analysis; however, these represented only 90 patients (0.6% of all screened cases), and substantial selection bias is unlikely. Second, detailed in-hospital treatment data were unavailable. Therefore, the observed associations between PAD shock and outcomes cannot be fully separated from the effects of post-resuscitation care. Third, arrest location and bystander characteristics were categorised using predefined groupings, and the consolidation of 53 original location categories into 13 analytical groups may have introduced misclassification bias and obscured location-specific heterogeneity. Additionally, granular data on AED accessibility, spatial proximity, and environmental factors such as pedestrian flow or bystander training were not available, limiting the ability to assess structural determinants at a finer scale. Fourth, information on comorbidities, frailty, and pre-arrest functional status was not available. Differences in outcomes across locations may therefore reflect underlying patient characteristics rather than differences in resuscitation performance alone. Finally, this study was conducted in a single large metropolitan city with a unified EMS system, limiting generalisability to other settings with different healthcare systems or urban structures.

Conclusion

This population-based study demonstrated that PAD shock is independently associated with favourable neurological outcomes after OHCA and revealed substantial heterogeneity in its PAD shock rates and effectiveness across urban locations. Factors beyond AED availability alone—including bystander density, occupancy characteristics, and organisational readiness—were found to influence real-world PAD performance, underscoring that survival relies on high-quality, comprehensive bystander interventions.

These findings suggest that optimising the urban chain of survival requires not only increasing AED availability but also aligning deployment, training, and system design with the structural characteristics of specific urban environments. A location-resolved framework may therefore serve as a practical tool for guiding targeted city-level resuscitation strategies.

CRediT authorship contribution statement

Hiroki Usui: Writing – original draft, Visualization, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Hayato Taniguchi: Writing – review & editing, Supervision, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Takeru Abe: Writing – review & editing, Supervision, Methodology, Formal analysis, Conceptualization. Ayako Nogaki: Writing – review & editing, Resources, Project administration, Funding acquisition. Ichiro Takeuchi: Writing – review & editing, Supervision, Resources, Project administration.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The research was supported by institutional funding from the hospital to which the first author is affiliated.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work the authors used ChatGPT and Elicit to identify relevant literature and to assist in structuring, refining, and critically reviewing the manuscript text. After using these tools, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Data statement

The data that support the findings of this study are not publicly available due to restrictions related to privacy and administrative regulations. However, the data are available from the corresponding author upon reasonable request and with appropriate institutional approval.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

The following are the Supplementary material to this article:

Supplementary Fig. 1.

Supplementary Figure 1 presents the location-resolved performance framework using prehospital ROSC as the outcome measure.

Supplementary Table 1

Supplementary Table 1 provides the mapping of original registry arrest location codes to the 13 analytical categories.

Supplementary Table 2

Supplementary Table 2 present the association between PAD shock and 30-day survival and prehospital ROSC, respectively, before and after propensity score matching.

Supplementary Table 3

Supplementary Table 3 present the association between PAD shock and 30-day survival and prehospital ROSC, respectively, before and after propensity score matching.