Socioeconomic status influences survival after out-of-hospital cardiac arrest outcomes

2.1 Study design and setting

This retrospective cohort study analyzed OHCA cases in Chiayi City, Taiwan, during 2024. The city covers an area of 60 km² and comprises 84 administrative villages (45 in the east and 39 in the west), with populations of 144,501 and 116,677 residents, respectively. The EMS system, managed by the Chiayi City Fire Bureau, operates seven fire/ambulance stations, 14 ambulances, and 242 public automated external defibrillators (AEDs) located in train stations, schools, and other public facilities. AEDs and EMS resources are distributed uniformly across all 84 villages, ensuring equitable access regardless of income quartile, as confirmed by geospatial mapping, with no significant differences in proximity to EMS stations or AED locations. In this OHCA cohort, the mean EMS response time was 5.23 minutes (standard deviation (SD) 2.3), with variations by income quartile described in the results section. The primary outcome was survival to hospital discharge, and the secondary outcomes were return of spontaneous circulation (ROSC) sustained for more than 2 h and more than 24 h.

2.2 Data sources and linkage

Data were obtained from the Chiayi City EMS registry and comprised patient demographics, incident details, bystander interventions, EMS metrics, and outcomes. Village-level income data (2021, expressed in thousands of NT$) were retrieved from government statistics and linked to OHCA cases by geocoded incident location. Cases were assigned to income quartiles (Q1–Q4) based on case-level distribution: Q1 (≤NT$823,000; 53 cases), Q2 (823,001–868,000; 52 cases), Q3 (868,001–931,000; 52 cases), and Q4 (>NT$931,000; 52 cases). Supplementary Tables 1,2 contains de-identified patient data and corresponding village income levels.

2.3 Study population

Eligible cases included adults (≥18 years) with non-traumatic OHCA occurring in 2024. Exclusion criteria were: (1) traumatic etiology, (2) age <18 years with distinct clinical etiologies and ethical considerations, and (3) incomplete EMS records. ROSC status was not an inclusion or exclusion criterion. Of the 243 OHCA cases recorded, 209 met the eligibility criteria (Fig. 1).

Fig. 1.Study flowchart for selecting OHCA cases. OHCA, out-of-hospital cardiac arrest; EMS, emergency medical service.

2.4 Data collection and variables

Collected variables included age, sex, incident location, witness status, bystander CPR (civilian, EMS, or none), public AED use (binary), number of AED shocks (categorized as 0, 1–2, or ≥3), initial cardiac rhythm (shockable: ventricular fibrillation/ventricular tachycardia; non-shockable: pulseless electrical activity/asystole), and outcomes (ROSC >2 h, ROSC >24 h, and survival to hospital discharge). All data were de-identified in compliance with Taiwan’s Personal Data Protection Act.

2.5 Statistical analysis

The cohort characteristics are summarized using descriptive statistics. Continuous variables were tested for normality using the Shapiro-Wilk test, and normally distributed variables were compared across groups using one-way analysis of variance (ANOVA), and non-normally distributed variables using the Kruskal-Wallis test. When significant differences were detected, post hoc pairwise comparisons were performed using Tukey’s Honestly Significant Difference (HSD) test for normal variables or Bonferroni-corrected Mann-Whitney U tests for non-normal variables. Linearity between continuous variables was evaluated using linear regression with the coefficient of determination (R²).

Multivariable logistic regression models were constructed with variable selection based on clinical relevance and univariate significance (p < 0.2), while ensuring events-per-variable ratios ≥10 to avoid model instability. Three models were specified: (1) survival to hospital discharge (income quartile, shockable rhythm); (2) ROSC >24 h (income quartile, shockable rhythm, age, witnessed status); and (3) ROSC >2 h (income quartile, shockable rhythm, age, witnessed status, bystander CPR, incident location). Incident location was categorized as residence, clinic/nursing home, public settings (roadside, public building, educational facility, transit station, sports center, workplace), or other. The number of AED shocks was treated as a categorical variable (0, 1–2, ≥3). Multicollinearity was assessed using the variance inflation factor (VIF), with VIF >5 indicating concern; all variables had VIF <2. Crude and adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were reported. Categorical variables were compared using Chi-square tests, with Fisher’s exact test or Monte Carlo simulation (10,000 iterations) applied when >20% of cells had expected frequencies <5. Bonferroni correction was applied for post hoc comparisons. Data analyses were conducted using the statsmodels package (v0.14.0) in Python within a Jupyter Notebook environment. Propensity score matching was not performed due to the limited sample size and low event rates, which increased the risk of overfitting.

3.1 Cohort characteristics

Of the 209 included OHCA cases, 60.3% were male, and the mean age was 71.8 years (SD 15.6). Cases were evenly distributed across income quartiles (Q1–Q4: 53, 52, 52, and 52 cases, respectively). Most incidents occurred in residences (72.7%). Bystander CPR was provided in 62.2% of cases (civilian: 46.9%, EMS: 15.3%), and public AED use was recorded in 2.9%. The mean number of AED shocks was 0.43 (SD 1.01). Age did not differ significantly across income quartiles (p = 0.88). EMS response time varied significantly (p = 0.03), with Q2 recording faster times than Q4 (post hocp = 0.02). There was no significant linear relationship between age and EMS response time (p = 0.32).

3.2 Incident location by income quartile

Incident locations varied significantly by income quartile (Table 1), with residential settings being predominant in lower quartiles, accounting for 84.9% of incidents in Q1, decreasing to 51.9% in Q4 (p < 0.01). In contrast, clinic/nursing home incidents were more frequent in Q4 (23.1%) than in Q1 (0.0%, p < 0.01). No significant differences were observed for public settings (p = 0.55) or other locations (p = 0.42). When location was included as a covariate in the multivariable model for ROSC >2 h, no independent association was found (p = 0.12).

3.3 ROSC distribution by income

A total of 68 patients (32.5%) achieved ROSC >2 h, 44 (21.1%) sustained ROSC >24 h, and 18 (8.6%) survived to hospital discharge (Table 1). Outcomes showed a non-linear distribution by income quartile. ROSC >2 h varied significantly (lowest in Q1 at 18.9%, peaking in Q2 at 40.4%; p = 0.02), with post hoc analysis confirming a difference between Q2 and Q1 (p = 0.01). ROSC >24 h demonstrated a borderline trend, ranging from 11.3% in Q1 to 26.9% in Q4 (p = 0.09). Survival to hospital discharge increased from 3.8% in Q1 to 13.5% in Q4 (p = 0.10), with post hoc analysis showing a significant difference between Q4 and Q1 (p = 0.04).

3.4 AED shocks and outcomes

AED shock categories (0, 1–2, ≥3) are presented in Table 2. Survival to hospital discharge was highest in patients receiving ≥3 shocks (28.6%), followed by those receiving 1–2 shocks (19.0%) and those with no shocks (5.0%) (adjusted odds ratio (aOR) 3.75, 95% CI 1.21–11.66, p = 0.02). However, higher shock counts (≥3) were associated with lower adjusted odds of survival in some cases, likely reflecting refractory arrhythmias (aOR 2.42, 95% CI 0.31–18.84, p = 0.40), as multiple shocks often indicate prolonged resuscitation rather than improved prognosis. In addition, there was no evidence of multicollinearity between public AED use and the number of shocks (VIF = 1.02).

There was no multicollinearity between the public AED usage and shock number (VIF = 1.02).

In summary, the higher income quartiles were associated with increased rates of ROSC >2 h and survival to hospital discharge; however, these associations were attenuated and no longer statistically significant after adjustment for confounders. No significant differences across income quartiles were observed for age, bystander CPR rates, or public AED use. EMS response time, however, differed significantly by income quartile.

3.5 Crude and adjusted associations

Crude and adjusted odds ratios for key outcomes by income quartile are shown in Table 3. For survival to hospital discharge, higher income quartiles had greater odds compared to Q1, but the associations did not reach statistical significance after adjustment (Q4 vs. Q1: aOR 2.06, 95% CI 0.76–5.64, p = 0.10). Shockable rhythm remained a strong independent predictor (aOR 17.4, 95% CI 4.2–71.7, p < 0.01). Similar non-significant patterns were observed for ROSC >24 h and >2 h after adjustment. Model performance was satisfactory, with AUC values ranging from 0.78 to 0.80, and no evidence of poor calibration (Hosmer-Lemeshow p > 0.05).

4.1 Non-linear SES impact and Q2 peak

Moderate-income neighborhoods showed higher early ROSC rates, potentially attributable to both equitable access to resources and community characteristics that facilitate rapid intervention. This pattern suggests that mid-tier SES communities may exhibit stronger social cohesion, which could enhance the likelihood of timely bystander CPR, consistent with prior observations in similar settings [11, 13]. The faster EMS response times observed in Q2 may have contributed to these results [12], although other factors such as the prevalence of initial shockable rhythms could also play a role. These findings indicate the importance of considering both clinical and community-level variables in urban EMS planning.

4.2 Shockable rhythm as a dominant predictor

An initial shockable rhythm was strongly associated with survival, aligning with extensive evidence supporting the effectiveness of rapid defibrillation in improving OHCA outcomes [14, 15]. Nevertheless, cases requiring multiple shocks may represent refractory arrhythmias or underlying cardiac pathology, which can limit prognosis despite prompt intervention [14]. The separate contributions of public AED use, and the number of shocks highlight opportunities for targeted community training to enhance early response in diverse SES contexts.

4.3 Incident location and SES

Incident location varied by SES, with wealthier areas having more arrests in institutional settings such as clinics and nursing homes. These environments may require specific response protocols and staff training to optimize early defibrillation and resuscitation efforts [16]. In contrast, lower SES areas showed a greater proportion of arrests in private residences, possibly reflecting reduced access to healthcare facilities. This contrasts with findings from larger cities, where high-SES areas more frequently experience arrests in public settings [17], and suggests that compact city design and uniform EMS coverage may reduce location-related disparities.

4.4 EMS response time and SES

Variation in EMS response times across quartiles may be attributable to subtle geographic or operational differences, such as accessibility in certain neighborhoods. Unlike patterns observed in larger cities, where low-SES areas often experience longer delays due to resource inequities [18], this small urban setting benefits from more consistent EMS coverage. Nonetheless, these differences emphasize the importance of optimizing station placement and community readiness to ensure equitable response times across all SES levels.

4.5 Implications and public health relevance

The patterns observed suggest that moderate-income neighborhoods could serve as models for enhancing bystander readiness through community-based strategies, given their relatively high bystander CPR rates. In high-income areas, the higher proportion of arrests occurring in clinics or nursing homes highlights the need for targeted staff training in early defibrillation. Interventions to improve AED availability and integration in lower SES areas may help reduce disparities. These findings support the development of localized strategies tailored to the social and spatial characteristics of small urban communities [19].

4.6 Comparison with broader literature

The non-linear SES associations observed in this study differ from the linear trends often reported in large metropolitan areas, where higher SES consistently correlates with better outcomes [20]. The importance of shockable rhythms aligns with international findings [15], yet the relatively low rate of public AED use suggests potential cultural, educational, or accessibility barriers compared to countries such as Japan, where public AED engagement is higher [21]. Further investigation incorporating individual-level SES data and detailed geospatial analyses could refine these urban-specific insights.

4.7 Strengths and limitations

This study’s strengths include balanced income quartile definitions, use of multiple outcome measures, and conservative statistical modeling to avoid overfitting. Limitations include the modest sample size, the potential influence of unmeasured confounders such as comorbidities, and reliance on village-level income data from 2021. Propensity score matching was not used to preserve statistical power, but multivariable regression accounted for key covariates. Although the incident location was recorded, housing type (e.g., apartment versus single-family home) was not specified, limiting the ability to assess housing-related influences on OHCA outcomes.