Enhancing out-of-hospital cardiac arrest survival in China through the 5-minute social rescue circle implementation

Out-of-hospital cardiac arrest (OHCA) survival in China remains critically low due to limited bystander cardiopulmonary resuscitation (CPR) training, insufficient automated external defibrillator (AED) deployment, and delayed community rescue responses. This study quantitatively assessed the impact of the intervention and the 5-minute social rescue circle (5MSRC) on OHCA outcomes in Baoan District, Shenzhen, China.

We employed a patient-centered approach to integrate three datasets from Baoan, Shenzhen from March 2019 to September 2024: (a) 3586 OHCA cases; (b) 332,507 CPR-trained residents; (c) 4,707 AED deployments. A WeChat-based 5MSRC system was developed to integrate video-guided CPR with emergency dispatch. Using the integrated dataset, we applied geospatial tracking, machine learning, and SHapley Additive exPlanations (SHAP) analysis to identify key factors associated with improvedreturn of spontaneous circulation (ROSC) and 30-day survival.

Among 3586 OHCA cases, 3.88 % achieved prehospital ROSC and 3.82 % survived after 30 days. Arrests in public locations showed significantly better outcomes than residential settings (ROSC: 66.2 % vs 33.8 %; 30-day survival: 68.6 % vs 31.4 %, both p < 0.001). During post-pandemic recovery (2020-2024), OHCA outcomes demonstrated significant improvement: bystander CPR rates increased from 12.32 % to 27.27 %, AED application rose from 1.96 % to 4.64 %, and 30-day survival improved from 2.27 % to 6.51 % (all p < 0.05). The machine learning models achieved excellent predictive performance for prehospital ROSC (best ROC-AUC: 0.92, 95 % CI 0.91-0.93) and good performance for 30-day survival (best ROC-AUC: 0.97, 95 % CI 0.96-0.98), demonstrating robust predictive capability for both acute and longer-term outcomes. Feature importance analysis revealed that community-level factors-particularly recent CPR training rates (e.g., 6-month), AED proximity, and rapid response systems-drove prehospital ROSC, while hospital-based interventions (e.g., percutaneous coronary intervention) became increasingly important for 30-day survival.

These findings underscore the critical synergy between community preparedness (frequent CPR training, AED accessibility) and advanced hospital care in optimizing OHCA outcomes. The high-performing prediction models demonstrate the potential of machine learning to identify strategic intervention points across the continuum of cardiac arrest care.