Abstract
Background
Out-of-hospital cardiac arrest (OHCA) remains a major public health challenge. Smartphone-based systems may facilitate early bystander response; however, their real-world associations with patient outcomes following implementation within emergency medical service (EMS) systems remain uncertain.
Objective
This study aimed to evaluate the association between system-level implementation of the MySOS application and prehospital and clinical outcomes among patients with OHCA in Bangkok.
Methods
We conducted a retrospective uncontrolled before–after observational study of patients with OHCA in Bangkok, Thailand, between December 2022 and May 2024. Patients were classified according to the pre- and post-implementation periods of the MySOS application. Inverse probability of treatment weighting (IPTW) and weighted multivariable regression were used to account for measured differences between the study periods. Given the temporal definition of exposure and absence of confirmed patient-level application use, findings were interpreted as associations rather than causal effects.
Results
Among 615 patients, 340 (55.3%) were included in the pre-implementation period and 275 (44.7%) in the post-implementation period. The post-implementation period was associated with a higher probability of return of spontaneous circulation (ROSC) (adjusted RD 6.28%, 95% CI 1.69–10.86; p = 0.007), although the corresponding RR was not statistically significant (RR 1.21, 95% CI 0.78–1.89). Differences in death at the scene were also observed between study periods (adjusted RD 14.70%, 95% CI 9.14–20.27; p < 0.001). No statistically significant differences were observed in 24-h survival, survival to hospital admission, or survival to hospital discharge.
Conclusions
The post-implementation period of the MySOS system was associated with a higher probability of ROSC, but not with improved longer-term survival outcomes. These findings should be interpreted cautiously because of the uncontrolled before–after design, residual baseline imbalance, potential temporal confounding, and the absence of patient-level data confirming actual application use. Further prospective studies incorporating direct measures of MySOS activation and utilization are warranted.
Keywords: Out-of-hospital cardiac arrest, Smartphone application, Bystander CPR, Emergency medical services, Return of spontaneous circulation, Thailand
Introduction
Out-of-hospital cardiac arrest (OHCA) remains a major global public health challenge, with survival critically dependent on early recognition, prompt bystander cardiopulmonary resuscitation (CPR), and rapid defibrillation.1, 2, 3, 4 Despite ongoing public education and training efforts,5 bystander CPR rates remain suboptimal in many regions, contributing to persistently poor survival outcomes.1, 2, 3, 4
Recent advances in digital health technologies, particularly smartphone-based applications, have introduced new opportunities to strengthen prehospital response. These applications can alert nearby responders, identify automated external defibrillator (AED) locations, and provide real-time CPR guidance to bystanders. Evidence from several countries suggests that such systems may improve early resuscitation processes. For example, the myResponder system in Singapore has been associated with increased bystander CPR and AED use, with potential benefits for early resuscitation outcomes.6 Similarly, smartphone-based responder systems such as GoodSAM have been implemented to facilitate earlier bystander intervention and access to defibrillation.7, 8
A recent meta-analysis further demonstrated that mobile application–based systems were associated with increased bystander CPR, AED utilization, and ROSC.9 However, evidence regarding longer-term survival outcomes remains inconsistent. In addition, simulation-based studies suggest that although smartphone-assisted systems may facilitate selected components of the emergency response, interaction with the application may also affect the timing of CPR initiation in some scenarios.10 These findings underscore the importance of evaluating smartphone-based systems under real-world EMS conditions.
In Thailand, the MySOS smartphone application was developed to support the early response to OHCA and incorporates emergency call support, global positioning system (GPS)-based identification of nearby AEDs, video-assisted CPR guidance, and communication support with emergency medical dispatchers.11, 12 The system was designed to complement, rather than replace, conventional emergency dispatch and professional EMS response.13 Although MySOS has demonstrated feasibility in simulation-based settings, evidence regarding its implementation within an operational EMS system, particularly in a large urban setting such as Bangkok, remains limited.
Importantly, many previous studies of smartphone-based responder systems have evaluated individual-level application activation or use. In contrast, evaluating system-level implementation in routine EMS practice presents additional methodological challenges because application use may not be confirmed for every OHCA event and outcomes may also be influenced by concurrent changes in patient characteristics, resuscitation practices, and EMS care over time. Consequently, evidence regarding the association between system-level implementation of smartphone-based CPR assistance and clinical outcomes remains limited.
Therefore, this study aimed to evaluate the association between the system-level implementation of the MySOS application and prehospital and clinical outcomes among patients with OHCA in Bangkok.
Methods
Study design and setting
This retrospective uncontrolled before–after observational study evaluated the association between the system-level implementation of a smartphone-based CPR assistance system (MySOS) and prehospital and clinical outcomes among patients with OHCA in Bangkok, Thailand.
The study compared outcomes between a pre-implementation cohort (historical control) and a post-implementation cohort defined according to the operational availability of the MySOS system. The pre-implementation cohort included eligible OHCA cases occurring before May 1, 2023. Following the historical control period, training and pilot implementation of the MySOS application were conducted among community volunteers and Bangkok emergency dispatch personnel between May and November 2023. The post-implementation cohort included eligible OHCA cases occurring on or after December 1, 2023, when the MySOS system became available for operational use within the EMS network. Cases occurring during the training and pilot implementation period (May–November 2023) were not included in either analytic cohort.
The study was conducted within the Bangkok Metropolitan Emergency Medical Services (EMS) system and coordinated through the Bangkok Dispatch Center (Erawan Center), which oversees emergency call triage, dispatch operations, and centralized EMS data collection. The Bangkok EMS system uses a tiered prehospital response involving basic life support (BLS) and advanced life support (ALS) units. BLS units include emergency medical personnel such as emergency medical technicians (EMTs) and provide basic resuscitative interventions, including CPR, basic airway management, oxygen administration, and automated external defibrillator (AED) use. ALS units include paramedics and other qualified advanced-level providers who can perform advanced resuscitative interventions, including advanced airway management, cardiac rhythm assessment, medication administration, and manual defibrillation. In OHCA care, AEDs may be used by BLS personnel and trained lay responders, whereas manual defibrillators are primarily used by ALS providers trained in cardiac rhythm interpretation and advanced cardiac life support.
MySOS was developed as a smartphone-based system to support early recognition and response to OHCA and to complement the existing Bangkok EMS response pathway. The application incorporates emergency call support, global positioning system (GPS)-based identification of nearby AEDs, video-assisted CPR guidance, and communication support with emergency medical dispatchers. During the implementation period, community volunteers and dispatch personnel received training in the use of the application. MySOS was intended to facilitate early EMS activation, support bystander CPR, and assist users in locating nearby AEDs while conventional BLS and ALS resources continued to be dispatched through the existing EMS system. Thus, the application functioned as an additional digital support component within the established EMS response pathway rather than replacing conventional emergency dispatch or professional EMS response.
Exposure was defined according to the temporal availability of the MySOS system rather than confirmed individual-level application activation or use. Therefore, the post-implementation cohort represented system-level exposure to an EMS environment in which MySOS was operationally available, rather than direct patient-level exposure to the application. Accordingly, the findings were interpreted as associations between the MySOS implementation period and observed outcomes rather than causal effects of actual application use.
Participants and eligibility criteria
Patients were eligible for inclusion if they: (1) experienced non-traumatic OHCA within the Bangkok Metropolitan Area; (2) were aged ≥15 years; (3) were recorded in the Bangkok EMS registry; and (4) experienced OHCA during either the predefined pre-implementation or post-implementation study period. Patients with trauma-related OHCA or those aged <15 years were considered ineligible. Cases with missing data required for the primary analysis were subsequently excluded from the analytic cohort.
Data collection
Data were obtained from the Bangkok EMS dispatch registry maintained by the Erawan Center and supplemented, where available, with prehospital care records and electronic medical records from receiving hospitals, including emergency department documentation. Where available, data were cross-referenced with national EMS databases to improve data completeness and consistency. Data were extracted using a structured data collection form based on predefined study variables and were de-identified prior to analysis.
Collected variables included patient characteristics (age, sex, and comorbidities); event-related characteristics (location of arrest, witnessed status, presumed etiology, and initial cardiac rhythm); bystander-related variables (bystander type, bystander CPR, and AED-related characteristics); EMS process variables (response time, scene time, and transport time); and prehospital resuscitation interventions, including airway management, intravenous fluid administration, adrenaline administration, and defibrillation.
Because exposure was defined according to the temporal availability of the MySOS system, individual-level data confirming application activation or use were not consistently available. Accordingly, available MySOS-related variables were treated as supplementary descriptive information rather than as definitive measures of individual-level intervention exposure.
Clinical outcomes included ROSC at the scene, sustained ROSC, death at the scene, 24-h survival, survival to hospital admission, and survival to hospital discharge. Definitions of cardiac arrest variables and outcomes were based on the 2024 update of the Utstein Out-of-Hospital Cardiac Arrest Registry Template.14
Definitions
OHCA and related cardiac arrest variables were defined in accordance with the 2024 update of the Utstein Out-of-Hospital Cardiac Arrest Registry Template.14 Initial cardiac rhythms were categorized as shockable or non-shockable. Shockable rhythms included ventricular fibrillation (VF) and pulseless ventricular tachycardia (pVT), whereas non-shockable rhythms included asystole and pulseless electrical activity (PEA). Return of spontaneous circulation (ROSC) was defined as the restoration of spontaneous circulation without the need for ongoing chest compressions. Sustained ROSC was defined as the maintenance of spontaneous circulation for at least 20 min. EMS response time was defined as the interval from EMS dispatch to arrival at the scene.
EMS response time was defined as the interval from EMS dispatch to arrival at the scene. Scene time was defined as the interval from EMS arrival at the scene to departure from the scene, and transport time as the interval from scene departure to arrival at the receiving hospital. Time to first AED shock was defined as the interval from cardiac arrest to delivery of the first AED shock.
Outcome measures
The primary outcomes were ROSC at the scene and 24-h survival following OHCA. Secondary outcomes were survival to hospital admission and survival to hospital discharge. Death at the scene was additionally evaluated as an exploratory outcome. Outcome definitions were based on the 2024 update of the Utstein Out-of-Hospital Cardiac Arrest Registry Template.14 Neurological outcomes at hospital discharge, including Cerebral Performance Category (CPC) scores, were not available in the dataset. Therefore, survival outcomes could not be accompanied by an assessment of functional neurological status.
Statistical analysis
Baseline characteristics were summarized using descriptive statistics. Continuous variables were reported as mean ± standard deviation (SD) or median with interquartile range (IQR), as appropriate, and categorical variables were presented as frequencies and percentages. Covariate balance between the pre- and post-implementation cohorts was assessed using standardized differences (STDs), rather than significance testing, with an absolute STD >0.10 considered indicative of meaningful imbalance.
To improve comparability between the pre- and post-implementation cohorts with respect to measured prognostic characteristics, inverse probability of treatment weighting (IPTW) based on propensity scores was applied. Propensity scores representing the probability of belonging to the post-implementation period, conditional on measured covariates, were estimated using multivariable logistic regression. The propensity-score model used a prognostic-balancing approach and included measured patient, event, EMS, and resuscitation-related characteristics considered clinically relevant to OHCA outcomes, including age, sex, comorbidities, arrest location, prehospital time intervals, bystander-related characteristics, initial cardiac rhythm, defibrillation, level of EMS care, resuscitation medications, and AED-related characteristics. The inclusion of resuscitation interventions was intended to improve balance in measured resuscitation-related case mix between the two observational cohorts rather than to characterize these variables as conventional pre-exposure confounders. Covariate balance before and after weighting was evaluated using absolute STDs and graphically assessed using a balance plot.
Variables with residual imbalance after weighting (absolute STD >0.10) were additionally included as covariates in the weighted outcome models to provide double adjustment. For binary outcomes, weighted generalized linear models with a binomial distribution and identity link were used to estimate absolute risk differences (RDs) with 95% confidence intervals (CIs). Weighted modified Poisson regression models with robust standard errors were used to estimate risk ratios (RRs) with 95% CIs. Outcomes evaluated in these models included ROSC at the scene, death at the scene, 24-h survival, survival to hospital admission, and survival to hospital discharge. Estimates from the weighted models were interpreted as associations between the MySOS implementation period and outcomes rather than causal effects of individual-level application use.
Missing data were assessed before analysis. Cases with missing data required for the primary analysis were excluded, and the analyses were therefore conducted using a complete-case approach without imputation. Prespecified sensitivity analyses included analyses stratified by initial cardiac rhythm (shockable vs non-shockable) and analyses restricted to cases with documented AED availability. All statistical analyses were performed using Stata version 18.0 (StataCorp LLC, College Station, TX, USA). All tests were two-sided, and a p-value < 0.05 was considered statistically significant.
Results
Baseline characteristics
A total of 695 patients were assessed for eligibility, of whom 80 were excluded, leaving 615 patients in the final analysis (340 pre-implementation and 275 post-implementation; Fig. 1). Important baseline differences were observed between the study periods, particularly in initial cardiac rhythm. IPTW improved covariate balance, although residual imbalance remained for selected prognostic variables (Table 1 and Fig. 2).
Fig. 1.
Study flow diagram of the pre–post implementation study.
Table 1.
Baseline demographic and cardiac arrest characteristics before and after implementation of the My SOS application.
| Characteristics |
All patients (n = 615) |
Pre-implementation period (n = 340) |
Post-implementation period (n = 275) |
Unweighted STD | Weighted STD | |||
|---|---|---|---|---|---|---|---|---|
| Overall | 615 | (100.0) | 340 | (55.3) | 275 | (44.7) | ||
| Gender (%) | ||||||||
| Male | 387 | (62.9) | 207 | (60.9) | 180 | (65.5) | 0.094 | 0.091 |
| Female | 228 | (37.1) | 133 | (39.1) | 95 | (34.5) | ||
| Age (years) | 63.97 ± 17.54 | 64.47 ± 17.57 | 63.36 ± 17.50 | 0.062 | 0.072 | |||
| Past medical history | 309 | (50.2) | 174 | (51.2) | 135 | (49.1) | 0.042 | 0.142 |
| Hypertension | 157 | (25.5) | 91 | (26.8) | 66 | (24.0) | 0.064 | 0.050 |
| Diabetes mellitus | 108 | (17.6) | 62 | (18.2) | 46 | (16.7) | 0.040 | 0.056 |
| Dyslipidemia | 20 | (3.3) | 12 | (3.5) | 8 | (2.9) | 0.035 | 0.025 |
| Heart disease | 83 | (13.5) | 41 | (12.1) | 42 | (15.3) | 0.094 | 0.036 |
| Old CVA | 16 | (2.6) | 11 | (3.2) | 5 | (1.8) | 0.090 | 0.028 |
| Renal disease/CKD | 33 | (5.4) | 18 | (5.3) | 15 | (5.5) | 0.007 | 0.132 |
| Cancer | 27 | (4.4) | 16 | (4.7) | 11 | (4.0) | 0.035 | 0.068 |
| Respiratory diseases/COPD | 10 | (1.6) | 3 | (0.9) | 7 | (2.5) | 0.128 | 0.040 |
| Others | 38 | (6.2) | 22 | (6.5) | 16 | (5.8) | 0.027 | 0.068 |
|
Cardiac arrest characteristics Location type (n, %) | ||||||||
| • Home residence | 482 | (78.4) | 266 | (78.2) | 216 | (78.5) | 0.008 | 0.046 |
| • Public/commercial building | 133 | (21.6) | 74 | (21.8) | 59 | (21.5) | ||
| Bystander CPR performed | ||||||||
| Relative | 444 | (72.2) | 245 | (72.1) | 199 | (72.4) | 0.026 | 0.043 |
| Bystanders | 99 | (16.1) | 56 | (16.5) | 43 | (15.6) | ||
| Healthcare personnel | 72 | (11.7) | 39 | (11.5) | 33 | (12.0) | ||
| First arrest rhythm (n, %) | ||||||||
| pVT | 8 | (1.3) | 4 | (1.2) | 4 | (1.5) | 0.427 | − |
| PEA | 101 | (16.4) | 66 | (19.4) | 35 | (12.7) | ||
| Asystole | 443 | (72.0) | 254 | (74.7) | 189 | (68.7) | ||
| VF | 63 | (10.2) | 16 | (4.7) | 47 | (17.1) | ||
| Initial rhythm category | ||||||||
| Shockable rhythms | 71 | (11.5) | 20 | (5.9) | 51 | (18.5) | 0.394 | 0.148 |
| Non-shockable rhythms | 544 | (88.5) | 320 | (94.1) | 224 | (81.5) | ||
Abbreviations: IQR, interquartile range; SD, standard deviation; STD, standardized difference.
Data are presented as number (%), mean ± standard deviation, or median (interquartile range).
An absolute STD >0.10 was considered indicative of meaningful covariate imbalance.
Fig. 2.
Propensity score distribution and covariate balance before and after inverse probability of treatment weighting (IPTW), (a) Propensity score distributions before and after inverse probability of treatment weighting (IPTW) and (b) Covariate balance before and after IPTW based on standardized differences.
Prehospital characteristics and resuscitation interventions
Prehospital response intervals and resuscitation characteristics differed between the study periods, particularly for selected resuscitation interventions and AED-related characteristics (Table 2).
Table 2.
Prehospital response, resuscitation interventions, and AED-related characteristics before and after implementation of the My SOS application.
| Characteristics |
All patients (n = 615) |
Pre-implementation period (n = 340) |
Post-implementation period (n = 275) |
p-value | Unweighted STD | Weighted STD | |||
|---|---|---|---|---|---|---|---|---|---|
| Overall | 615 | (100.0) | 340 | (55.3) | 275 | (44.7) | |||
| Prehospital time intervals (min) | |||||||||
| Cardiac arrest to detection | 10.09 ± 16.96 | 11.05 ± 15.19 | 8.90 ± 18.88 | 0.135 | 0.120 | 0.046 | |||
| Cardiac arrest to dispatch | 13.53 ± 17.09 | 14.55 ± 15.39 | 12.28 ± 18.93 | 0.101 | 0.132 | 0.058 | |||
| Cardiac arrest to scene | 25.75 ± 19.32 | 26.69 ± 16.44 | 24.59 ± 22.35 | 0.194 | 0.107 | 0.025 | |||
| Cardiac arrest to EMS departure | 57.53 ± 24.88 | 59.03 ± 24.66 | 55.69 ± 25.06 | 0.098 | 0.134 | 0.023 | |||
| Cardiac arrest to hospital | 68.80 ± 29.08 | 70.02 ± 29.75 | 67.30 ± 28.21 | 0.250 | 0.094 | 0.003 | |||
| Cardiac arrest to CPR | 10.55 ± 17.70 | 11.53 ± 15.48 | 9.34 ± 20.06 | 0.126 | 0.122 | 0.059 | |||
| Response time | 15.37 ± 8.20 | 15.35 ± 6.46 | 15.39 ± 9.96 | 0.957 | 0.004 | 0.038 | |||
| Scene time | 31.30 ± 15.89 | 31.85 ± 16.85 | 30.61 ± 14.63 | 0.337 | 0.079 | 0.006 | |||
| Transfer time | 10.63 ± 11.4 | 10.17 ± 11.27 | 11.19 ± 11.56 | 0.271 | 0.089 | 0.033 | |||
| Resuscitation interventions | |||||||||
| Defibrillation | 77 | (12.5) | 28 | (8.2) | 49 | (17.8) | <0.001 | 0.288 | 0.162 |
| Number of defibrillation (times) | 2 | (1–3.5) | 2.5 | (1–3) | 2 | (1–4) | 0.512 | 0.233 | 0.184 |
| Advanced life support | 612 | (99.5) | 338 | (99.4) | 274 | (99.6) | 1.000 | 0.033 | 0.019 |
| Endotracheal intubation | 519 | (84.4) | 306 | (90.0) | 213 | (77.5) | <0.001 | 0.345 | 0.008 |
| Chest compression | 611 | (99.3) | 339 | (99.7) | 272 | (98.9) | 0.330 | 0.096 | 0.020 |
| Intravenous fluid administration | 585 | (95.1) | 338 | (99.4) | 247 | (89.8) | <0.001 | 0.435 | 0.084 |
| Adrenaline | 570 | (92.7) | 335 | (98.5) | 235 | (85.5) | <0.001 | 0.496 | 0.027 |
| Number of adrenaline (amp) | 5 | (3–7) | 5 | (3–7) | 5 | (3–7) | 0.661 | 0.055 | 0.081 |
| AED-related characteristics | |||||||||
| My SOS-assisted AED localization, n (%) | 236 | (38.4) | 148 | (43.5) | 88 | (32.0) | <0.001 | 1.040 | 0.022 |
| Not available | 229 | (37.2) | 171 | (50.3) | 58 | (21.1) | |||
| Available | 150 | (24.4) | 21 | (6.2) | 129 | (46.9) | |||
| Time to first AED shock (min), (n = 172) | 16.48 ± 14.71 | 3.81 ± 4.76 | 18.24 ± 14.77 | <0.001 | |||||
Weighted univariable analyses
In the IPTW-weighted analyses, the post-implementation period was associated with differences in ROSC, death at the scene, and survival to hospital admission based on absolute risk differences, whereas no significant associations were observed for 24-h survival or survival to hospital discharge. Corresponding risk-ratio estimates were significant only for death at the scene (Table 3).
Table 3.
Propensity score–weighted univariable analysis of clinical outcomes before and after MySOS implementation.
| Outcomes |
Pre-implementation period (n = 340) |
Post-implementation period (n = 275) |
Univariable analysis |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|
| n | (%) | n | (%) | Risk difference†(95% CI) | p-value | Risk ratio‡(95% CI) | p-value | |||
| Return of spontaneous circulation | 62 | (18.2) | 71 | (25.8) | 5.68 | (0.94, 10.43) | 0.019 | 1.27 | (0.82, 1.96) | 0.288 |
| Death at scene | 109 | (32.1) | 77 | (28.0) | 14.34 | (9.22, 19.46) | <0.001 | 1.58 | (1.14, 2.17) | 0.006 |
| Survival in hospital 24 h | 62 | (18.2) | 61 | (22.2) | −3.39 | (−7.94, 1.15) | 0.143 | 0.85 | (0.53, 1.36) | 0.500 |
| Survival in hospital admission | 46 | (13.5) | 22 | (8.0) | −4.78 | (−8.65, −0.91) | 0.015 | 0.71 | (0.36, 1.40) | 0.320 |
| Survival to hospital discharged | 8 | (2.4) | 6 | (2.2) | −1.53 | (−3.40, 0.35) | 0.111 | 0.58 | (0.17, 2.00) | 0.389 |
Abbreviations: CI, confidence interval; RD, risk difference; RR, risk ratio.
Risk difference estimated by generalized linear model with identity link in binomial family, with inverse probability of treatment weighting (IPTW) based on propensity scores (PS).
Risk ratio estimated by a modified Poisson regression with robust standard errors, with inverse probability of treatment weighting (IPTW) based on propensity scores (PS).
Weighted multivariable analyses with double-adjustment prognostic variables
After IPTW, variables with residual imbalance were additionally included in the weighted outcome models. In the double-adjusted analyses, the post-implementation period was associated with ROSC based on the adjusted risk difference, although the corresponding risk ratio was not statistically significant. An association with death at the scene was also observed, whereas no significant associations were found for 24-h survival, survival to hospital admission, or survival to hospital discharge (Table 4).
Table 4.
Propensity score–weighted and double-adjusted analyses of clinical outcomes before and after MySOS implementation.
| Outcomes |
Univariable analysis |
Multivariable analysis§ |
||||||
|---|---|---|---|---|---|---|---|---|
| Risk difference†(95% CI) | p-value | Risk ratio‡(95% CI) | p-value | Risk difference (95% CI) | p-value | Risk ratio (95% CI) | p-value | |
| Return of spontaneous circulation | 5.68 | 0.019 | 1.27 | 0.288 | 6.28 | 0.007 | 1.21 | 0.400 |
| (0.94, 10.43) | (0.82, 1.96) | (1.69, 10.86) | (0.78, 1.89) | |||||
| Death at scene | 14.34 | <0.001 | 1.58 | 0.006 | 14.70 | <0.001 | 1.46 | 0.014 |
| (9.22, 19.46) | (1.14, 2.17) | (9.14, 20.27) | (1.08, 1.97) | |||||
| Survival in hospital 24 h | −3.39 | 0.143 | 0.85 | 0.500 | 0.35 | 0.877 | 0.83 | 0.434 |
| (−7.94, 1.15) | (0.53, 1.36) | (−4.12, 4.83) | (0.51, 1.33) | |||||
| Survival in hospital admission | −4.78 | 0.015 | 0.71 | 0.320 | −2.02 | 0.316 | 0.71 | 0.311 |
| (−8.65, −0.91) | (0.36, 1.40) | (−5.96, 1.93) | (0.37, 1.38) | |||||
| Survival to hospital discharged | −1.53 | 0.111 | 0.58 | 0.389 | 0.34 | 0.781 | 0.64 | 0.453 |
| (−3.40, 0.35) | (0.17, 2.00) | (−2.06, 2.75) | (0.20, 2.07) | |||||
Abbreviations: CI, confidence interval; RD, risk difference; RR, risk ratio.
Risk difference estimated by generalized linear model with identity link in binomial family, with inverse probability of treatment weighting (IPTW) based on propensity scores (PS).
Risk ratio estimated by a modified Poisson regression with robust standard errors, with inverse probability of treatment weighting (IPTW) based on propensity scores (PS).
Multivariable analysis using double adjustment in the weighted samples with comorbidity, first shockable, medication use during CPR process (defibrillation).
Discussion
Principal findings
In this study, the post-implementation period of a smartphone-based CPR assistance system was associated with a higher probability of ROSC compared with the pre-implementation period. This association was observed for the adjusted absolute risk difference, although the corresponding risk ratio was not statistically significant. No statistically significant associations were observed for 24-h survival, survival to hospital admission, or survival to hospital discharge. An association with death at the scene was also observed in the weighted analyses. Overall, these findings suggest that differences in early resuscitation outcomes between the pre- and post-implementation periods did not translate into improved longer-term survival outcomes.1, 2, 3 However, the findings should be interpreted cautiously because of the uncontrolled before–after observational design, substantial baseline differences in important prognostic factors, residual imbalance after weighting, potential temporal confounding, and the absence of patient-level data confirming actual MySOS use. Accordingly, the observed associations cannot be attributed independently or causally to MySOS implementation.
Comparison with previous studies
The findings of this study are broadly consistent with previous research suggesting that smartphone-based systems may facilitate early bystander response and are associated with favorable early resuscitation outcomes. Systems such as myResponder and GoodSAM have been associated with increased bystander CPR and AED utilization, with potential benefits for ROSC and other early resuscitation outcomes.6, 7, 8 Similarly, a recent meta-analysis reported that mobile application–based interventions were associated with increased bystander CPR, AED use, and ROSC.9 However, evidence regarding longer-term outcomes, including survival to hospital discharge, remains inconsistent across studies.4, 9 This variability likely reflects the multifactorial nature of OHCA survival, which depends not only on early recognition, CPR, and defibrillation but also on subsequent advanced life support, post-resuscitation care, and neurological recovery.3, 4 The present findings should also be considered in the context of previous OHCA data from Thailand. In the Pan-Asian Resuscitation Outcomes Study (PAROS), 65.1% of OHCA cases in Thailand were witnessed by a bystander, whereas bystander CPR and bystander defibrillation were reported in only 15.8% and 0.3% of cases, respectively. Survival to hospital admission was 27.7%, and survival to hospital discharge was 4.2%.15, 16, 17 These historical data illustrate gaps in early community resuscitation, particularly in bystander CPR and defibrillation, and provide important context for the development and implementation of digital systems intended to facilitate early bystander intervention and AED localization in Thailand.18 However, these historical estimates should not be directly compared with those of the present study because of differences in study periods, populations, EMS systems, and outcome definitions.
Importantly, many previous studies evaluated settings in which application activation or use could be measured at the individual level, whereas the present study evaluated outcomes according to system-level availability of MySOS during the post-implementation period rather than confirmed individual-level application use. This methodological distinction limits direct comparison with studies evaluating confirmed application activation and is important when interpreting the observed associations. Specifically, differences between the pre- and post-implementation periods cannot be attributed solely to MySOS and may also reflect differences in patient characteristics, resuscitation practices, and other temporal changes in EMS care.
Interpretation of findings
Several factors may have contributed to the observed findings. First, the higher ROSC observed in the post-implementation cohort should be interpreted cautiously because this cohort had a substantially higher proportion of shockable rhythms than the pre-implementation cohort (18.5% vs 5.9%) and differed markedly in AED-related characteristics. Both initial shockable rhythm and access to early defibrillation are important prognostic factors for successful resuscitation.19 These imbalances may therefore have contributed substantially to the observed difference in ROSC and limit the extent to which this association can be attributed independently to MySOS implementation. Furthermore, residual imbalance in initial rhythm and defibrillation-related characteristics remained after propensity score weighting, despite additional adjustment in the weighted outcome models.
Second, the association between the post-implementation period and death at the scene warrants cautious interpretation, particularly because the direction and magnitude of the model-based estimates differed from the unadjusted proportions. Potential explanations may include differences in case mix, illness severity, EMS operational practices, or termination-of-resuscitation decisions between the study periods. However, these factors were not fully captured in the available dataset and therefore cannot be directly evaluated. Given that exposure was defined by the system-level availability of MySOS rather than confirmed individual-level application use, the observed association may also reflect unmeasured temporal or system-level changes rather than a direct effect of the application.
Finally, the absence of statistically significant improvements in 24-h survival, survival to hospital admission, or survival to hospital discharge, despite the observed association with ROSC based on the adjusted absolute risk difference, highlights the complexity of translating early resuscitation outcomes into longer-term survival. Survival after OHCA depends on the entire chain of care, including early recognition, CPR and defibrillation, advanced life support, post-resuscitation care, and subsequent in-hospital management.3, 4, 20, 21, 22 These findings therefore reinforce the importance of evaluating smartphone-based resuscitation systems within the broader continuum of OHCA care. Given the uncontrolled before–after observational design, residual baseline imbalance, potential temporal confounding, and lack of confirmed patient-level MySOS utilization, the observed associations should not be interpreted as causal effects of the application.
Implications for clinical practice and EMS systems
These findings have important implications for the implementation and evaluation of smartphone-based systems within EMS. Digital tools designed to support early recognition, bystander CPR, and AED localization should be integrated within the broader chain of survival,14, 23 including emergency dispatch, timely professional EMS response, early defibrillation, and post-resuscitation care.2, 3 The absence of corresponding improvements in longer-term survival in this study highlights that implementation of a smartphone-based system alone may be insufficient to improve outcomes across the entire OHCA care pathway.
From an EMS system perspective, future implementation should incorporate mechanisms to routinely capture patient-level application activation, responder engagement, AED localization and use, and linkage with clinical outcomes. Such data would allow more precise evaluation of how smartphone-based systems influence specific components of the resuscitation pathway and help distinguish the contribution of the application from concurrent changes in EMS practice. Although propensity score–based methods can help address measured differences between implementation periods, they cannot fully account for unmeasured confounding or temporal changes inherent in uncontrolled before–after studies.
Strengths and limitations
This study has several strengths. It used real-world data from a large urban EMS system and evaluated multiple clinically relevant outcomes spanning early resuscitation and short-term survival. Propensity score weighting and additional adjustment for residual covariate imbalance were used to reduce measured differences between the pre- and post-implementation cohorts. The study also provides evidence regarding the system-level implementation of a locally developed smartphone-based CPR assistance system within an operational EMS setting.24 Several limitations should be acknowledged. First, exposure was defined according to the temporal availability of MySOS rather than confirmed individual-level application activation or use. Consequently, patients in the post-implementation cohort cannot be assumed to have been directly exposed to the application, introducing potential exposure misclassification and limiting attribution of the observed associations specifically to MySOS. Second, this was an uncontrolled before–after observational study using a historical control group, without a concurrent control region or interrupted time-series analysis. Therefore, underlying temporal trends and contemporaneous changes in EMS practice, bystander behavior, data collection, termination-of-resuscitation practices, hospital care, or other system-level factors could not be separated from the implementation period. Third, important differences in prognostic characteristics were present between the study periods, particularly in initial cardiac rhythm and AED- and defibrillation-related characteristics. Although IPTW and additional covariate adjustment improved balance, residual imbalance remained, and unmeasured confounding cannot be excluded. Fourth, cases with missing data required for the primary analysis were excluded, and complete-case analysis may have introduced selection bias if missingness was related to exposure or outcomes. Finally, neurological outcomes, such as Cerebral Performance Category scores, were unavailable, preventing assessment of neurologically favorable survival. In addition, the propensity-score model incorporated selected resuscitation interventions as part of a broader prognostic-balancing strategy. Because these interventions occurred after classification into the implementation period and may have been influenced by factors on the causal pathway, their inclusion may complicate causal interpretation of the weighted estimates; accordingly, these estimates should be regarded as adjusted associations rather than causal effects.
Conclusion
The post-implementation period of the MySOS smartphone-based CPR assistance system was associated with a higher probability of ROSC based on the adjusted absolute risk difference, although the corresponding risk ratio was not statistically significant. No significant associations were observed with 24-h survival, survival to hospital admission, or survival to hospital discharge. Given the uncontrolled before–after design, residual baseline imbalance, potential temporal confounding, and lack of confirmed patient-level MySOS use, these findings should be considered hypothesis-generating rather than causal. Future prospective studies incorporating direct measures of application activation and utilization are warranted.
CRediT authorship contribution statement
Duangpon Thepmanee: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Pramote Papukdee: Writing – review & editing, Methodology. Rossakorn Klaiangthong: Writing – review & editing, Methodology. Piti Chalongviriyalert: Writing – review & editing.
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.
References
- 1.Kiguchi T., Okubo M., Nishiyama C., Maconochie I., Ong M.E.H., Kern K.B., et al. Out-of-hospital cardiac arrest across the world: first report from the International Liaison Committee on Resuscitation (ILCOR) Resuscitation. 2020;152:39–49. doi: 10.1016/j.resuscitation.2020.02.044. [DOI] [PubMed] [Google Scholar]
- 2.Nishiyama C., Kiguchi T., Okubo M., Alihodžić H., Al-Araji R., Baldi E., et al. Three-year trends in out-of-hospital cardiac arrest across the world: second report from the International Liaison Committee on Resuscitation (ILCOR) Resuscitation. 2023;186 doi: 10.1016/j.resuscitation.2023.109757. [DOI] [PubMed] [Google Scholar]
- 3.Wissenberg M., Lippert F.K., Folke F., Weeke P., Hansen C.M., Christensen E.F., et al. Association of national initiatives to improve cardiac arrest management with bystander intervention and survival. JAMA. 2013;310(13):1377–1384. doi: 10.1001/jama.2013.278483. [DOI] [PubMed] [Google Scholar]
- 4.Grasner J.T., Herlitz J., Tjelmeland I.B.M., Wnent J., Masterson S., Lilja G., et al. European Resuscitation Council guidelines 2021: epidemiology of cardiac arrest in Europe. Resuscitation. 2021;161:61–79. doi: 10.1016/j.resuscitation.2021.02.007. [DOI] [PubMed] [Google Scholar]
- 5.Isbye D.L., Rasmussen L.S., Ringsted C., Lippert F.K. Disseminating CPR training among school children. Circulation. 2007;116(12):1380–1385. doi: 10.1161/CIRCULATIONAHA.107.710616. [DOI] [PubMed] [Google Scholar]
- 6.Takahashi H., Ain N., Fook-Chong S., Qiao F., Shahidah N., Okada Y., et al. Impact of smartphone activated first responders on bystander CPR and outcomes. Resuscitation. 2025;212 doi: 10.1016/j.resuscitation.2025.110645. [DOI] [PubMed] [Google Scholar]
- 7.Smith C.M., Wilson M.H., Ghorbangholi A., Hartley-Sharpe C., Gwinnutt C., Dicker B., et al. Smartphone apps to alert volunteers to cardiac arrest: the GoodSAM experience. Resuscitation. 2017;121:118–123. doi: 10.1016/j.resuscitation.2017.10.020. [DOI] [PubMed] [Google Scholar]
- 8.Ringh M., Rosenqvist M., Hollenberg J., Jonsson M., Fredman D., Nordberg P., et al. Mobile-phone dispatch of laypersons for CPR in out-of-hospital cardiac arrest. N Engl J Med. 2015;372:2316–2325. doi: 10.1056/NEJMoa1406038. [DOI] [PubMed] [Google Scholar]
- 9.Tong Q., Zhou M., Liu X., Long J., Li L., Pan X., et al. Mobile applications enhance OHCA outcomes: systematic review and meta-analysis. BMC Health Serv Res. 2025;25(1):256. doi: 10.1186/s12913-025-12416-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Vallianatou L., Kapadohos T., Polikandrioti M., et al. Real-time smartphone guidance improves CPR performance: simulation study. Cureus. 2025;17(9) doi: 10.7759/cureus.92196. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Thepmanee D., Tanaka H., Hiroshi T., Hara T. Development of smartphone applications for bystander CPR in Thailand. J EMS Med. 2022;2(1):21–30. [Google Scholar]
- 12.Boland L.L., Formanek M.B., McGovern P.M., Hannan P.J., et al. Dispatch-assisted CPR performance. Resuscitation. 2022;171:41–48. [Google Scholar]
- 13.Berdowski J., Beekhuis F., Zwinderman A.H., Tijssen J.G.P., Koster R.W. Importance of first link recognition in OHCA. Circulation. 2009;119(15):2096–2102. doi: 10.1161/CIRCULATIONAHA.108.768325. [DOI] [PubMed] [Google Scholar]
- 14.Bray J.E., Grasner J.T., Nolan J.P., Iwami T., Ong M.E.H., Finn J., et al. Cardiac arrest and cardiopulmonary resuscitation outcome reports: 2024 update of the Utstein out-of-hospital cardiac arrest registry template. Circulation. 2024;150(9):e203–e223. doi: 10.1161/CIR.0000000000001243. [DOI] [PubMed] [Google Scholar]
- 15.Ong M.E.H., Shin S.D., Tanaka H., Ma M.H.M., Khruekarnchana P., Hisamuddin N., et al. PAROS study design. Resuscitation. 2011;82(4):420–426. [Google Scholar]
- 16.Ong M.E.H., Shin S.D., De Souza N.N.A., Tanaka H., Nishiuchi T., Song K.J., Ko P.C.-I., Leong B.S.-H., Khunkhlai N., Naroo G.Y., Sarah A.K., Ng Y.Y., Li W.Y., Ma M.H.-M. Outcomes for out-of-hospital cardiac arrests across 7 countries in Asia: the Pan Asian Resuscitation Outcomes Study (PAROS) Resuscitation. 2015;96:100–108. doi: 10.1016/j.resuscitation.2015.07.026. [DOI] [PubMed] [Google Scholar]
- 17.Ng Y.Y., Wah W., Liu N., Zhou S.A., Ho A.F.W., Pek P.P., Shin S.D., Tanaka H., Khunkhlai N., Lin C.-H., Wong K.D., Cai W.W., Ong M.E.H. Associations between gender and cardiac arrest outcomes in pan-Asian out-of-hospital cardiac arrest patients. Resuscitation. 2016;102:116–121. doi: 10.1016/j.resuscitation.2016.03.002. [DOI] [PubMed] [Google Scholar]
- 18.Pluempremchit C. Outcomes of OHCA after PAD program in Thailand. Thai J Emerg Med. 2023;5(1):1–17. [Google Scholar]
- 19.Berdowski J., Blom M.T., Bardai A., Tan H.L., Tijssen J.G., Koster R.W. AED use and survival after OHCA. Circulation. 2011;124(20):2225–2232. doi: 10.1161/CIRCULATIONAHA.110.015545. [DOI] [PubMed] [Google Scholar]
- 20.Perkins G.D., Graesner J.T., Semeraro F., Olasveengen T.M., Soar J., Lott C., et al. ERC guidelines 2021: executive summary. Resuscitation. 2021;161:1–60. doi: 10.1016/j.resuscitation.2021.02.003. [DOI] [PubMed] [Google Scholar]
- 21.Nolan J.P., Sandroni C., Böttiger B.W., Cariou A., Cronberg T., Friberg H., et al. Post-resuscitation care guidelines. Intensive Care Med. 2021;47(4):369–421. doi: 10.1007/s00134-021-06368-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Callaway C.W., Donnino M.W., Fink E.L., Geocadin R.G., Golan E., Kern K.B., et al. AHA guidelines 2020: post-cardiac arrest care. Circulation. 2020;142:S465–S482. doi: 10.1161/CIR.0000000000000262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kragh A.R., Ravn-Finseth J., Claesson A., Hollenberg J., Ringh M., Nordberg P. Public involvement in OHCA response. Resusc Plus. 2023;15 [Google Scholar]
- 24.Andersen L.W., Granfeldt A., Callaway C.W., Bradley S.M., Soar J., Nolan J.P., et al. Implementation science in resuscitation. Resuscitation. 2021;162:327–334. [Google Scholar]