RESTORE-ZG: Return of Effective Spontaneous Circulation Resuscitation Evaluation Canton of Zug

Overview

This study looked at cardiac arrest that happens outside a hospital. Cardiac arrest means the heart suddenly stops pumping blood. When this happens, a person collapses and needs help right away. The study looked at care given by the emergency medical service of the Canton of Zug in Switzerland. The main goal was to learn what makes it more likely that the heart starts to pump blood again on its own. Doctors call this return of spontaneous circulation (ROSC). The researchers also wanted to describe how well the service performed over four years. The researchers did not test any new treatment. They did not contact any people. Instead, they reviewed records that the service already had, from the years 2022 to 2025. All records were coded so that no person could be identified. The study included 240 people who were treated for cardiac arrest. In about 4 out of 10 people (43 out of 100, or 43%), the heart started to pump blood again during care. The researchers checked whether some factors were linked to this result. These factors were: the heart rhythm found at the start of care how quickly the ambulance reached the person whether someone saw the person collapse whether a bystander or an emergency call handler started cardiopulmonary resuscitation (CPR), which is chest compressions and rescue breaths whether a paramedic or a doctor led the care The heart rhythm at the start was the factor most strongly linked to the heart starting again. The result was similar whether a specially trained paramedic or a doctor led the care. The researchers also wanted to know how well these factors can predict whether the heart will start again. They built two simple tools to estimate the chance of this result. One tool uses information known during the emergency call. The other tool uses information known when the ambulance team arrives. They tested how well the tools worked, and they compared the second tool with a tool that already exists. These tools are early versions. They need to be tested on new patients before they can be used in care. The researchers also looked at whether the result differed by place within the service area and whether it changed over the four years. This study cannot show that one factor causes another. It looked back at past records from one service, so the results may not apply to other places. The study did not follow people after they reached the hospital, so it does not report longer term survival. The findings can help the service review its own care and plan future work.

This study looked at cardiac arrest that happens outside a hospital. Cardiac arrest means the heart suddenly stops pumping blood. When this happens, a person collapses and needs help right away. The study looked at care given by the emergency medical service of the Canton of Zug in Switzerland. The main goal was to learn what makes it more likely that the heart starts to pump blood again on its own. Doctors call this return of spontaneous circulation (ROSC). The researchers also wanted to describe how well the service performed over four years. The researchers did not test any new treatment. They did not contact any people. Instead, they reviewed records that the service already had, from the years 2022 to 2025. All records were coded so that no person could be identified. The study included 240 people who were treated for cardiac arrest. In about 4 out of 10 people (43 out of 100, or 43%), the heart started to pump blood again during care. The researchers checked whether some factors were linked to this result. These factors were: the heart rhythm found at the start of care how quickly the ambulance reached the person whether someone saw the person collapse whether a bystander or an emergency call handler started cardiopulmonary resuscitation (CPR), which is chest compressions and rescue breaths whether a paramedic or a doctor led the care The heart rhythm at the start was the factor most strongly linked to the heart starting again. The result was similar whether a specially trained paramedic or a doctor led the care. The researchers also wanted to know how well these factors can predict whether the heart will start again. They built two simple tools to estimate the chance of this result. One tool uses information known during the emergency call. The other tool uses information known when the ambulance team arrives. They tested how well the tools worked, and they compared the second tool with a tool that already exists. These tools are early versions. They need to be tested on new patients before they can be used in care. The researchers also looked at whether the result differed by place within the service area and whether it changed over the four years. This study cannot show that one factor causes another. It looked back at past records from one service, so the results may not apply to other places. The study did not follow people after they reached the hospital, so it does not report longer term survival. The findings can help the service review its own care and plan future work. Study design and setting RESTORE-ZG is a retrospective, single-centre registry analysis conducted at the emergency medical service (EMS) of the Canton of Zug (Rettungsdienst Kanton Zug, RDZ), Switzerland. The service operates a mixed response model in which critical care paramedics and emergency physicians both lead resuscitation, with helicopter EMS support when indicated, and it covers the Canton of Zug and adjacent parts of the cantons of Zurich and Schwyz. The analysis is a full enumeration of resuscitation-attempted out-of-hospital cardiac arrest managed by the service during the observation period rather than a sample. The study has several objectives: to describe the cohort and the performance of the service, to identify prehospital predictors of return of spontaneous circulation, to compare resuscitation led by critical care paramedics with resuscitation led by emergency physicians, to develop and internally validate two prehospital prognostic scores and compare one of them with an existing score, and to describe the geographic distribution of outcomes and response times. Reporting follows the STROBE statement for observational studies and the structure of the Utstein template for out-of-hospital cardiac arrest. Registry description and standard operating procedures The internal resuscitation registry is maintained continuously alongside routine operations. Cases are documented within the electronic patient care record using a defined set of variables and a structured documentation process. For this analysis the study team extracted the registry, merged it with dispatch time stamps, and validated each record. Data extraction, management, and analysis were performed by the study team. The registry is maintained internally by the service. No third-party certification and no independent on-site audit were performed, and this is stated for transparency. Data sources and source data verification Two routinely maintained operational sources were used. Clinical variables were taken from the internal RDZ resuscitation registry and include initial rhythm, witnessed status, bystander and dispatcher-assisted cardiopulmonary resuscitation, first responder involvement, attending qualification, the UB-ROSC severity score, and the outcome classification. Dispatch time stamps for call receipt, alarm, departure, and arrival at the scene were taken from the Sanostat dispatch system. As source data verification, each case was validated manually against the internal registry, and the outcome and clinical variables were cross-checked against the underlying documentation. Time stamps were sourced directly from the dispatch system rather than re-entered. Data checks and quality assurance Range and consistency checks were applied during data preparation. Time intervals were computed from the dispatch time stamps with correction for midnight crossover, so that a negative raw interval was adjusted rather than retained. Dispatch time values above 60 minutes were treated as implausible and set to missing. Duplicate mission numbers were identified and removed as technical duplicates. Two cases could not be merged into the primary dataset because of non-matching time stamps and were excluded. Missing values were reported and were not imputed for the descriptive analysis. Multiple imputation was used only for the multivariable sensitivity analysis described below. Data dictionary The main variables, their source, and their coding are as follows. Outcome: derived from the National Advisory Committee for Aeronautics (NACA) score in the registry, coded as NACA 6 for return of spontaneous circulation and NACA 7 for death at the scene without sustained circulation. Initial rhythm: registry, six categories (ventricular fibrillation, ventricular tachycardia, pulseless electrical activity, asystole, extreme bradycardia, unknown), grouped as shockable or non-shockable, with the two rarest categories combined into an other category for modelling. Attending qualification: registry, coded as critical care paramedic, emergency physician, or paramedic. Witnessed collapse: registry, yes or no. Bystander cardiopulmonary resuscitation: registry, yes or no. Dispatcher-assisted cardiopulmonary resuscitation: registry, five categories (performed, not performed, declined, not possible, already in progress at the call). Patient status at the time of the call: derived from the dispatcher-assisted cardiopulmonary resuscitation field, coded as not yet in arrest versus already in arrest. First responder dispatched and first responder on scene: registry, yes or no. UB-ROSC severity score: registry, numeric. Age and sex: registry. Call receipt, alarm, departure, and arrival times: Sanostat dispatch system, used to derive dispatch, turnout, response, and travel time in minutes. Incident coordinates: registry, geocoded for the spatial analysis. The registry uses service-defined categorical fields for clinical variables; external terminologies such as the World Health Organization Drug Dictionary or MedDRA were not applicable, because no drug exposure or adverse event coding was required. Sample size assessment The analysis was a full enumeration of all eligible cases during the observation period, so no a priori sample size calculation was performed. The achieved sample was 240 cases. Post-hoc power was estimated for the primary comparisons. At the achieved sample size, the minimum detectable effect at 80 percent power was approximately a standardised mean difference of 0.363 for continuous comparisons and an effect size w of about 0.181 for categorical comparisons. Analyses with limited power are identified as such in the reporting. Plan for missing data Variables with missing values were reported as missing and were not filled in for the descriptive analysis. The multivariable models were fitted on complete cases, and a sensitivity analysis then addressed missing values using multiple imputation by chained equations with predictive mean matching and five imputations. The variables carrying missing values in the restricted model related mainly to bystander and dispatcher-assisted cardiopulmonary resuscitation. The direction and significance of the main effects were compared between the complete-case and imputed analyses to assess the influence of missing data. Statistical analysis plan * Descriptive and univariable analysis: Categorical variables were summarised as counts and percentages and non-normally distributed continuous variables as median and interquartile range. Univariable associations were tested with the chi-squared or Fisher exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. * Multivariable analysis: Two multivariable binary logistic regression models were fitted. Following the analysis code, the outcome was coded with No ROSC as the modelled event. Model 1 included all cases and adjusted for initial rhythm, witnessed collapse, patient status at the call, attending qualification, response time, and turnout time. Model 2 was restricted to cases already in cardiac arrest at the time of the call and additionally included bystander and dispatcher-assisted cardiopulmonary resuscitation status. Model discrimination was summarised by the area under the receiver operating characteristic curve with 95 percent confidence interval by the DeLong method. * Response time analysis: For response time, the area under the curve and the Youden-optimal threshold were estimated, and the response-time effect was illustrated with a descriptive time-to-event visualisation that is not a survival analysis. A prespecified subgroup analysis compared the response-time association with the outcome in shockable and non-shockable rhythms. * Comparison by attending qualification: Resuscitation led by critical care paramedics and by emergency physicians was compared using the Fisher exact test and the unadjusted odds ratio. Confounding was assessed by comparing the UB-ROSC severity score between qualification groups with the Kruskal-Wallis test and by comparing the rhythm distribution. The comparison was supported by a 1:1 nearest-neighbour propensity score matched analysis with a conditional logistic model, by an E-value for the upper confidence bound of the unadjusted odds ratio, and by an exploratory likelihood ratio test for a qualification by rhythm interaction. * Prognostic score development and validation: Two prehospital prognostic scores were developed with logistic regression: a dispatch score, intended for use during the emergency call, and an on-scene score, intended for use on arrival of the team. Internal validation used non-parametric bootstrap resampling with 1000 resamples, and an optimism-corrected c-statistic and a calibration assessment were reported. Each continuous model was converted to an integer point system with defined risk groups, and the monotonic gradient across point levels was tested with the Cochran-Armitage trend test. The on-scene score was compared with the existing UB-ROSC score on the shared sample using the DeLong test, an equivalence test based on two one-sided tests, and reclassification metrics, namely the net reclassification improvement and the integrated discrimination improvement. * Temporal and spatial analysis: Temporal stability of the annual outcome rate was tested with the Cochran-Armitage trend test. A spatial analysis geocoded incident locations and summarised case counts, outcome rates, and response times across the service area and by canton. All tests were two-sided and a p value below 0.05 was considered statistically significant. Analyses were performed in R version 4.5. Data handling The analysis used pseudonymised data, and only aggregated results without personal identifiers are reported. Reporting The study addresses several objectives, and the results are reported in more than one publication, each of which references this registration.