Types of studies

We included randomized and quasi‐randomized trials (RCTs and q‐RCTs) that compared routine (usual) emergency medical services (EMS) care versus EMS care plus mobilization of community first responders (CFRs) in instances of out‐of‐hospital cardiac arrest (OHCA). Trials with randomization by cluster were eligible for inclusion, including cluster‐design studies with intervention cross‐over.

A trial was considered eligible if, on the basis of the best available information, we judged that participants followed in the trial were assigned prospectively to either routine EMS care or routine EMS care with the addition of CFR mobilization, using a random or quasi‐random method of allocation (Higgins 2011).

Mobilization of CFRs to OHCA represents a complex community intervention that may rely on organizational structures outside the control of the healthcare system. It is likely that in some instances, it would not be feasible for trial designs to use random allocation with individual participants representing the unit of allocation. For this reason, both randomized and quasi‐randomized trials including cluster methods were eligible for inclusion in this Review.

We excluded studies that primarily considered OHCA due to traumatic causes, as the core interventions provided by CFRs, namely, cardiopulmonary resuscitation (CPR) and early defibrillation, are unlikely to be of significant benefit in such circumstances.

Types of participants

We included adults and children older than four weeks of age who had experienced an OHCA.

We excluded studies primarily considering OHCA in infants at birth.

Types of interventions

We included studies that compared routine EMS care (control group) versus EMS care plus mobilization of CFRs (intervention group) in instances of OHCA.

Community first responders were defined as per the Description of the intervention section (above). Community first responders were individuals within a community that were organized in a framework that offered OHCA care within that community to supplement the standard EMS response.

Mobilization of CFRs to the scene of an OHCA event represented a complex intervention with variation in components depending on the community setting and its system of emergency healthcare delivery. Key features that defined CFRs across different settings and systems of care included the following.

• Community first responders were present in the community where cardiac arrest occurred.

• Community first responders did not have statutory responsibility for cardiac arrest response but rather served to supplement the statutory EMS response.

• Community first responders were mobilized to an OHCA event by an active and predetermined alert mechanism.

In some communities, the statutory EMS or ambulance service is routinely provided by the local fire service. For the purposes of this Review, this group represents statutory EMS, as distinct from CFRs, and was not included as an eligible intervention.

We did not include studies primarily focused on opportunistic bystanders. Individuals who were present at the scene of an OHCA event and who performed CPR according to telephone instruction provided by EMS call takers were not considered to be CFRs.

We also excluded studies primarily assessing the impact of specific additional interventions such as administration of naloxone in narcotic overdose or adrenaline in anaphylaxis.

Types of outcome measures

Electronic searches

We identified RCTs and q‐RCTs through literature searching designed to identify relevant trials, as outlined in Chapter 6.4 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We did not apply restrictions by language or publication status.

We searched the following databases for relevant trials.

• Cochrane Central Register of Controlled Trials (CENTRAL), in the Cochrane Library, on 16 February 2018.

• MEDLINE (Ovid SP, 1946 onwards), on 16 February 2018.

• Embase (Ovid SP, 1974 onwards), on 19 February 2018.

• Web of Science (1960 to present), on 16 February 2018.

We listed the search strategies used for each database in Appendix 1. We updated the search strategy in January 2019 and re‐ran the searches. We screened all new references obtained but detected no additional eligible studies.

We scanned the following trials registries on 20 August 2018 for ongoing and unpublished trials.

• World Health Organization International Clinical Trials Registry Platform (www.who.int/ictrp/en/).

• ClinicalTrials.gov.

Searching other resources

We scanned the reference lists and citations of included trials for further references to additional trials. We also scanned the abstracts of conference proceedings of the American Heart Association and the European Resuscitation Council. When necessary, we contacted trial authors to request additional information.

Selection of studies

Two review authors (TB and MD) independently reviewed all titles and abstracts received to assess potential eligibility, using the inclusion criteria outlined above. We obtained and examined in detail full‐text copies of all papers considered potentially eligible, and we approached authors of trials for additional information when necessary. We resolved disagreements by discussion, and when necessary, we involved a third review author (GB or RS). We have summarized the process of study selection in a PRISMA flowchart (Moher 2009).

Data extraction and management

Three review authors (TB, GB, and MD) independently extracted relevant data using our standard data extraction form (Appendix 2), which we adapted from the version used by the Cochrane Effective Practice and Organisation of Care Group (EPOC 2013).

We collected information on study design, study setting, participant characteristics, eligibility criteria, details of intervention(s) given, outcomes assessed, sources of study funding, and any conflicts of interest. We contacted authors of included trials to request additional information when this was necessary. We resolved any disagreements by discussion and consensus.

Assessment of risk of bias in included studies

Three review authors (TB, GB, and MD) independently assessed the validity of each included trial using the Cochrane 'Risk of bias' tool and provided a summary assessment of risks of bias across studies (Higgins 2011).

We assessed each included trial according to the following domains: sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other potential sources of bias. Where relevant, the latter included sources related to a cluster‐randomized design such as (1) recruitment bias; (2) baseline imbalance; (3) loss of clusters; (4) incorrect analysis; and (5) comparability with individually randomized trials, as outlined in Section 16.3.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). For the domain of 'incomplete outcome data', we assessed risk of bias at the outcome level rather than at the study level.

We considered low risk of bias to represent studies with plausible bias unlikely to seriously alter the results; unclear risk of bias to represent studies with plausible bias that raises doubts about the results; and high risk of bias to represent studies with plausible bias that seriously weakens confidence in the results (Higgins 2011). We resolved any disagreements by discussion and consensus.

We constructed a 'Risk of bias' table and generated plots of risk of bias assessments using Review Manager 5 (Review Manager 2014).

Measures of treatment effect

We used odds ratios (ORs) with 95% confidence intervals (CIs) to measure the following dichotomous outcomes.

• Survival at hospital discharge.

• Survival at 30 days.

• Survival to hospital admission.

• Cardiopulmonary resuscitation performed before EMS arrival.

It was not possible to calculate an OR for 'defibrillation performed before ambulance service arrival', as in the only study that reported this outcome, no cases occurred in the control group.

We planned to group neurological outcomes into categories of favourable (CPC score of 1 or 2) and unfavourable (CPC score of 3, 4, or 5), as suggested in a previous systematic review concerning OHCA (Huang 2014); however, data for this outcome were not available.

Health‐related quality of life can be measured by many different tools, including the Quality of Life Scale, the Personal Wellbeing Index, Short Form‐36, and the Satisfaction With Life Survey (Dronavalli 2015), with potentially varying validity for this target population. We anticipated substantial heterogeneity in measuring this outcome, and for this reason, planned to assess treatment effects of health‐related quality of life by narrative description and tabulation in this Review. Unfortunately, outcome data for health‐related quality of life were not available.

Unit of analysis issues

We planned that if we included studies with multiple treatment groups, we would follow the recommendations of Higgins 2011 and would combine groups to create a single pair‐wise comparison or would select one pair of groups and exclude the other groups. This was not necessary, as we included no such studies.

We included one cluster‐randomized trial and evaluated whether clustering was accounted for in the determination of required sample size, whether assessment for design effect was carried out, and whether methods used in analysis are appropriate to the cluster design. Had such a study been inappropriately analysed, as though randomization was performed by individual rather than by cluster, we would have adhered to the advice provided in the Cochrane Handbook for Systematic Reviews of Interventions, Section 16.3.4 (Higgins 2011), and we would have adjusted for design effect when possible. This would have necessitated a request to investigators for additional individual‐level data to allow assessment of the intraclass correlation coefficient (ICC) in clusters. As we judged the methods used in the analysis to be appropriate, we believed this was not necessary.

Dealing with missing data

When summary statistics were missing, we contacted the first author of the trial to try to retrieve relevant data in the first instance.

When individual studies did not account appropriately for missing data, or did not report how these were handled, we considered whether data were likely to be missing at random or otherwise, and we assessed the resulting risk of bias.

When outcome data were missing and could not be recovered, we adopted the approach suggested in the Cochrane Handbook for Systematic Reviews of Interventions, Section 16.2.1 (Higgins 2011), and we used available‐case analysis. We included data only for participants whose results were known, and we addressed the potential impact of the missing data by using the 'Risk of bias' tool. Ultimately, we considered the potential impact of including such studies in the overall assessment of intervention effect.

Assessment of heterogeneity

We considered clinical heterogeneity, methodological heterogeneity, and statistical heterogeneity as outlined by Higgins 2011.

We addressed clinical heterogeneity through detailed reporting of the diagnostic and clinical definitions and characteristics of the included studies. We planned to conduct meta‐analysis if we considered the included studies to be sufficiently homogeneous for participants, interventions, and outcomes. However, the included studies were too heterogeneous for review authors to proceed.

We assessed methodological heterogeneity using the Cochrane 'Risk of bias' tool.

We planned to assess statistical heterogeneity by considering the consistency of study results, and by examining how this impacted the planned meta‐analysis. However as already stated, the included studies were too heterogeneous for review authors to proceed with meta‐analysis. Formerly, we had planned to use the Chi² test and to consider P < 0.10 to represent significant heterogeneity. We also had planned to use the I² statistic to describe the percentage of variability in effect estimates that is due to heterogeneity rather than to sampling error (chance), and to assess its potential impact on the planned meta‐analysis. We had planned to interpret the I² result in keeping with guidance provided in Section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Assessment of reporting biases

We planned to create a funnel plot to explore publication bias if at least 10 studies were included in the meta‐analysis. However, we found too few studies and did not proceed with meta‐analysis for the reasons already outlined.

Data synthesis

We planned to perform a meta‐analysis; however as already stated, the included studies were too heterogeneous for us to proceed.

We had planned to use a random‐effects meta‐analysis to provide some robustness against the presence of heterogeneity, with inverse variance weighting provided by the DerSimonian‐Laird estimate of between‐study variance (tau²) (DerSimonian 1986), and with all analyses carried out in Review Manager 5 (Review Manager 2014).

As we deemed individual study designs to be too diverse, and thus statistical combination to be inappropriate, we have presented the findings in a narrative fashion.

Subgroup analysis and investigation of heterogeneity

We planned a priori subgroup analysis for our primary outcomes in keeping with the following characteristics and rationale.

• Geographical setting (primarily urban or non‐urban): urban CFR mobilization may allow shorter response time.

• Cadre of CFRs (trained laypersons, police service, fire service, and off‐duty paramedics): this may influence CFR training, scope of the intervention, and response time.

• Community first responders routinely equipped with a defibrillator: defibrillation is a key time‐critical intervention after OHCA.

• Witnessed OHCAs: these are likely to have shorter intervals to initiation of CPR and defibrillation from time of OHCA.

• Age groups (children defined as individuals up to 15 years old vs adults): common causes of OHCA are different in children (Meyer 2012), and this may affect the efficacy of interventions.

However, we did not undertake pooled subgroup analysis because we found insufficient data.

Sensitivity analysis

We planned to perform sensitivity analysis by excluding studies considered to have high risk of bias; however, the heterogeneity of included studies and insufficient data precluded both meta‐analysis and sensitivity analysis.