Introduction

Out of hospital cardiac arrest (OHCA) has a devastatingly high mortality rate [1]. Survival to hospital discharge ranges between countries from < 1% [2] to 25% in the best European centres [3], reflecting differences in case identification, demography, geography and emergency service provision [4]. Reducing the mortality associated with OHCA is a strategic priority of many countries [5,6,7,8,9,10].

Prompt, effective bystander cardiopulmonary resuscitation (CPR) is the most important factor determining survival from OHCA, increasing survival almost 4-fold [11, 12]. Registry data show most OHCA occur at home [2, 13, 14]. Even the most prompt emergency medical response will take at least a few minutes (median 6 mins.) [15], and so the response of others in the home is critical.

Governments and charities invest significantly in training lay-people in CPR [16,17,18]. Despite this, those in OHCA often do not receive CPR prior to the arrival of emergency services [19]. Even amongst those who are trained, less than half attempt CPR when required [20]. Increasing the proportion of lay-people trained in CPR who actually apply their skills in a real emergency situation is essential [21] as otherwise much of the effort expended in training lay-people will not improve outcomes for patients.

Research relating to CPR training of lay-people has largely been concerned with increasing knowledge and achieving competence in the skill of CPR. Questions of how best to teach CPR tend to be answered by studies using skills performance (e.g. compression depth) and assessment of knowledge as outcome measures [22, 23]. However the International Liaison Committee on Resuscitation [24] and behavioural science [25] would suggest that psychological factors (e.g. people’s attitudes about CPR) are likely to be critical in explaining whether or not people initiate CPR. To date there has not been a systematic synthesis of this literature.

The aim of this review was to synthesise evidence relating to lay-people initiating CPR and to identify the psychological and behavioural factors that facilitate or inhibit people’s willingness to perform CPR.

Method

Protocol and registration

In line with best practice, a review protocol was published (2018) and registered with the PROSPERO International Prospective Register of systematic reviews (protocol number 117438): https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=117438.

Eligibility criteria

Inclusion

Types of study

All primary study designs.

Types of participants

Lay members of the public (i.e. not healthcare professionals or others who receive CPR training as a part of their job, e.g. lifeguards) of any age.

Types of outcome measure

Studies which contained psychological/behavioural data (not CPR knowledge or training status) related to 1) why the participants did or did not perform CPR in real emergencies or 2) would or would not perform CPR in a hypothetical or simulated situation. CPR was defined as performing chest compressions (CC), mouth-to-mouth ventilations, applying an Automated External Defibrillator (AED) or any combination of these.

Exclusion

Papers which did not report a primary empirical study (e.g. reviews, editorials, opinion pieces) were excluded.

Information sources and search strategy

Six electronic databases - Cochrane Library, MEDLINE, EMBASE, CINAHL, PsycInfo and Google Scholar- were searched for publications from inception of each database to 13th December 2019 (search strategy is supplied in supplementary materials Additional file 1). Supplementary searches included: a) reference lists of included studies, b) citations of included studies (Science Citation Index (SCI), Social Sciences Citation Index (SSCI) and Arts and Humanities Citation Index (A&HCI), c) hand-searches of titles (Jan 2005 – Jan 2020) of Resuscitation and a further update database search performed 01/06/21.

Study selection

Screening of titles was undertaken independently by two reviewers (BF and DD) to exclude titles that were obviously irrelevant. The inclusion/exclusion criteria were applied to abstracts of studies and irrelevant abstracts were excluded. Inter-rater agreement kappa was 0.85, pabak kappa = 0.85. Full texts considered potentially relevant by either reviewer were screened independently (BF and DD). At full-text stage, any disagreements between the reviewers were resolved by discussion.

Assessment of methodological quality and risk of bias

The methodological quality of studies was assessed using the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for quantitative studies [26] and the Joanna Briggs Institute’s Quality Assessment and Review Instrument (QARI) for qualitative studies [27]. Included studies were independently assessed by two reviewers (BF and CT) for methodological quality, with discrepancies being resolved through discussion.

Data extraction

Guided by the CONSORT guidelines [28] and the published protocol, the following data were extracted for each study: study details (author & date, location, study duration, objectives), study methods (design, setting, target population, sample size estimation, actual sample size, sampling and recruitment method, behavioural and psychological data, analysis, dates of recruitment) and study results.

BF and SM independently performed data extraction on 20% of the included studies (n = 20) to assess reliability. No discrepancies in independently extracted data were found and the remainder were extracted by a single researcher (BF or SM).

Synthesis and analysis

Behavioural and psychological factors identified during extraction were grouped into conceptually similar ‘factors’ by BF: 51 individual factors were identified. To facilitate interpretation, this large number of factors were grouped using categorisations or domains from the Theoretical Domains Framework Version 2 [29] (a validated, comprehensive, theory-informed approach to identifying determinants of behaviour). Definition of domains referred to in this paper are provided in Box 1. Domain categorisations were confirmed by a second reviewer (DD).

Included studies were differentiated according to the study population, study design and whether factors were identified by participants in response to an open question or endorsed from a list of factors presented by researchers. In order to facilitate comparisons studies were grouped according to the summary statistics used and p-values and Odds Ratios compared where possible. We prioritised 1) the most ecologically valid data [30, 31] (i.e. real-life OHCA calls and accounts of people who had actually witnessed OHCA), 2) studies which formally assessed posited relationships and 3) methodologically strong studies (i.e. assessed as low risk of bias) in the findings section.

Results

Original database searches conducted on 13th Dec 2018 (see PRISMA diagram, Fig. 1) identified 17,309 citations with 87 studies included after screening for eligibility. An update search conducted 01/06/21 identified 1119 additional titles, 15 of which were assessed as eligible. Hand-searching of Resuscitation (Jan 2005-Dec 2021) identified 96 potentially relevant titles, seven of which had not already been identified by database screening, none met the inclusion criteria. Reference lists of included studies identified an additional 136 papers, 26 of which had not been previously identified, two studies were eligible and included. Finally, citation tracking identified 35 potentially relevant titles, seven not previously screened and one study included. Therefore, a total of 105 studies were included in the narrative synthesis.

Fig. 1
figure 1

PRISMA diagram

Full size image

Description of included studies

Table 1 summarises the main characteristics of the 105 included studies comprising a total of 150,820 participants. The studies were published between 1989 and 2021 and conducted across 30 countries. The studies were heterogenous in design and included: randomised controlled trials (n = 6); non-randomised trials (n = 1); a quasi-experimental deign (n = 1), prospective cohort study (n = 1); before and after studies (n = 15); cross sectional studies (n = 67), qualitative studies (n = 9) and studies examining actual OHCA calls to Emergency Medical Services (n = 5).

Table 1 List of included studies
Full size table

Methodological quality

Of the quantitative studies, four [58, 69, 103, 110] were identified as strong, six as moderate [83, 90, 102, 117, 125, 131] with the remaining 87 quantitative studies rated ‘weak’ (see Table 2). There was a predominance of non-randomised designs, uncontrolled confounders, and use of unvalidated data collection methods. All qualitative studies were assessed as of sufficient quality for inclusion but also varied in quality (n = 8).

Table 2 EPHPP Quality Assessment of included studies
Full size table

The psychological and behavioural factors identified from the included studies are reported below and summarised in Tables 3, 4, 5, 6, 7, 8, 9 and 10 below. Studies were divided into subgroups according to the study population (i.e. results from those with direct experience versus general samples responding to a ‘hypothetical’ OHCA); study design and statistics used. Data were further categorised depending on whether the ‘predictor’ was identified by participants in response to an open question or whether it was presented as a possible factor and subsequently endorsed. Factors are presented in relation to the domains of the Theoretical Domains Framework so that theoretically similar factors are grouped together and can be compared across study designs (Fig. 2).

Table 3 Psychological and behavioural factors associated with LOWER actual/intended CPR initiation (grouped using Theoretical Domains Framework V.2 [29])
Full size table
Table 4 Psychological and behavioural factors associated with GREATER actual/intended CPR initiation (grouped by Theoretical Domains Framework V.2 [29])
Full size table
Table 5 Studies which formally assess association of variables with measures of CPR initiation/intention (grouped by Theoretical Domains Framework V.2 [29])
Full size table
Table 6 Summary of studies exploring relationship to victim (Domain 11. Environmental context and resources [29])
Full size table
Table 7 Studies exploring relationship with victim (Likert Scale) ((Domain 11. Environmental context and resources [29])
Full size table
Table 8 Studies exploring mouth-to-mouth ventilation as a deterrent (Domain 6. Beliefs about consequences [29])
Full size table
Table 9 Studies exploring mouth-to-mouth ventilation as a deterrent (Likert Scale) ((Domain 6. Beliefs about consequences [29])
Full size table
Table 10 Studies exploring disagreeable characteristics (Domain 11. Environmental context and resources [29])
Full size table
Fig. 2
figure 2

Theoretical Domains Framework definitions [29]

Full size image

Studies involving those with direct experience of OHCA

Sixteen studies involving people with direct experience of OHCA were identified. These included five studies which analysed recorded calls involving OHCA [47, 58, 67, 83, 110], four qualitative studies exploring the experiences of people who had witnessed an OHCA [93, 94, 98, 130] and seven cross-sectional surveys which asked open questions about people’s experiences of facilitators and barriers to them having performed CPR [20, 36, 46, 95, 122, 126, 136].

Real-life calls

TDF domain 4: beliefs about capabilities

Limitations in the physical capacity of the caller was also identified in all five studies. Physical capability was a barrier to CPR in 15% [47], 51% [58], 11% [67], 35% [110] and 8% [83] of calls. Difficulties moving the person who had collapsed to a flat position in order to perform CPR and the rescuer being frail or with a condition making CPR difficult were described. Uncertainty about whether cardiac arrest was happening (e.g. person still making some respiratory sounds) was reported in 28% of calls by Case (2018) [47] and in 6% by Hauff (2003) [67].

Case (2018) [47] reported that “many callers” reported a lack of confidence.

TDF domain 6: beliefs about consequences

Concerns that CPR was futile (e.g. that the person was already dead/beyond help) were reported in 50% of calls analysed by Riou et al. (2020) [110], in 28% of calls analysed by Case (2018) [47] and in 23% by Hauff (2003) [67]. Concern about infection (4%) [58], fear of doing harm (3%) and fear of legal consequences (1%) [83] were reported in a small minority of calls.

TDF domain 11: environmental context

Disagreeable characteristics associated with the victim was identified as a factor in 3% [83] and 2% [67] of calls.

TDF domain 13: emotion

All five studies of real-life calls analysed calls where the layperson hesitated or refused to provide CPR identified the strong emotion of the situation as a factor that prevented initiation of CPR. Elements of emotional distress, such as panic, upset and stress were identified in 20% [47], 42% [58], 11% [67] and 14% [83] of calls where callers expressed reluctance. ‘Being shaken’ and ‘fear’ were described in 2 example quotations by Riou et al. (2020) [110].

Qualitative studies of people who have witnessed OHCA

Four qualitative accounts of people’s experiences of encountering OHCA and CPR were identified [93, 94, 98, 130] comprising interviews with a total of 107 participants (aged 24 [93] to 87 [130]).

TDF domain 2: skills

Feeling unprepared as to what to expect in a cardiac arrest was a theme identified by Mausz (2018) [94] and Moller (2014) [98], in particular that reality was very different from training with a manikin [98].

TDF domain 3: social/professional role and identity

A sense of community or social responsibility were described as encouraging performance of CPR, some stating it was expected of any responsible citizen [93].

TDF domain 4: beliefs about capabilities

Problems identifying whether cardiac arrest had actually occurred (and thus whether CPR was indicated) were identified [93, 94].

TDF domain 6: beliefs about consequences

Fear of doing the patient harm was identified as a cause for hesitation [130]. Recognising the extreme seriousness of the situation led people to erroneously assume that the person was already dead and that CPR would be futile [130]. However, anticipating feeling guilty if they didn’t perform CPR and the person died as a result was a motivation for participants [93].

Concerns about personal safety [93] and liability in the context of a workplace [94] were also expressed.

TDF domain 11: emotion

Participants also described experiencing panic and extreme emotions which inhibited their ability to perform CPR actions [94, 130].

Cross-sectional surveys

Eight cross sectional surveys included analyses of barriers and facilitators of CPR identified by participants who had direct experience of OHCA [20, 46, 93, 95, 100, 122, 126, 136]. Issues identified were very similar to those already described above in the qualitative studies:

Studies of participants where direct experience of CPR was not required

Studies examining the relationship between psychological/behavioural variables and willingness/confidence/intention to perform CPR

Thirteen studies formally explored the relationship between behavioural and psychological predictor variables and willingness to initiate CPR (see Table 5).

TDF domain 1: knowledge

Knowing the importance of CPR (OR 1.9) was positively and significantly related to willingness to perform CPR [80].

TDF domain 2: skills

Having previous experience of CPR or OHCA was the strongest predictor of anticipated willingness to perform CPR [80, 113, 116]. Odds ratios across four studies ranged from 1.5 [68] to 4.8 [113].

TDF domain 4: beliefs about capabilities

Those with good self-rated health status (AOR, 1.26) were more likely to report that they could provide bystander CPR than those reporting poor health [111] and feeling confident (OR 1.9) [119] was positively and significantly related to willingness to perform CPR [97]. Perceiving a lack of expertise was negatively related to willingness (OR 0.6) [119]. Nolan et al. (1999) [101] also showed that confidence differed significantly between those willing and unwilling to initiate CPR. Those unwilling to act also perceived a greater number of psychosocial barriers than those willing (p ≤ .05).

Vaillancourt (2013) [131] and Magid (2019) [92] explored the ability of constructs from the Theory of Planned Behaviour, to predict intention to perform CPR in the event of a cardiac arrest. Attitudes (e.g. I could save someone’s life with CPR) were with the strongest predictor of respondents’ intentions to perform CPR (OR 1.63) identified by Vaillancourt (2013) and also found to be significant in predicting intention to perform CPR. Similarly, both Vaillancourt (2013) and Magid (2019) found higher control beliefs (I feel confident in my abilities to perform CPR) to be significantly related to increased intentions to perform CPR on a cardiac arrest victim (OR 1.16) [131].

Domain 12: social influences

Normative beliefs (derived from the Theory of Planned Behaviour (TPB)) (e.g. My friends and family expect me to do CPR) were found to be modestly but significantly related to people’s intentions to perform CPR on a cardiac arrest victim (OR: 1.07) [131]. Magid (2019) [92] also found subjective norms predictive of intention to perform CPR.

TDF domain 13: emotion

Nolan et al. (1990) [101] showed that confidence differed significantly between those willing and unwilling to initiate CPR. Those unwilling to act anticipated a higher number of negative emotions (afraid, sad, angry, anxious, confused) if they were to perform CPR compared to those who were willing to act (p ≤ .02).

Studies which have compared responses to scenarios -varying psychological/behavioural factors

Sixty-two studies explored a variety of other factors related to willingness to perform CPR (see Tables 6, 7, 8, 9 and 10). Respondents were more willing to perform CPR on their family and friends compared to strangers and in situations that did not involve disagreeable characteristics (TDF Domain 11: Environmental context and resources).

Respondents were more willing to perform compression-only CPR compared to mouth-to-mouth CPR and in situations where there was a perceived risk of transmissible infection willingness to perform CPR was reduced, e.g. after a SARS outbreak (TDF Domain 6: Beliefs about consequences).

Studies of people’s anticipated barriers and facilitators to CPR

Qualitative studies

Four studies provided qualitative accounts of people’s perceptions of CPR [61, 114, 115, 135]. Many of the barriers anticipated by participants in these studies were similar to those identified by people with direct experience of OHCA, as reported above. Additionally, issues around a general fear of ‘getting involved’ with possible consequences in relation to immigration status/law enforcement [115] were identified (TDF Domain 6: Beliefs about Consequences).

Cross-sectional data

Twelve studies [32, 39, 51, 60, 71, 73,74,75, 84, 85, 91, 104] explored the reasons people indicated a reluctance or unwillingness to perform CPR using open questions (rather than presenting possible reasons).

  • Unprompted reasons provided by those categorised as ‘unwilling’

  • TDF domain 4: beliefs about capability Concerns around capability were reported by 11% of unwilling high school students [104], and by 45% of those not willing to use an AED [91]. Concerns about physical capability in particular were reported by 11% of unwilling general public [74]. Low confidence was also reported (4%) [74] and 6–12% [51].

    TDF domain 6: beliefs about consequences The reasons most commonly volunteered by those categorised as unwilling were concerns about to the risk to self: 56% of unwilling general public [74] with 24% [51], 35% [71] and 19% [74] concerned about the risk of infection in particular. Concerns about doing harm to the casualty were reported by 25% [71] and 23% [104]. Legal concerns were reported by 13% [71] and 19% [74] of people unwilling to provide CC and by 16% [71] and 4% [74] of those unwilling to provide mouth-to-mouth ventilation. CPR violating beliefs about death were also reported (4%) [71].

    TDF domain 13: emotion Being too stressed (4%) [91] was also reported as a reason for unwillingness.

  • Prompted reasons

  • The reasons for not performing CPR most commonly proposed by researchers were: fear of doing harm (27 studies); concerns about infection (29 studies); legal concerns (24 studies); concerns about capability (26 studies) and concerns about mouth-to-mouth ventilation (10 studies). Averaging across the studies, the reasons endorsed by the largest proportion of unwilling participants were Lack of confidence (TDF Domain 4: Beliefs about capabilities), Fear of doing it wrong (TDF Domain 6: Beliefs about consequences) and Concerns about capability (TDF Domain 4: Beliefs about capabilities).

Discussion

We have conducted a comprehensive, high-quality, pre-registered systematic review of the psychological and behavioural factors relating to initiation of CPR. This provides a useful synthesis of the evidence to date and identifies promising avenues for intervention and further research. The prominence of two themes: the overwhelming emotion of the OHCA situation and concerns about physical capability in the more methodologically strong studies [58, 83] and evident across the various designs suggests these may be particularly important to address in order to increase CPR initiation.

Emotion of the situation

All five studies [47, 58, 67, 83] that analysed call-recordings involving actual CPR attempts identified the emotion of the situation as an important factor delaying initiation of CPR, as did studies of people who had witnessed OHCA [20, 94, 130].

In hypothetical studies, the expectation of high emotion was significantly associated with not being prepared to act [101] and identified as a likely barrier to CPR by high school students [32]. However, interestingly the potential impact of strong emotions was not frequently anticipated by those without experience of CPR (even when prompted) suggesting people may under-estimate the impact of emotion on their behaviour. Helping people to prepare for the unanticipated impact of strong emotions and providing strategies to perform CPR despite their emotional response might be helpful.

Concerns about capability

Concerns about physical capability were identified as a barrier to initiation in all five studies that analysed emergency call recordings [47, 58, 67, 83], identified in a survey of people who had witnesses an OHCA [20] and provided unprompted as an issue by 11% of the general public [74]. Further, those with good self-rated health were more likely to report being able to perform CPR than those with poor health [111]. Evidence also identified that feeling confident about one’s capability [119] and self-perceived capability [97, 117] are associated with increased willingness to perform CPR and conversely that a lack of confidence reduces willingness [101]. Concerns about capability were identified unprompted by 11% of students [104] and endorsed when prompted by up to 80% of participants. This triangulation of evidence from very different sources suggests concerns about capability as a key issue. Concerns may reflect actual physical limitations amongst potential rescuers but are also likely to reflect people’s beliefs about their capabilities; both are amenable to intervention but importantly will require very different approaches.

Predictors of CPR that have been formally tested

Studies which statistically tested the relationship between variables of interest and intention to perform CPR or actual behaviour were few, highlighting a need for more definitive studies to confirm posited relationships. Previous experience in performing [80, 113] or witnessing CPR [116] and self-perceived ability [97] were the variables most strongly associated with willingness suggesting interventions that improve perceptions of capability may be helpful.

Six studies found evidence to support predictors derived from behavioural theory such as the Theory of Planned Behaviour [137], highlighting the potential utility of an approach to intervention that is based on behavioural theory. Positive attitudes about CPR [92, 106, 131], perceived behavioural control [92, 131] and normative beliefs [92, 131] were significantly associated with intention to perform CPR and Magid (2019) [92] found the theory accounted for 51% of the variance in intention to perform CPR overall. These belief-based constructs are amenable to change and thus are promising targets for intervention. Resources such as the Behaviour Change Technique Taxonomy [138] and the Theory and Techniques resource (https://theoryandtechniquetool.humanbehaviourchange.org/) are available to help researchers and practitioners identify techniques to include in interventions based on their likely mode of action and their likely effectiveness to change the behaviour of interest (in this case initiation of CPR) in the required situation of OHCA.

Overall, it was notable how few papers explicitly discussed underlying theory and how multiple terms were used to refer to highly similar constructs (e.g. intention, willingness, readiness, prepared to act, capable in an emergency). Construct proliferation [139] and lack of precision in defining and labelling of constructs limits our collective ability to synthesise available evidence and to build a cumulative science [140]. This may lead to wasteful duplication of effort and hinder our ability to identify factors that increase initiation of CPR and, importantly, the factors that make initiation of CPR less likely. Greater attention to robust study design, explicit use of theory or at least consistent definitions of terms might bring us more quickly to our collective goal of increasing CPR initiation.

Limitations

This review is limited as we have only assessed published materials. There is thus the potential that publication bias has resulted in studies with negative findings being less likely to be identified [141]. We identified a preponderance of cross-sectional surveys using unvalidated measures with relatively little formal testing of posited ‘predictors’ meaning that it is difficult to draw robust and reliable conclusions from the literature.

Conclusion

Many psychological and behavioural factors associated with CPR initiation can be identified from the current literature with varying degrees of supporting evidence. Preparing people to manage strong emotions and increasing their perceptions of capability are likely important foci for interventions aiming to increase CPR initiation.

Greater use of theory and more robust study designs would strengthen knowledge in this area.

PROSPERO registration number: CRD42018117438.