Abstract
Aim
To investigate the role of witnessed arrest as an effect modifier on neurological outcomes following bystander cardiopulmonary resuscitation (CPR) in patients with out-of-hospital cardiac arrest (OHCA).
Methods
This cohort study was conducted using an OHCA patient registry collected between September 2019 and March 2021 in Japan. The effect modification by witnessed arrest was analyzed using additive and multiplicative interactions with risk adjustment. The main outcome was a cerebral performance category of 1 or 2 at 30 days after OHCA.
Results
A total of 7496 patients from 42 hospitals were analyzed. Witnessed arrest and bystander CPR occurred in 3053 (40.7 %) and 3152 (42.0 %) patients, respectively. The adjusted odds ratio (OR) of witnessed arrest and bystander CPR for favorable neurological outcomes was 11.36 (95 % confidence interval [CI]: 7.10–18.17). The ORs for bystander CPR within each stratum of witnessed arrest were 2.38 (95 % CI: 1.34–4.24) for non-witnessed arrest and 3.80 (95 % CI: 2.61–5.55) for witnessed arrest. The additive interaction was 7.01 (95 % CI: 3.34–10.68) and the multiplicative interaction was 1.60 (95 % CI: 0.80–3.17). In the non-home arrest setting, bystander CPR had an adjusted OR of 1.23 (95 % CI: 0.58–2.62) for non-witnessed arrest and 2.77 (95 % CI: 1.69–4.53) for witnessed arrest. The additive interaction was 4.98 (95 % CI: 1.67–8.29) and the multiplicative interaction was 2.25 (95 % CI: 0.91–5.53).
Conclusions
The interaction effect of bystander CPR and witnessed arrest exhibited the second strongest form of interaction. The non-significant effect of bystander CPR in non-witnessed arrest in the non-home cardiac arrest setting warrants further investigation.
Keywords: Effect modification, Witnessed arrest, Bystander cardiopulmonary resuscitation, Out-of-hospital cardiac arrest
Introduction
Out-of-hospital cardiac arrest (OHCA) poses an important threat to public health, and prognostic factors in OHCA patients have been studied extensively. In the prehospital setting, factors such as witnessed arrest and bystander cardiopulmonary resuscitation (CPR) are critical for improving the neurological outcomes of patients.1 Although prognostic factors are well studied, they are often analyzed as individual effects affecting the outcome.2, 3, 4 However, prognostic factors may interact with each other and have a complex effect on the outcome. A simple analytical model is easy to understand, but the reality is often complex. Analyzing the interaction effects can provide deeper insights into the pathway between prognostic factors and outcome.5 Reporting guidelines, such as Strengthening the Reporting of Observational Studies in Epidemiology (STROBE), recommend describing whether interaction effects have been studied.6 Even if interaction effects are analyzed, most studies only report the multiplicative interactions in the form of subgroup analysis; the results of additive interactions are rarely presented.7 However, additive interactions are more informative when assessing the impact on public health because they can be used to identify target populations for more effective and efficient use of resources.5 Whether an effect is stronger in one subgroup than in another is scale-dependent5; therefore, studies that only analyze multiplicative interactions can miss important aspects of subgroup analyses.
It remains unclear whether the effect of bystander CPR differs between witnessed arrest and non-witnessed arrest. The term “bystander” is defined as a person who is not part of an organized emergency response system for cardiac arrest.8 This definition does not specify whether the bystander witnessed the arrest. Although bystander CPR is potentially effective, its effect may vary depending on whether the arrest was witnessed. We hypothesized that bystander CPR is less effective or even ineffective if the arrest is not witnessed. Therefore, the aim of this study was to investigate the effect modification by witnessed arrest on favorable neurological outcomes following bystander CPR. If the effect of bystander CPR depends on the presence of witnessed arrest, different strategies can be developed and employed to improve the outcomes of patients with OHCA from a public health perspective.
Methods
Study design
We conducted an observational cohort study using data from the SOS-KANTO 2017 study. The SOS-KANTO 2017 study was originally a prospective cohort study conducted between September 2019 to March 2021 that aimed to address multiple clinical questions on patients with OHCA in the Kanto region of Japan.9 Our study is a post hoc analysis of the SOS-KANTO 2017 study, which collects clinical information based on the Utstein template. In Japan, a person calls 119 to request an ambulance and activate emergency medical services (EMS). The dispatcher provides telephone CPR instructions if cardiac arrest is suspected. Local fire departments and private organizations also provide CPR courses for citizens. About one-third of the Japanese population aged ≥15 years received certification for government-organized CPR courses in 2020.10
Patients with OHCA of a non-traumatic origin who were aged ≥18 years were included in this analysis. Patients with cardiac arrest witnessed by a member of the EMS were excluded for the reason explained later in the Methods section.
To prepare this manuscript, we followed the STROBE guidelines.6 The present study was approved by the Research Ethics Committee of the University of Tokyo (approval number: 2024363NI; approval date November 15, 2024). HE had full access to all the data in the study and takes responsibility for its integrity and the data analysis.
Outcomes
The primary outcome was a favorable neurological outcome, defined as a cerebral performance category of 1 or 2 at 30 days after OHCA.11 Secondary outcomes included 30-day survival, favorable neurological outcomes at 90 days, and 90-day survival after OHCA.
Variables
The exposures of interest were bystander CPR and witnessed arrest. Bystander CPR was defined as CPR performed by a person who is not a member of an organized emergency response system, in alignment with the definition of the Utstein reporting template.12 Witnessed arrest was defined as cardiac arrest that was seen or heard by another person.12 The definition of witnessed arrest in the Utstein definition includes all persons who witnessed the arrest. However, because this study investigated the interaction with bystander CPR, witnessed arrest by a member of the EMS was excluded because such cases usually receive CPR from the EMS staff, who are considered non-bystanders. Therefore, a combination of witnessed arrest by a member of an EMS and bystander CPR is clinically unlikely. Variables that were considered potential confounders included age, location of arrest, clinical frailty scale, and comorbidities. The location of arrest was divided into home and non-home settings.13 The non-home setting category included all the other patients who were not included in the home setting category. The clinical frailty scale is a 9-point scale that is widely used for frailty assessment.14 The clinical frailty scale was categorized into a scores of ≥5 and <5. A clinical frailty scale of 5 indicates mild frailty and is frequently used as a cutoff value to indicate higher vulnerability to stressors.15 Comorbidities were assessed using the 17 items included in the Charlson Comorbidity Index,16 namely myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatologic diseases, peptic ulcer disease, mild liver disease, diabetes without chronic complications, diabetes with chronic complications, hemiplegia or paraplegia, moderate or severe renal disease, any malignancy including leukemia and lymphoma, moderate or severe liver disease, metastatic solid tumors, and symptomatic acquired immune deficiency syndrome (AIDS). The items were separately entered in the regression model and were not summed up to yield an index score. We also included dialysis in this list of comorbidities. The clinical frailty scale and the presence of comorbidities were assessed as a pre-arrest state. Shockable rhythm is a critical prognostic factor but was assumed to be a mediator between witnessed arrest/bystander CPR and the patient’s outcome,17 and was therefore not included as a confounder. A directed acyclic graph to conceptualize the relationships between variables is provided in Supplementary Fig. 1.
Data analysis
A logistic regression model was used to estimate the association between the exposures and the outcome. The 10 events per variable rule was applied to determine the number of variables that could be entered in the logistic regression analysis.18 To account for variability among hospitals and clustering of patients within hospitals, the hospital identifier was included as a random intercept in the regression model.19 We used a generalized linear mixed effects model with a binomial distribution and the logit link function. The modifying effect of witnessed arrest on bystander CPR was assessed using additive and multiplicative interactions.5 Additive interaction is also known as relative risk due to interaction (RERI) and is calculated using odds ratios (ORs) derived from the regression model as follows: OR11 − OR10 − OR01 + OR00, where OR11 is the OR for patients with witnessed arrest who received bystander CPR, OR00 is the OR for patients with non-witnessed arrest without bystander CPR (which equals 1 as the reference), OR10 is the OR for patients with witnessed arrest without bystander CPR, and OR01 is the OR for patients with non-witnessed arrest who received bystander CPR. Multiplicative interaction was assessed with the OR of the interaction term of witnessed arrest and bystander CPR. A significant OR indicates the presence of multiplicative interaction, defined as OR11 > OR10OR01. This would imply that witnessed arrest and bystander CPR interact with each other and have a synergistic effect on the outcome, yielding a stronger effect than the sum of the individual effects. Although studies rarely assess RERI by comparison with interactions on the multiplicative scale, reporting additive interaction has implications from a public health perspective.5 For example, if resources are restricted, we may know which subgroup to focus on for efficient use of resources. This is a difficult decision when we only have information on multiplicative interaction. The magnitude of the interaction was evaluated using the scale proposed by VanderWeele.20 The strongest form of interaction is present if both RERI and multiplicative interaction are significantly positive. The second strongest interaction occurs if RERI is significantly positive but multiplicative interaction is not significant. The ORs for bystander CPR were also calculated within subgroups of patients according to the presence of witnessed arrest. The results are presented using the format recommended by Knol and VanderWeele.21 ORs can only be used to calculate the RERI if the outcome is relatively rare (<10 %); otherwise, risk ratios should be used.
As a prespecified additional analysis, we conducted the same analysis according to the location of arrest because the effect modification was assumed to differ between patients with arrest in home or non-home settings.13 Patients with missing data were not excluded. Missing data were imputed using multiple imputation, in which missingness was assumed to be missing at random (MAR).22 Multiple imputation with the chained imputation method was used for data imputation.23 Twenty datasets were generated for multiple imputation. Further details are provided in Supplementary Methods 1. Two-sided P-values of ≤0.05 were considered statistically significant. All statistical analyses were conducted using R version 4.4.1 (2024, R Foundation for Statistical Computing, Vienna, Austria).
Results
After applying the exclusion criteria, 7496 patients were eligible for the study (Fig. 1). Witnessed arrest occurred in 3053 patients (40.7 %) and bystander CPR was performed in 3152 patients (42.0 %). The median age was 77 years (interquartile range 67–85 years), and 2839 patients were female (37.9 %). There were 2798 (37.3 %) patients with non-witnessed arrest who did not receive bystander CPR, 1645 (21.9 %) patients with non-witnessed arrest who received bystander CPR, 1546 (21.9 %) patients with witnessed arrest who did not receive bystander CPR, and 1507 (20.1 %) patients with witnessed arrest who received bystander CPR. The clinical characteristics of these four groups are presented in Table 1. The proportion of females and the median age was highest in the group of patients with non-witnessed arrest who received bystander CPR (42.2 % and 79 [interquartile range: 67–86] years, respectively). The proportion of patients with arrest in a home setting and a clinical frailty scale of <5 was greatest in the group of patients with witnessed arrest who received bystander CPR (43.4 % and 53.0 %, respectively). Table 2 shows the outcomes of each group. The proportion of patients with favorable neurological outcomes at 30 days after OHCA was greatest in the group of patients with witnessed arrest who received bystander CPR (9.3 %). This group also had the highest proportions for the secondary outcomes of favorable neurological outcome at 90 days, survival at 30 days, and survival at 90 days (8.4 %, 14.4 %, and 10.6 %, respectively). The numbers of patients with missing data for baseline characteristics and outcomes are provided in Supplementary Tables 1 and 2, respectively.
Fig. 1.
Study population flowchart. EMS, emergency medical service; OHCA, out-of-hospital cardiac arrest.
Table 1.
Baseline characteristics.
| Non-witnessed arrest, without bystander CPR | Non-witnessed arrest, with bystander CPR | Witnessed arrest, without bystander CPR | Witnessed arrest, with bystander CPR | Total | |
|---|---|---|---|---|---|
| n | 2798 | 1645 | 1546 | 1507 | 7496 |
| Female, n (%) | 1062 (38.0) | 695 (42.2) | 541 (35.0) | 541 (35.9) | 2839 (37.9) |
| Age, years (median [IQR]) | 78 [69, 85] | 79 [67, 86] | 76 [66, 84] | 75 [63, 84] | 77 [67, 85] |
| Location of arrest: home, n (%) | 519 (18.5) | 549 (33.4) | 399 (25.8) | 654 (43.4) | 2121 (28.3) |
| Clinical frailty scale <5, n (%) | 1292 (46.2) | 807 (49.1) | 805 (52.1) | 798 (53.0) | 3702 (49.4) |
| Myocardial infarction, n (%) | 44 (1.6) | 17 (1.0) | 48 (3.1) | 73 (4.8) | 182 (2.4) |
| Congestive heart failure, n (%) | 50 (1.8) | 24 (1.5) | 48 (3.1) | 49 (3.3) | 171 (2.3) |
| Peripheral vascular disease, n (%) | 4 (0.1) | 1 (0.1) | 10 (0.6) | 11 (0.7) | 26 (0.3) |
| Cerebrovascular disease, n (%) | 39 (1.4) | 29 (1.8) | 52 (3.4) | 49 (3.3) | 169 (2.3) |
| Dementia, n (%) | 39 (1.4) | 29 (1.8) | 35 (2.3) | 41 (2.7) | 144 (1.9) |
| Chronic pulmonary disease, n (%) | 31 (1.1) | 16 (1.0) | 35 (2.3) | 34 (2.3) | 116 (1.5) |
| Rheumatologic disease, n (%) | 7 (0.3) | 4 (0.2) | 5 (0.3) | 6 (0.4) | 22 (0.3) |
| Peptic ulcer disease, n (%) | 10 (0.4) | 4 (0.2) | 10 (0.6) | 10 (0.7) | 34 (0.5) |
| Mild liver disease, n (%) | 12 (0.4) | 4 (0.2) | 14 (0.9) | 7 (0.5) | 37 (0.5) |
| Diabetes without chronic complications, n (%) | 50 (1.8) | 33 (2.0) | 59 (3.8) | 54 (3.6) | 196 (2.6) |
| Diabetes with chronic complications, n (%) | 29 (1.0) | 19 (1.2) | 24 (1.6) | 35 (2.3) | 107 (1.4) |
| Hemiplegia or paraplegia, n (%) | 4 (0.1) | 4 (0.2) | 9 (0.6) | 7 (0.5) | 24 (0.3) |
| Renal disease, n (%) | 48 (1.7) | 13 (0.8) | 35 (2.3) | 44 (2.9) | 140 (1.9) |
| Dialysis, n (%) | 20 (0.7) | 5 (0.3) | 14 (0.9) | 20 (1.3) | 59 (0.8) |
| Any malignancy including leukemia and lymphoma, n (%) | 4 (0.1) | 0 (0.0) | 1 (0.1) | 2 (0.1) | 7 (0.1) |
| Moderate or severe liver disease, n (%) | 9 (0.3) | 3 (0.2) | 6 (0.4) | 9 (0.6) | 27 (0.4) |
| Metastatic solid tumor, n (%) | 26 (0.9) | 5 (0.3) | 19 (1.2) | 12 (0.8) | 62 (0.8) |
| AIDS, n (%) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 1 (0.1) | 1 (0.0) |
AIDS, acquired immune deficiency syndrome; CPR, cardiopulmonary resuscitation; IQR, interquartile range.
Table 2.
Summary of outcomes.
| Non-witnessed arrest, without bystander CPR | Non-witnessed arrest, with bystander CPR | Witnessed arrest, without bystander CPR | Witnessed arrest, with bystander CPR | Total | |
|---|---|---|---|---|---|
| n | 2798 | 1645 | 1546 | 1507 | 7496 |
| Cerebral performance category ≤2 at 30 days, n (%) | 22 (0.8) | 27 (1.6) | 43 (2.8) | 140 (9.3) | 232 (3.1) |
| Survival at 30 days, n (%) | 63 (2.3) | 53 (3.2) | 88 (5.7) | 217 (14.4) | 421 (5.6) |
| Cerebral performance category ≤2 at 90 days, n (%) | 22 (0.8) | 25 (1.5) | 41 (2.7) | 127 (8.4) | 215 (2.9) |
| Survival at 90 days, n (%) | 36 (1.3) | 32 (1.9) | 59 (3.8) | 159 (10.6) | 286 (3.8) |
CPR, cardiopulmonary arrest.
When all of the potential confounders listed above were included in the regression model, the model failed to converge because of the low numbers of patients with any malignancy, including leukemia and lymphoma, and symptomatic AIDS. Therefore, these two variables were excluded from the model. After adjusting for the confounders, the OR of witnessed arrest and bystander CPR for favorable neurological outcomes at 30 days after OHCA was 11.36 (95 % confidence interval [CI]: 7.10–18.17, P < 0.001). The ORs for bystander CPR within each stratum of witnessed arrest were 2.38 (95 % CI: 1.34–4.24, P = 0.003) for non-witnessed arrest and 3.80 (95 % CI: 2.61–5.55, P < 0.001) for witnessed arrest. The RERI was 7.01 (95 % CI: 3.34–10.68, P < 0.001) and the multiplicative interaction was 1.60 (95 % CI: 0.80–3.17, P = 0.18), corresponding to the second strongest type of interaction of the interaction spectrum. The results of the regression analyses for the secondary outcomes are shown in Supplementary Table 3. The results of the regression analysis for the outcome of favorable neurological outcomes at 90 days after OHCA were similar to the main results. For the survival outcomes at 30 and 90 days, the RERI and multiplicative interactions showed significant positive effects, corresponding to the strongest type of interaction.
The prespecified subgroup analysis for the outcome of favorable neurological outcomes at 30 days after OHCA is shown in Table 3. For cardiac arrest in the home setting, the results were similar to those of the main results. For the non-home cardiac arrest setting, the OR of witnessed arrest and bystander CPR for favorable neurological outcomes at 30 days after OHCA was 8.11 (95 % CI: 4.66–14.12, P < 0.001). The ORs for bystander CPR within each stratum of witnessed arrest were 1.23 (95 % CI: 0.58–2.62, P = 0.59) for non-witnessed arrest and 2.77 (95 % CI: 1.69–4.53, P < 0.001) for witnessed arrest. The RERI was 4.98 (95 % CI: 1.67–8.29, P = 0.002) and the multiplicative interaction was 2.25 (95 % CI: 0.91–5.53, P = 0.08), corresponding to the second strongest type of interaction of the interaction spectrum.
Table 3.
Interaction effects for favorable neurological outcomes at 30 days after OHCA.
| ORs (95 % CI) for no bystander CPR | ORs (95 % CI) for bystander CPR | ORs (95 % CI) for bystander CPR within strata of witnessed arrest | |
|---|---|---|---|
| All patients | |||
| Non-witnessed arrest | 1 (reference) | 2.38 (1.34–4.24); P = 0.003 |
2.38 (1.34–4.24); P = 0.003 |
| Witnessed arrest | 2.98 (1.75–5.07); P < 0.001 |
11.36 (7.10–18.17); P < 0.001 |
3.80 (2.61–5.55); P < 0.001 |
| RERI (95 % CI) | 7.01 (3.34–10.68); P < 0.001 |
||
| Measure of interaction on multiplicative scale (95 % CI) | 1.60 (0.80–3.17); P = 0.18 |
||
| Subgroups by location | |||
| Home | |||
| Non-witnessed arrest | 1 (reference) | 3.31 (1.34–8.18); P = 0.01 |
3.31 (1.34–8.18); P = 0.01 |
| Witnessed arrest | 2.63 (1.05–6.57); P = 0.04 |
12.47 (5.58–27.86); P < 0.001 |
4.74 (2.60–8.65); P < 0.001 |
| RERI (95 % CI) | 7.57 (1.12–14.01); P = 0.01 |
||
| Measure of interaction on multiplicative scale (95 % CI) | 1.43 (0.49–4.21); P = 0.51 |
||
| Non-home | |||
| Non-witnessed arrest | 1 (reference) | 1.23 (0.58–2.62); P = 0.59 |
1.23 (0.58–2.62); P = 0.59 |
| Witnessed arrest | 2.92 (1.61–5.31); P < 0.001 |
8.11 (4.66–14.12); P < 0.001 |
2.77 (1.69–4.53); P < 0.001 |
| RERI (95 % CI) | 4.98 (1.67–8.29); P = 0.002 |
||
| Measure of interaction on multiplicative scale (95 % CI) | 2.25 (0.91–5.53); P = 0.08 |
CI, confidence interval; CPR, cardiopulmonary resuscitation; OHCA, out-of-hospital cardiac arrest; OR, odds ratio; RERI, relative risk due to interaction.
The confounders included in the model were age, sex, clinical frailty scale, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatologic diseases, peptic ulcer disease, mild liver disease, diabetes without chronic complications, diabetes with chronic complications, hemiplegia or paraplegia, moderate or severe renal disease, dialysis, moderate or severe liver disease, and metastatic solid tumors.
The subgroup analyses for the secondary outcomes are provided in Supplementary Table 4. For all secondary outcomes, the non-home setting had non-significant ORs for bystander CPR within the stratum of non-witnessed arrest, in terms of survival at 30 days (1.02; 95 % CI: 0.56–1.88, P = 0.94), favorable neurological outcomes at 90 days (1.51; 95 % CI: 0.62–3.69, P = 0.36), and survival at 90 days (1.10; 95 % CI: 0.49–2.47, P = 0.83). By comparison, positive ORs were observed for bystander CPR within the stratum of witnessed arrest for survival at 30 days (2.59; 95 % CI: 1.75–3.84, P < 0.001), favorable neurological outcomes at 90 days (2.92; 95 % CI: 1.70–5.03, P < 0.001), and survival at 90 days (3.22; 95 % CI: 1.98–5.24, P < 0.001). In the non-home arrest setting, the RERI and multiplicative interaction were significantly positive for the outcomes of survival at 30 and 90 days. The other results indicated a non-significant multiplicative interaction but a significantly positive RERI.
Discussion
This study investigated the interaction effect between witnessed arrest and bystander CPR on favorable neurological outcome, and we found that it had the second strongest form of interaction, namely a significant RERI with no multiplicative interaction. Both witnessed arrest and bystander CPR were positively associated with better neurological outcomes. For each stratum of witnessed arrest, bystander CPR had a greater OR in the witnessed arrest stratum. However, according to the non-significant multiplicative interaction, it was not apparent which effect was greater. Considering that RERI was significantly positive, we may conclude that, if bystander CPR is given to both groups, the number of patients with favorable neurological outcomes will be greater among patients with witnessed arrest than in patients with non-witnessed arrest. From a public health perspective, this suggests that implementing a policy aimed at improving the rate of bystander CPR will have a greater effect for patients with witnessed arrest.
Our results carry two major public health implications. First, if the arrest is witnessed, encouraging bystander CPR at the scene is necessary to maximize the public health impact. Our hypothesis was that bystander CPR would be less effective in patients with non-witnessed arrest and the dispatcher might not need to encourage bystander CPR in such cases. However, this was not confirmed in this study. Encouraging bystander CPR can be achieved by interventions such as dispatcher assistance and volunteer community responders. Dispatcher assistance is when instructions are given to the caller to the EMS for performing CPR to the person in arrest. A systematic review has shown that dispatcher assistance improves the rate of bystander CPR and the neurological outcomes of the patients.24 There may be instances when the caller is unable to perform CPR, for example if the caller is too old. In such cases, volunteer community responders may increase the likelihood that the patient receives CPR.25 Although barriers may still exist, volunteer community responders will be a promising intervention in the future. Second, witnessed arrest is not an easily modifiable factor, but wearable devices that can detect cardiac arrest may offer a potential public health intervention.26, 27 Wearable devices can mimic the effect of a witness and could save more patients when combined with bystander CPR. Although current devices are not yet optimal, further technological advances could provide a substitute for a witness in person.
The prespecified subgroup analysis provides further insight into the role of witnessed arrest and bystander CPR. In the arrest at home setting, the interaction effect was similar to that of the main analysis, but in the non-home setting, the association between bystander CPR and favorable neurological outcomes was not significant for non-witnessed arrest. This indicates that bystander CPR has less effect if an arrest in a non-home setting is not witnessed. The multiplicative interaction was not significant, but the P-value was close to 0.05, which suggests that there might be a difference in the effect of bystander CPR between witnessed and non-witnessed arrest on the multiplicative scale. For the survival outcomes, namely survival at 30 and 90 days after OHCA, the multiplicative interaction was significant in the non-home setting but the effect of bystander CPR was not statistically significant for non-witnessed arrest. Therefore, the location of arrest may be crucial in the effectiveness of bystander CPR, highlighting the need for further investigation to explore the low effectiveness of bystander CPR for non-witnessed arrest in non-home settings. Recognition of the arrest and an early response may be more difficult if the arrest is not witnessed in non-home settings.28 Because the actual time of cardiac arrest is unknown in non-witnessed cases, we can only speculate that the failure to recognize the arrest may delay bystander CPR and diminish its effect. By comparison, in patients with cardiac arrest at home, bystander CPR was still effective even if the arrest was not witnessed. The bystander may be more familiar with the patient and their past medical history than with a stranger outside the home, prompting an early response and initiation of bystander CPR.28 In Japan, people who need help because of disabilities or comorbidities wear a “Help Mark” in public.29 Although the recognition of the Help Mark is still low, a sign worn by a person in public settings may encourage a quicker response following cardiac arrest. From a public health standpoint, non-witnessed arrest in non-home settings represents a greater opportunity for promoting bystander CPR to achieve more favorable neurological outcomes.
This study has several limitations. First, we could not control for some variability that may exist in witnessed arrest and bystander CPR, such as who witnessed the arrest and gave bystander CPR, that may affect the course of recovery from cardiac arrest.30 If the witness is old and does not have the strength or capacity to respond to the arrest and perform effective CPR, it may affect the patient’s neurological outcome. Second, some measurement bias of the exposures and other variables may be present. The people who registered the data may judge the witness and bystander differently.31, 32 However, because we used the hospital identifier as a random intercept, we could at least adjust for the variability among hospitals. Third, although we controlled for as much confounding as permitted by the database, unmeasured confounding may still exist. Fourth, missing data were handled under the MAR assumption, which we considered plausible considering the observed data structure. However, the MAR assumption cannot be empirically verified, and no sensitivity analyses exploring any departure from MAR were conducted. Therefore, the results should be interpreted with caution. Fifth, because the study period included the coronavirus disease 2019 pandemic, the interaction effects might have been affected by the pandemic. Finally, different results may be observed in different settings and different EMS. For example, the Japanese EMS attempts resuscitation in most cases and this may affect the clinical course of the patients, even if the exposure status is the same. Hence, the effect of the interaction may not be homogenous or universal. Further research should be conducted if the environment changes.
Conclusions
This research elucidated the effect of an interaction between bystander CPR and witnessed arrest on favorable neurological outcomes of patients with OHCA. Bystander CPR was significantly effective in patients with witnessed and non-witnessed arrest. The significant RERI without a corresponding multiplicative interaction, the second strongest form of interaction, suggests that although both factors independently contribute to better outcomes, the effect of bystander CPR is particularly pronounced in patients with witnessed arrest because an increase in the number of patients with favorable neurological outcome could be achieved in this setting. This underscores the necessity for public health initiatives aimed at increasing bystander CPR rates in patients with witnessed arrest. Additionally, further research into the effectiveness of bystander CPR in patients with non-witnessed arrest in non-home settings is needed to optimize the interventions and resource allocation.
Declaration of AI and AI-assisted technologies in the writing process
None.
CRediT authorship contribution statement
Hideki Endo: Writing – original draft, Visualization, Methodology, Formal analysis, Data curation, Conceptualization. Takahiro Miyoshi: Writing – review & editing, Conceptualization. Hiroyuki Yamamoto: Writing – review & editing, Conceptualization. Nobuya Kitamura: Writing – review & editing, Project administration. Takashi Tagami: Writing – review & editing, Project administration, Data curation. Kiyotsugu Takuma: Writing – review & editing, Supervision, Conceptualization. Kiyoshi Murata: Writing – review & editing, Supervision.
Funding
This study did not receive any funding.
Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Hideki Endo and Hiroyuki Yamamoto are affiliated with the Department of Healthcare Quality Assessment at the University of Tokyo. This department is a social collaboration department supported by grants from the National Clinical Database, Intuitive Surgical Sarl, Johnson & Johnson K.K., and Nipro Co.
Acknowledgements
The authors would like to thank everyone who participated in the SOS-KANTO 2017 study.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.resplu.2025.101066.
Appendix A. Supplementary material
The following are the Supplementary data to this article:
References
- 1.Sasson C., Rogers M.A.M., Dahl J., Kellermann A.L. Predictors of survival from out-of-hospital cardiac arrest: a systematic review and meta-analysis. Circ Cardiovasc Qual Outcomes. 2010;3:63–81. doi: 10.1161/CIRCOUTCOMES.109.889576. [DOI] [PubMed] [Google Scholar]
- 2.Kragholm K., Wissenberg M., Mortensen R.N., et al. Bystander efforts and 1-year outcomes in out-of-hospital cardiac arrest. N Engl J Med. 2017;376:1737–1747. doi: 10.1056/NEJMoa1601891. [DOI] [PubMed] [Google Scholar]
- 3.Geri G., Fahrenbruch C., Meischke H., et al. Effects of bystander CPR following out-of-hospital cardiac arrest on hospital costs and long-term survival. Resuscitation. 2017;115:129–134. doi: 10.1016/j.resuscitation.2017.04.016. [DOI] [PubMed] [Google Scholar]
- 4.Hollenberg J., Herlitz J., Lindqvist J., et al. Improved survival after out-of-hospital cardiac arrest is associated with an increase in proportion of emergency crew-witnessed cases and bystander cardiopulmonary resuscitation. Circulation. 2008;118:389–396. doi: 10.1161/CIRCULATIONAHA.107.734137. [DOI] [PubMed] [Google Scholar]
- 5.VanderWeele T.J., Knol M.J. A tutorial on interaction. Epidemiol Method. 2014;3:33–72. doi: 10.1515/em-2013-0005. [DOI] [Google Scholar]
- 6.von Elm E., Altman D.G., Egger M., et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Epidemiology. 2007;18:800–804. doi: 10.1097/EDE.0b013e3181577654. [DOI] [PubMed] [Google Scholar]
- 7.Knol M.J., Egger M., Scott P., Geerlings M.I., Vandenbroucke J.P. When one depends on the other: reporting of interaction in case-control and cohort studies. Epidemiology. 2009;20:161–166. doi: 10.1097/EDE.0b013e31818f6651. [DOI] [PubMed] [Google Scholar]
- 8.Grasner J.-T., Bray J.E., Nolan J.P., et al. Cardiac arrest and cardiopulmonary resuscitation outcome reports: 2024 update of the Utstein out-of-hospital cardiac arrest registry template. Resuscitation. 2024;201 doi: 10.1016/j.resuscitation.2024.110288. [DOI] [PubMed] [Google Scholar]
- 9.Kitamura N., Tagami T., Takeda M., Shinozaki K., SOS-KANTO 2017 Study Group Changes of practice on out of hospital cardiopulmonary arrest during the COVID-19 pandemic: a cross-sectional survey of SOS-KANTO 2017 study. Ann Clin Epidemiol. 2017;2024(6):12–16. doi: 10.37737/ace.24003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Yamaguchi T., Nakai M., Kodama T., et al. Impact of a national initiative to provide civilian cardiopulmonary resuscitation training courses on the rates of bystander intervention by citizens and survival after out-of-hospital cardiac arrest. Resuscitation. 2024;195 doi: 10.1016/j.resuscitation.2024.110116. [DOI] [PubMed] [Google Scholar]
- 11.Jennett B., Bond M. Assessment of outcome after severe brain damage. Lancet. 1975;1:480–484. doi: 10.1016/s0140-6736(75)92830-5. [DOI] [PubMed] [Google Scholar]
- 12.Perkins G.D., Jacobs I.G., Nadkarni V.M., et al. Cardiac arrest and cardiopulmonary resuscitation outcome reports: update of the Utstein resuscitation registry templates for out-of-hospital cardiac arrest: a statement for healthcare professionals from a task force of the international liaison committee on resuscitation (American Heart Association, European Resuscitation Council, Australian and New Zealand Council on Resuscitation, Heart and Stroke Foundation of Canada, InterAmerican Heart Foundation, Resuscitation Council of Southern Africa, Resuscitation Council of Asia); and the American heart association emergency cardiovascular care committee and the council on cardiopulmonary, critical care, perioperative and resuscitation. Resuscitation. 2015;96:328–340. doi: 10.1016/j.resuscitation.2014.11.002. [DOI] [PubMed] [Google Scholar]
- 13.Kiyohara K., Nishiyama C., Matsuyama T., et al. Out-of-hospital cardiac arrest at home in Japan. Am J Cardiol. 2019;123:1060–1068. doi: 10.1016/j.amjcard.2018.12.038. [DOI] [PubMed] [Google Scholar]
- 14.Pulok M.H., Theou O., van der Valk A.M., Rockwood K. The role of illness acuity on the association between frailty and mortality in emergency department patients referred to internal medicine. Age Ageing. 2020;49:1071–1079. doi: 10.1093/ageing/afaa089. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hamlyn J., Lowry C., Jackson T.A., Welch C. Outcomes in adults living with frailty receiving cardiopulmonary resuscitation: a systematic review and meta-analysis. Resusc Plus. 2022;11 doi: 10.1016/j.resplu.2022.100266. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Quan H., Sundararajan V., Halfon P., et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43:1130–1139. doi: 10.1097/01.mlr.0000182534.19832.83. [DOI] [PubMed] [Google Scholar]
- 17.Nguyen D.D., Spertus J.A., Kennedy K.F., et al. Association between delays in time to bystander CPR and survival for witnessed cardiac arrest in the United States. Circ Cardiovasc Qual Outcomes. 2024;17 doi: 10.1161/CIRCOUTCOMES.123.010116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Peduzzi P., Concato J., Kemper E., Holford T.R., Feinstein A.R. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996;49:1373–1379. doi: 10.1016/s0895-4356(96)00236-3. [DOI] [PubMed] [Google Scholar]
- 19.Finch W.H., Bolin J.E., Kelly K. 2nd ed. CRC Press; 2019. Multilevel modeling using R. [Google Scholar]
- 20.VanderWeele T.J. The interaction continuum. Epidemiology. 2019;30:648–658. doi: 10.1097/EDE.0000000000001054. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Knol M.J., VanderWeele T.J. Recommendations for presenting analyses of effect modification and interaction. Int J Epidemiol. 2012;41:514–520. doi: 10.1093/ije/dyr218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Little R.J., Rubin D.B. 2nd ed. John Wiley & Sons; 2002. Statistical analysis with missing data. [Google Scholar]
- 23.van Buuren S., Groothuis-Oudshoorn K. Mice: multivariate imputation by chained equations in R. J Stat Softw. 2011;45:1–67. doi: 10.18637/jss.v045.i03. [DOI] [Google Scholar]
- 24.Wang J., Zhang H., Zhao Z., et al. Impact of dispatcher-assisted bystander cardiopulmonary resuscitation with out-of-hospital cardiac arrest: a systemic review and meta-analysis. Prehosp Disaster Med. 2020;35:372–381. doi: 10.1017/S1049023X20000588. [DOI] [PubMed] [Google Scholar]
- 25.Ringh M., Fredman D., Nordberg P., Stark T., Hollenberg J. Mobile phone technology identifies and recruits trained citizens to perform CPR on out-of-hospital cardiac arrest victims prior to ambulance arrival. Resuscitation. 2011;82:1514–1518. doi: 10.1016/j.resuscitation.2011.07.033. [DOI] [PubMed] [Google Scholar]
- 26.Khalili M., Lingawi S., Hutton J., et al. Detecting cardiac states with wearable photoplethysmograms and implications for out-of-hospital cardiac arrest detection. Sci Rep. 2024;14:23185. doi: 10.1038/s41598-024-74117-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Edgar R., Scholte N.T.B., Ebrahimkheil K., et al. Automated cardiac arrest detection using a photoplethysmography wristband: algorithm development and validation in patients with induced circulatory arrest in the DETECT-1 study. Lancet Digit Health. 2024;6:e201–e210. doi: 10.1016/S2589-7500(23)00249-2. [DOI] [PubMed] [Google Scholar]
- 28.Litwin P.E., Eisenberg M.S., Hallstrom A.P., Cummins R.O. The location of collapse and its effect on survival from cardiac arrest. Ann Emerg Med. 1987;16:787–791. doi: 10.1016/s0196-0644(87)80576-0. [DOI] [PubMed] [Google Scholar]
- 29.Hashimoto H., Maeda K., Sato K. Future-oriented thinking promotes positive attitudes toward the “Help Mark” in Japan. Front Rehabilit Sci. 2022;3 doi: 10.3389/fresc.2022.967033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Park H.J., Jeong W.J., Moon H.J., et al. Factors associated with high-quality cardiopulmonary resuscitation performed by bystander. Emerg Med Int. 2020;2020 doi: 10.1155/2020/8356201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Rittenberger J.C., Martin J.R., Kelly L.J., Roth R.N., Hostler D., Callaway C.W. Inter-rater reliability for witnessed collapse and presence of bystander CPR. Resuscitation. 2006;70:410–415. doi: 10.1016/j.resuscitation.2005.12.015. [DOI] [PubMed] [Google Scholar]
- 32.Lewis M.M., Stubbs B.A., Eisenberg M.S. Determining witnessed status for out-of-hospital cardiac arrest. Resuscitation. 2016;109:133–137. doi: 10.1016/j.resuscitation.2016.08.022. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.