Documented Use of Emergency Medical Dispatch Protocols is Associated with Improved Survival in Out of Hospital Cardiac Arrest

Abstract Objective There are over 300,000 out-of-hospital cardiac arrests (OHCA) annually in the United States (US) and despite many scientific advances in the field, the survival rate remains low. We seek to determine if return of spontaneous circulation (ROSC) is higher when use of emergency medical dispatch (EMD) protocols is documented for OHCA calls compared to when no EMD protocol use is documented. We also seek identify care-related processes that differ in calls that use EMD protocols. Methods This is a retrospective cohort study of U.S. adults with OHCA prior to emergency medical services (EMS) arrival using 2019 National EMS Information System data. The primary exposure was EMD usage during EMS call. The primary outcome was prehospital ROSC, and secondary outcomes included automated external defibrillator (AED) use before EMS arrival, bystander CPR, and end-of-event EMS survival (survival to the end of the EMS care at transport destination). Multivariable logistic regression adjusted for age, sex, race/ethnicity, primary insurance, rurality, initial rhythm, arrest etiology, and witnessed arrest. Results Of the 96,269 OHCA cases included, EMD use was documented in 73%. Overall, 26% of subjects achieved ROSC in EMS care. EMD subjects were more likely to achieve ROSC (27.2% vs. 23.5%, uOR 1.22, 95%CI 1.18 − 1.26) even after adjusting for subject and arrest characteristics (aOR 1.13, 95%CI 1.08 − 1.17). EMD subjects also had higher end-of-event survival (19.1% vs. 16.4%, aOR 1.20, 95%CI 1.15 − 1.25). AED use before EMS arrival was more common in the EMD group (28.3% vs. 26.3% %diff 2.0, 95%CI 1.4 to 2.6), as was CPR before EMS arrival (63.8% vs. 55.1%, difference 8.6%, 95%CI 7.9 to 9.3%). Conclusions In this retrospective analysis, the rate of ROSC was higher in adult OHCA patients when EMD protocol use was reported compared to when it was not reported. The group with documented EMD use also experienced higher rates of bystander AED use, bystander CPR, and end-of-event survival.


Introduction
Management of out-of-hospital cardiac arrest (OHCA) is one of the most challenging clinical encounters for emergency medical services (EMS).Despite advances in care and decades of research, overall OHCA survival with good neurologic outcome remains low, and survival to discharge remains a target for ongoing improvement (1).In 2021, there were over 300,000 OHCA in the United States (U.S.), with only 10.5% surviving to hospital discharge (2,3).There is significant regional variation in EMS systems, with survival rates ranging from less than 6% to as high as 39.9% (4).
Public safety answering points (PSAP) are typically the first point of contact during OHCA and have an opportunity to guide callers in basic resuscitation techniques.The importance of PSAPs and their role in OHCA care was highlighted by Kurz et al. who noted that telecommunicators were the true first responders and a critical link in the cardiac arrest chain of survival (5).There is a growing body of literature detailing patient-level interventions to improve outcomes in OHCA, but little attention is given to the effectiveness of emergency dispatch interventions in improving the lay response.One such system-level emergency dispatch intervention is the use of emergency medical dispatch (EMD) protocols, which provide standardized just-in-time guidance to a caller while summoning EMS to the OHCA scene.There are several EMD products commercially available that ask structured questions to determine if OHCA is present, and then provide subsequent structured pre-arrival instructions and dispatch life support.Currently, there is no U.S. national standard for medical protocols for emergency telecommunicators, and not all PSAPs use EMD protocols in response to possible OHCA calls.
This study tests the hypothesis that the proportion of return of spontaneous circulation (ROSC) is higher when EMD protocol use is documented for OHCA calls.We also examine care-related procedures that are different in calls when EMD protocol use is reported and measure the calllevel application of EMD pathways.We theorize that EMD will be associated with automated external defibrillator (AED) use and bystander CPR occurrence.

Study Design, Setting, and Participants
This analysis is a retrospective study of EMS calls in the U.S. for adult (!18 years) OHCA events in 2019, using the National EMS Information System (NEMSIS) national EMS database (data elements from the version 3.4.0standard), which has been previously described (6).The EMS response type was primary response.We included cardiac arrest events prior to EMS arrival, and we only included primary EMS response with arrest etiology of medical causes (i.e., respiratory, presumed cardiac, drug overdose); events with an arrest etiology of drowning, trauma, exsanguination, medical, and "other" were excluded.We excluded patients with do not resuscitate status or specialty EMS transport.Records missing exposure (EMD) or outcome (ROSC status) were excluded.This project was determined not to be human subjects research by the local institutional review board, and it is reported according to The STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines (7).

Variables
The primary exposure was EMD use documentation, operationalized as a binary outcome of EMD or no EMD usage in an event-level analysis.The primary outcome was any ROSC during the EMS encounter, which was defined as any ROSC during the time between arrival of EMS to the end of the encounter (i.e., transition to emergency department team or death) (Table S1) (8).The secondary outcome was survival at the end of EMS event, as defined by a binary endof-event derived from multiple event fields from triage to hospital discharge, which describes the patient's disposition status as dead or alive at the time an EMS clinician stops providing care at transport destination or scene of event (9).While favorable neurologic status at hospital discharge would be a preferred, patient-relevant outcome for this study of OHCA events, the patient's neurologic status at hospital discharge is not available in the NEMSIS dataset and much of the hospital and emergency department disposition data has large amounts of missingness (9).Layperson care process outcomes included CPR provided prior to EMS arrival and AED use prior to EMS arrival.Finally, an EMS care process outcome of any resuscitation provided by EMS (includes any attempt by EMS to resuscitate patient including defibrillation, ventilation, chest compressions) was included to assess for possible underlying differences between the exposure groups (i.e., we hypothesized there should be no difference in this outcome).Covariates included patient age ( 44, 45-64, 65-84 and >85 years), sex, race/ethnicity, primary insurance, rural/urban location of incident (classified based on 2013 Urban Influence Codes) (10), initial rhythm, etiology of arrest, witness of cardiac arrest, and level of service provided by EMS for the encounter (ALS Level 1 or 2, Air, BLS Emergency, BLS Regular, Paramedic Intercept, Unknown).Down time interval was defined as the time in minutes from reported cardiac arrest event time to EMS clinician arrival on scene time.If cardiac arrest time is unknown (such as in an unwitnessed arrest), the down time interval would be missing.Data are reported by EMS clinicians, and cardiac arrest variables followed the standard Utstein cardiac arrest criteria (6).

Data Sources/Measurement
The data source was the NEMSIS national EMS database, which is the largest public prehospital dataset in the US (6,11).This dataset is a convenience sample of EMS calls submitted by 10,062 EMS agencies in 47 states for over 34 million EMS activations (6).Data received from EMS agencies are checked for completeness and consistency by the NEMSIS team.Data not passing validation and data cleaning are dropped from the dataset prior to public release (12).

Quantitative Variables
Categorical variables were grouped based upon available categories in the NEMSIS dataset.For continuous variables, the linearity between the variable and the log odds of ROSC was assessed.If the assumption of linearity was met, the variable remained continuous in the model.Otherwise, it was categorized (Table S2).

Missing Data
Subjects with missing exposure (EMD) or outcome (ROSC) information were excluded.Subjects with missing covariate data remained in the study.As most covariates with missingness had missingness greater than 10%, imputation methods were not used, and a "missing" category was included for these covariates in the main analysis models.Sensitivity analyses for missing data assumptions included a complete case analysis.

Study Size
Assuming a baseline proportion of 29% achieving ROSC (13,14), 30% receiving EMD, and standard error rates (a ¼ 0.05 with a two-sided test and b ¼ 0.20), the minimal detectable difference is a risk ratio of 0.968 with the sample size of 90,000 patient encounters.

Statistical Methods
The characteristics of subjects and the cardiac arrest events were described using overall summary statistics and summary statistics by group.Percent differences with 95% confidence intervals were used to compare characteristics by group to allow for assessment of clinically significant differences in this relatively large sample.Unadjusted analyses of the differences in outcomes between groups are presented as percent differences and unadjusted logistic regression.We adjusted estimates of the association between EMD and outcomes for patient and cardiac arrest characteristics (primary payer, age (categorized), sex, race, urbanicity, first cardiac rhythm, cardiac arrest etiology, and witnessed arrest) multivariable logistic regression.We additionally adjusted for down time interval in a sequential fashion (planned a priori), as there was missingness in the down time interval variable.Covariates were specified a priori based on purposeful selection using hypothesized causal relationships informed by scientific literature and biologic mechanisms.All analyses were performed in SAS version 9.4.

Secondary & Sensitivity Analyses
In a secondary analysis, to assess a "dose-response" relationship, the exposure of EMD for each OHCA event was redefined from a binary variable (yes/no) to a four-category variable (EMD with instructions, EMD with unknown instructions, EMD with no instructions, no EMD).This variable was self-reported by the EMS agency based on answering whether EMD was performed for this event with the following choices: "No," "Yes, With Pre-Arrival Instructions," "Yes, Without Pre-Arrival Instructions," or "Yes, Unknown if Pre-Arrival Instructions Given."As many subjects had missing covariates, a sensitivity analysis was performed using a complete case analysis that excluded subjects with missing covariates.Additionally, in the main analysis ROSC was defined as any ROSC, so in a sensitivity analysis we assessed the association of EMD with "sustained ROSC," or ROSC for 20 min, which is a commonly used definition for ROSC in many registries (3,15).Finally, as a sensitivity analysis we calculated an E-value to estimate the minimum strength of association needed for an unmeasured confounder to explain the association found (16,17).

Association of EMD Recorded Use and Bystander and EMS
Care Processes.There were care processes that occurred more frequently in the EMD group, including behaviors before and after EMS arrival (Table 2).CPR before EMS arrival (63.8% vs. 55.1%,%diff 8.6, 95% CI 7.9 to 9.3) and AED use before EMS arrival (28.3% vs. 26.3%,%diff 2.0, 95% CI 1.4 to 2.6) were both more common in the EMD group.Both CPR and AED use before EMS arrival remained associated with EMD use after adjustment for patient and event characteristics and down time interval (aOR 1.39, 95%CI 1.34 À 1.44; aOR 1.09, 95%CI 1.04 À 1.13, respectively).

Sensitivity Analyses
In a dose-response analysis, the binary exposure of EMD was re-categorized into no EMD (26.9%),EMD with instructions (35.9%),EMD without instructions (1.7%), and EMD with unknown instructions (35.5%).EMD with instructions was associated with the greatest odds of ROSC (aOR 1.17, 95%CI 1.12 À 1.23), and EMD with unknown instructions had an attenuated relationship with ROSC (aOR 1.08, 95%CI 1.03 À 1.14) (Figure 2).There was no significant association between EMD with no instructions and ROSC, although the confidence interval is large because this exposure was rare and the point estimate is similar to EMS with unknown instructions (aOR 1.11, 95%CI 0.96 À 1.28).Results were similar for the outcome of survival.When assessing the outcome of sustained ROSC for 20 min, there remained an association between reported EMD use and sustained ROSC (Table 2).In the complete case analyses (Table S4), the adjusted odds of both ROSC and survival were similar.The sensitivity analysis using the E-value suggests that an unmeasured confounder with an aOR of 1.41 with both EMD and ROSC would be required to explain the observed effect size.

Discussion
In this analysis of OHCA from the largest national sample of public prehospital data, we identified documentation of EMD protocol use was associated with higher incidence of ROSC.This finding is relevant, because cardiac arrest is a time-sensitive condition, and the interval prior to EMS arrival is a particularly vulnerable time for cardiac arrest patients.High-performing systems have a model of emergency activation that involves not only the dispatch of emergency responders but also focuses on providing telephone instructions to the callers (18).These telephone instructions are given by emergency telecommunicators who are performing a myriad of tasks, and task saturation is a risk that can impede the delivery of instructions (19).Any targeted intervention to improve system-wide performance in OHCA must ensure that lay person responder actions are guided and supported by emergency telecommunicators, and EMD protocols may be a key component in achieving that goal (20)(21)(22)(23).
Also noted in our study, increased CPR performance and AED use before EMS arrival was associated with recorded use of EMD, which is important because bystander CPR has been independently associated with improved ROSC and survival in multiple studies (21)(22)(23).Siman-Tov et al. described that ROSC on scene was achieved nearly twice as often in patients whose callers accepted "dispatcher assisted CPR" (DA-CPR) (24).Additional studies have shown that DA-CPR (which can be facilitated by using EMD protocols) not only improves the initiation rates of bystander CPR, but also CPR quality (25,26).
The use of an AED has also been shown to be effective in OHCA care, with prehospital AED deployment nearly doubling survival from OHCA (27)(28)(29)(30).Furthermore, AED use paired with bystander CPR has been shown to be even more effective than either alone (30).This finding also provides a feasible potential mechanism for our main association between EMD and ROSC.
Regional variation in EMS performance remains one of the biggest challenges faced by EMS systems.There are many obstacles in rural OHCA responses including fewer emergency telecommunicators staffing the dispatch center which may lead to more difficult multi-tasking (call-taking, dispatching, providing EMD instructions), longer EMS response intervals, fewer responders, and longer transport intervals to definitive care.In addition, EMD was more prevalent in urban than rural EMS systems.Despite adjusting for longer down-times and fewer witnessed arrests, adjusted OHCA outcomes in rural areas remain lower than urban areas.Connolly et al. described in their Nova Scotia experience that patients with OHCA were 107% more likely to survive in non-rural areas than those in rural areas (31).These findings were supported by Peters et al. who studied 64,489 cardiac arrest incidents and reported that rural patients had lower odds of favorable outcomes compared to their urban counterparts (8).Nichol et al. suggested that if OHCA survival in North American EMS agencies could be improved to the maximum observed EMS survival, an estimated 15,000 annual premature cardiac arrest deaths could be prevented.
Our study showed that the association between EMD and ROSC persisted even after adjusting for rural incident location and downtime.An opportunity to narrow the outcome gap could be for rural areas to adopt EMD protocols for their emergency telecommunicators to improve CPR use and quality by lay responders using standardized emergency telecommunicator instructions (32).Robust research has

Limitations
Our study has several limitations.First, we are unable to infer causality with this observational study design.The sensitivity analysis of dose-response relationship between EMD  and lay cardiac interventions and known temporality (exposure before outcome) help support the theory that EMD is responsible for improved care, but unmeasured confounders may still exist (e.g., AED availability for the outcome of "AED use prior to EMS arrival" and accreditation status of EMD protocols used).Second, NEMSIS data are at the run level, meaning we are unable to account for duplication or use hierarchical modeling to account for similarities between dispatches from the same agency or dispatch center.There may be EMS runs where two agencies responded to the same call, which are counted as two distinct observations, or a patient who had multiple EMS calls where we do not have the ability to account for non-independence of all observations.While the true magnitude of duplication is unknown, one state dataset found duplication occurred in <0.5% of general EMS calls (35).Third, many variables in NEMSIS dataset (particularly race and ethnicity) are clinicianreported and may be estimated based on the clinician's best guess.For the primary exposure variable of EMD, this variable can be entered directly from the computer-aided dispatch or entered from the individual EMS agency, which could introduce misclassification bias.Variables, such as down time interval, are likely to be estimates and not perfectly reliable even when reported.In both cases, we hypothesize that the misclassification bias would be non-differential by outcome and therefore would bias the results toward the null.Additionally, because NEMSIS uses a convenience sample, agencies that report to the data set may differ from those that do not, which affects generalizability.For example, it is possible that agencies reporting to NEMSIS may be more likely to have EMD protocols and have protocols that are accredited, so this study may overestimate the percent of OHCA events where EMD protocols are used in the US.Finally, NEMSIS data include significant missingness, which may lead to selection bias.

Conclusion
Reported use of EMD procedures was associated with increased ROSC, increased bystander CPR, increased AED use, and increased end-of-event survival in adult patients experiencing OHCA.Future work will focus on ways of optimizing EMD-based coaching as a strategy to improving care very early in OHCA resuscitation.
attempted to define the most effective lay interventions in OHCA, and widespread public CPR training programs, promotion of AED use, and standardized use of EMD may improve implementation of these overall quality improvement strategies(33,34).Future work should focus on the most efficient tactics to coach emergency callers effectively, so that even novice lay rescuers can improve cardiac arrest outcomes.Implementation and dissemination science are critical to optimizing this earliest link in the chain of survival, and the interventions delivered during this vulnerable time may have outsized effects on survival.

Figure 2 .
Figure 2. Association of clinical outcomes with EMD by instructions type.A (Top) Main analysis showing association of any EMD with ROSC and end-of-event survival; B (Bottom) Secondary analysis of EMD levels based on instructions with ROSC and end-of-event survival.Adjusted for age, sex, race/ethnicity, primary insurance, rurality, initial rhythm, etiology of arrest, and witnessed arrest.EMD ¼ emergency medical dispatch; ROSC ¼ return of spontaneous circulation.

Table 1 .
Subject demographics by EMD status.Down time interval missing for 28,914 cases; N ¼ 67,355 (18,829 no EMD and 48,526 EMD).The % difference report for this variable is the difference in medians not proportions. a

Table 2 .
Clinical and process outcomes by EMD status.NEMSIS binary mortality was not determined for 9,916 subjects (7.290 EMD and 2,626 No EMD); this is in line with original estimates from outcome derivation paper and these subjects are excluded from analyses using this outcome.b Not recorded for 4576 subjects (3127 EMD and 1449 no EMD), and these are excluded.c Not recorded for 1552 subjects (1001 EMD and 551 no EMD), and these are excluded.d Not recorded for 4079 subjects (3076 EMD and 1003 no EMD), and these are excluded.Resuscitation attempted included defibrillation (25.7% of total), ventilation (37.9%), and compressions (18.8%).e Adjusted for age, sex, race/ethnicity, primary insurance, rurality, initial rhythm, etiology of arrest, and witnessed arrest.f This model includes covariates and down time interval.Down time interval is missing in 30% of subjects.Sample size for ROSC and survival indicator models are 67,355 and 60,154, respectively. a