Methods

Results

Between June 1, 2004, and January 31, 2007, emergency dispatchers offered telephone CPR instructions for 1700 patients (Figure). Fifty-five percent (938 of 1700) of patients were in cardiac arrest on EMS arrival, with 774 receiving resuscitation efforts by EMS personnel. Forty-five percent (762 of 1700) were not in cardiac arrest. Of the patients not in arrest with a known level of CPR performance, 46% (313 of 686) received chest compressions from a bystander as a consequence of dispatcher assistance. Overall, 18% (313 of 1700) of dispatcher-assisted CPR instruction resulted in bystander chest compressions for patients not in cardiac arrest.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Flow diagram of subject eligibility.

Table 1 presents the characteristics of patients not in arrest who received dispatcher-assisted CPR stratified by chest compression status. Overall, the average age was 57 years; just over half were male. The most common condition as assessed by EMS was a drug or alcohol overdose (21%), followed by seizure (18%), syncope (15%), cerebrovascular event (11%), and hypoglycemia (9%). Those who received chest compressions were more likely to have experienced an overdose and less likely to have had a syncopal episode or other illness than those who did not progress to chest compressions (P<0.0001). The large majority were transported to the hospital by EMS (85%), most commonly by paramedics. Overall, half of the patients not in cardiac arrest were randomized to chest compression alone, and half were randomized to chest compression plus ventilations. However, a larger proportion of patients who received chest compressions were randomized to the chest compressions–only arm (P<0.0001). When stratified by randomization status, 57.1% (194 of 340) of chest compression–only CPR instruction compared with 34.4% (119 of 346) of chest compression plus ventilation instruction resulted in bystander chest compressions for the patient not in cardiac arrest. The median duration of bystander CPR for patients who progressed to chest compressions, as determined from our review of a random 20% sample of cases, was 91 seconds, with a minimum duration of 1 second and a maximum duration of 405 seconds (interquartile range, 44 to 189 seconds).

Hospital chart reviews were completed for 247 of the 281 patients (88%) who received dispatcher-assisted chest compressions and were subsequently transported to the hospital. We were unable to review 34 hospital charts because the prehospital data could not be directly matched with the hospital record. Analyses indicated that the EMS diagnostic profiles, transport level, and Glasgow coma scores at the end of EMS care were comparable between the patients with and without an available hospital chart. Overall, one third of the patients transported were treated in the emergency department and released; two thirds were admitted to the hospital (Table 2). There were no significant differences in hospital admission status or discharge disposition according to randomization group (P=0.31 and P=0.89, respectively).

No deaths were attributed to bystander CPR. Of the 247 patients whose charts were reviewed, 22 (9%) experienced chest discomfort or pain as a probable consequence of dispatcher-assisted bystander CPR (Table I of the online-only Data Supplement). Four patients (2%) sustained injuries as a result of the bystander CPR: 1 patient had a clavicular fracture; 1 patient had confirmed rib fractures; another had suspected rib fractures; and 1 patient had a hip fracture caused by the bystander pulling the patient to the floor to begin CPR. There were no reports of sternal fracture, organ laceration, or other injuries. Another 3% (7 of 247) experienced chest discomfort or pain, and 1% (2 of 247) suffered injuries that were a possible consequence of bystander CPR (Table II of the online-only Data Supplement). Of the 2 injured patients, 1 had rib fractures, and 1 had tracheal bleeding possibly resulting from the CPR. Review of this case indicated that the tracheal bleeding was self-limited and required no additional intervention. No other injuries were noted for these 9 patients. Analyses suggested that the 31 patients who experienced pain or injuries as a likely or possible consequence of bystander CPR were similar in age, randomization status, and EMS diagnostic profiles to all other patients not in arrest who received chest compressions. However, injured patients were significantly more likely to be female than those not injured (68% and 41% female, respectively; P=0.005).

Telephone interviews were attempted with the 313 patients who received chest compressions from a bystander with dispatcher assistance. Seventy-three patients (23%) consented to participate, 8 (3%) refused, and 5 (2%) were unable to complete the interview because of a language barrier or an inability to recall the incident. For 157 patients (50%), we had no functioning contact information, and we were unable to contact the remaining 70 patients (22%) despite multiple attempts. There were no statistical differences between those interviewed and those not interviewed according to randomization group (chest compression only, 60% versus 63%), EMS diagnostic profiles (eg, seizure, 19% versus 15%), transport level (advanced life support, 63% versus 50%), and patient outcomes (death, 6% versus 7%). Those completing the interview were significantly older than those not interviewed (60.4 versus 52.0 years of age; P=0.004). Of the 73 patients who completed the interview, 14 (19%) reported discomfort or injuries resulting from the bystander CPR. All 14 stated that they experienced soreness in their chest. In addition, 2 reported rib fractures, and 1 reported lower back pain. There were no reports of lacerated organs, internal bleeding, or other complications. Of the patients reporting injuries, 3 required treatment, including pain medication, breathing treatments, or physical therapy. No patients required hospitalization for their injuries.

Discussion

In this cohort study of dispatcher-assisted bystander CPR, nearly half of the patients for whom emergency dispatchers offered CPR instructions were not in cardiac arrest. Approximately 18% of dispatcher-assisted CPR instruction resulted in bystander chest compressions for patients not in cardiac arrest. For these patients, EMS most often determined that the patient had a drug or alcohol overdose, seizure, syncope, cerebrovascular event, or hypoglycemia. A total of 9% of these patients experienced discomfort, and 2% sustained injuries likely attributed to the bystander CPR; an additional 3% possibly experienced discomfort, and 1% possibly suffered injuries resulting from bystander CPR. However, only 2% of patients not in arrest suffered a fracture, and no patients experienced visceral organ injury or death as a consequence of bystander CPR.

Ideally, emergency dispatchers would be able to identify each case of cardiac arrest comprehensively and precisely so that CPR instructions could be provided in a timely and directed manner. The goal is to apply simple questions that can quickly identify most true arrest patients while excluding patients without arrest, given the potential risk attributed to unnecessary CPR. The optimal balance between sensitivity and specificity depends in part on the operating characteristics of the identification questions and the risks associated with CPR for those not in arrest.

The 2-question approach used in the present study to identify potential cardiac arrest is straightforward and now commonly used in various dispatch protocols. With this approach, 18% of dispatcher CPR instruction resulted in chest compressions for patients not in arrest. Although exacting comparisons are difficult, this “false-positive” rate is comparable to prior reports from communities with well-integrated and established dispatch systems.^13,14 One consideration is to alter or add questions to limit the number of these false-positive cases. However, prior investigation indicates that some patients with cardiac arrest are not identified with the 2-question approach.^6,7,9,15 Moreover, additional efforts to decrease false-positive CPR would likely delay and further reduce the yield of true arrests. For example, the current question asks whether the patient is breathing normally. In several of the current false-positive conditions, breathing is present but abnormal. False-positives would likely be reduced if the question simply asked whether the patient was breathing, but then true arrest patients with agonal respirations would also be excluded. Patients with agonal respirations are potentially the most viable arrest patients, and early identification and bystander CPR may be especially important for resuscitation.¹⁶

A second consideration when determining the optimal balance of sensitivity and specificity is the risk associated with unnecessary bystander CPR, a risk not previously well investigated. Serious injuries such as rib fracture resulting from unnecessary bystander CPR were uncommon, occurring in 2% of patients who received bystander chest compressions. This rate of injury is far lower than what has been observed in previous studies of CPR complications, which have reported rates of injury ranging from 21% to 65%.^17,18 These prior studies, however, were primarily autopsy studies conducted on true cardiac arrest patients who had undergone extended resuscitation efforts by medical personnel. Longer duration of CPR is associated with an increased risk of injury.¹⁰ In the present study, patients not in cardiac arrest received only a median duration of 91 seconds of (bystander) CPR. CPR was interrupted by the arrival of EMS or when characteristics indicating that the patient was not in arrest became apparent to the bystander or dispatcher. Moreover, simulation studies suggest that CPR by laypersons often does not produce guideline-directed compression depth and thus would presumably be less likely to cause injury.¹⁹

On the basis of evidence indicating at least comparable outcomes for chest compression alone compared with chest compression plus ventilation CPR, some but not all CPR guideline organizations advocate chest compression alone for presumed cardiac arrest with witnessed collapse in an effort to increase the proportion of patients receiving bystander CPR.³ The present investigation is set within a randomized trial of dispatcher instruction comparing chest compression alone with chest compression plus ventilation, providing the opportunity to evaluate how instruction type would affect CPR among patients not in cardiac arrest. As might be anticipated, chest compressions were more frequently provided to patients randomized to chest compression only. Given the low observed rate of serious injury to these nonarrest patients regardless of the type of CPR instruction, the choice of CPR type (chest compression alone versus chest compression plus ventilation) should depend on which instruction type provides greater benefit to the true arrest patient rather than the potential risk to the patient not in arrest.

This study has limitations. We relied primarily on hospital record review supplemented by telephone survey to ascertain adverse events related to CPR. The study used hospital records available as part of regular clinical care and did not specifically mandate chest x-ray or other diagnostic evaluations that might be directed toward complete ascertainment of CPR injury. For this reason, the injuries identified in this study should be considered clinically apparent but may underestimate the extent and amount that might have been identified with a more comprehensive ascertainment protocol. Hospital charts could not be obtained for ≈10% of patients receiving chest compressions who were transported to the hospital and another 10% who did not receive hospital evaluation. The frequency of serious injury could differ in cases that could not be reviewed, although clinical characteristics based on review of the EMS record were similar between those with and without hospital information among those transported to hospital. We did not systematically evaluate patients who were not in arrest and did not receive chest compressions. Potential exists for injury before chest compression, as illustrated by the patient who suffered a hip fracture during initial positioning, but we believe the risk of injury is lower given the absence of any chest compressions. With this consideration in mind, the current program of dispatcher CPR instruction produced serious injury as defined by fracture or documented visceral injury in 0.3% of patients (5 of 1700) for whom dispatcher CPR was initiated. Conversely, dispatcher-assisted CPR enabled bystander CPR in ≈25% of cardiac arrest patients in the study community.⁵ Finally, the results are derived from a heterogeneous community in which dispatchers have a mature experience with CPR instruction and are currently participating in a randomized trial of dispatcher-assisted CPR. Results may differ in other communities that use distinct protocols or have a different emphasis. The investigators regard the dispatchers of the study community as generally assertive and thus would not typically expect false-positive CPR instruction or CPR-induced injury to occur more frequently in most communities.

In this prospective cohort study, the patient was not in arrest in almost half of cases in which dispatcher-assisted CPR was initiated. Ultimately, 18% of dispatcher instruction resulted in chest compression for patients not in arrest. However, the frequency of serious injury in the nonarrest group was very low. Coupled with the established benefits of bystander CPR among those with arrest, these results support an assertive program of dispatcher-assisted CPR as a meaningful approach to improve survival from out-of-hospital cardiac arrest.