Potential impact of public access defibrillators on survival after out of hospital cardiopulmonary arrest: retrospective cohort study

Setting

The Scottish Ambulance Service is the sole provider of emergency pre-hospital ambulance care in Scotland. During the period of study no public access defibrillators were deployed in Scotland. Pre-hospital defibrillations were undertaken by ambulance personnel (98.9%) or general practitioners (1.1%).

Data

Ambulance crews collect data on all resuscitation attempts after out of hospital cardiopulmonary arrests in Scotland. The forms include the time from the emergency telephone call to the arrival of the ambulance crew at the arrest scene, location of the arrest, and whether defibrillation was performed. Since 1991 the presumed cause of arrest has been classified in accordance with the Utstein convention, whereby arrests are defined as due to cardiac disease if the cause is recorded as heart disease or unknown or the record is missing.4 Forms completed by hospital staff document whether patients admitted to hospital after cardiac arrest survive to discharge. Data from both sets of forms are collated in an electronic register at Glasgow University.

Cohort study

Our cohort comprised all out of hospital cardiopulmonary arrests over a seven year period from May 1991 that fulfilled the Utstein definition of a cardiac cause, were not witnessed by the ambulance crew, and did not occur in an ambulance, ambulance helicopter, general practice, dental surgery, or non-acute hospital.

We reached a consensus on individual arrest sites that were suitable for locating public access defibrillators, possibly suitable (if defibrillators were to be very widely distributed), or unsuitable. In general, sites were given the benefit of doubt as to their suitability. Patients' homes were classified as unsuitable because, although automated external defibrillators are sometimes provided for use at home by patients judged to be at high risk of cardiopulmonary arrest, they are available only for personal use by trained family members and so are not accessible to the public. Unspecified sites were classified as possibly suitable. Actual survival to discharge was calculated for each of the three types of site.

Statistical model

Some patients have asystole from the onset of arrest and are unsuitable for defibrillation by the ambulance crew or bystanders. We assumed that public access defibrillators could increase the use of defibrillation up to a maximum equivalent to the use associated with early ambulance attendance—and therefore that survival would be improved, at best, to that obtained by early ambulance attendance. In each of the three suitability groups we divided arrests into those that were attended by the ambulance in 3 minutes of the emergency telephone call and those that were not. We calculated actual survival for each subgroup.

We calculated predicted survival in the suitable and possibly suitable sites by applying the survival rate among patients attended in 3 minutes to the number of patients who experienced a delay of >3 minutes. Calculations of the predicted numbers of survivors, survival rates, and associated 95% confidence intervals were made assuming that future arrests would have the same distribution of arrest sites and attendance times as the cohort data. We calculated two predicted overall survival rates, according to location of public access defibrillators only in suitable sites or in both suitable and possibly suitable sites. We calculated standard errors and 95% confidence intervals for predicted survival rates by propagating the binomial error from estimating the survival rates in the site groups and the time from emergency call to ambulance arrival through the formula for the estimated rates.

Cohort study

Over the seven year study period, 21 481 cardiopulmonary arrests were attended by the Scottish Ambulance Service, and 15 189 fulfilled the inclusion criteria. Of these, 12 004 (79.0%) occurred in sites not suitable for the location of public access defibrillators, such as the person's home or a friend's home, 453 (3.0%) occurred in possibly suitable sites, such as buses and multistorey car parks, and 2732 (18.0%) occurred in suitable sites, such as shops, places of business, and sports centres (table 1). The 453 possibly suitable sites included 105 sites recorded only as “other” and three cases in which no information on site was recorded.

Data on defibrillation and outcome were available for 14 947 (98.4%) arrests. Of these, 8895 (59.5%) people received defibrillation from the ambulance crew (table 2). However, the proportion of people who received defibrillation varied significantly between unsuitable, possibly suitable, and suitable sites (P<0.0001). Just over half (56.2%) of the patients who had an arrest in sites not suitable for public access defibrillators received defibrillation, compared with 67.9% in possibly suitable sites and 72.9% in suitable sites. Less than 1% of patients who were not in ventricular tachycardia or fibrillation survived, irrespective of site of arrest. In all types of site, patients who received defibrillation were significantly more likely to survive. People who had an arrest in sites suitable for public access defibrillators had the highest baseline survival rate.

Overall, 70.1% of patients who were attended in 3 minutes received defibrillation, compared with 58.3% in patients who experienced longer delays (P<0.0001) (table 3). Among the 14 850 patients with complete data on ambulance response time and defibrillation, 744 (5.0% (95% confidence interval 4.7% to 5.4%)) survived to discharge from hospital.

Statistical model

The model predicted that locating public access defibrillators only in suitable sites would increase the number of survivors from 744 to 942, a survival rate of 6.3% (5.6% to 7.1%). If defibrillators were located in suitable and possibly suitable sites, the predicted number of survivors increased only slightly further to 959, a survival rate of 6.5% (5.7% to 7.2%) (table 3). In Scotland, with a population of 5.1 million, these rates equate to an average number of additional lives saved a year of 28 and 31, respectively.

A response time cut-off of 2 minutes rather than 3 minutes produced a predicted number of survivors of 1008 (6.8% (5.6% to 7.9%)) for suitable sites only and 1028 (6.9% (5.7% to 8.1%)) for suitable and possibly suitable sites. A cut-off of 4 minutes produced figures of 892 (6.0% (5.7% to 6.3%)) and 903 (6.1% (5.5% to 6.7%)).

Reducing time to defibrillation

Nearly all survivors of out of hospital cardiopulmonary arrest are in ventricular fibrillation when the ambulance arrives, because asystole and pulseless electrical activity usually imply a prolonged interval from the onset of initial ventricular fibrillation or the presence of irreversible cardiac damage.5 The proportion of patients in ventricular fibrillation is inversely associated with time from cardiac arrest, with no evidence of a threshold effect. 2 6 Therefore, any intervention that reduces time to defibrillation should increase the proportion of patients suitable for defibrillation and improve survival.

The American Heart Association proposed four strategies to reduce time to defibrillation.6 Traditional first responders other than ambulance crews (such as police and firefighters) could be provided with automated external defibrillators and trained in their use. In Scotland provision of automated external defibrillators to other first responders could double overall survival to discharge from hospital.7

Formal training could be given to personnel not in the traditional emergency services—flight attendants, security personnel, and voluntary workers. 8 9 An Australian airline installed automated external defibrillators in its aircraft and the airports it used.10 Over a period of 64 months 46 arrests occurred, 23 from ventricular fibrillation. Twenty one of these patients were successfully defibrillated and six survived. Similar results were reported in a more recent study of automated external defibrillators in a US airline.8 In a 32 month study on the use of automated external defibrillators by trained security personnel in 32 casinos, 148 people had a confirmed cardiac arrest. Defibrillators were used on 105 people whose initial cardiac rhythm was ventricular fibrillation. Overall 56 survived to discharge from hospital.

Automated external defibrillators could be provided to people at high risk of cardiopulmonary attack and to friends and family members trained in their use. Almost two thirds of all arrests occur in patients' own homes, and patients who have recently suffered a myocardial infarction or non-fatal arrest are at increased risk of cardiopulmonary arrest. However, these defibrillators would not be publicly available and would not reduce time to defibrillation among people not identified as at high risk or people in whom cardiopulmonary arrest is the first manifestation of cardiovascular disease, who account for most cardiopulmonary arrests in the population.

Finally, automated external defibrillators could be located in public places for use by the general public, most of whom will have little or no training in their use. This option is analogous to the situation in relation to fire extinguishers, which are widely distributed throughout public buildings. Public access defibrillators would need to take the form of simple, “intelligent” devices that require minimal training and maintenance and could be supported by telephone or interactive instruction.

Cost effectiveness

Making public access defibrillators as widely available as fire extinguishers would greatly increase early access but would incur considerable costs and would be problematic in terms of maintenance and avoiding misuse and vandalism. Nichol et al modelled the incremental cost effectiveness of expanding access to automated external defibrillators beyond the emergency medical services in the United States.11 Their model excluded the 25% of arrests in the United States that occur in non-urban areas. They reported a baseline survival of 7.9%, which increased to a predicted 8.7% with lay responder defibrillation and to 11.8% with the police as additional first responders. The incremental costs per patient treated were estimated as $7100 (£4590; £;7148) and $9200, respectively, and the cost per additional quality adjusted life year (QALY) was $44 000 and $27 200. The authors concluded that these costs were similar to those of a number of existing medical interventions. However, their lay responder model used costs from a police responder pilot. It is likely that the cost per QALY of a public access scheme will be much higher, owing to the large number of defibrillators needed to provide adequate coverage. Unlike automated external defibrillators used by a police service, public access defibrillators are fixed to a location, and the location of future arrests cannot be predicted precisely from the location of previous arrests.

Conclusions

Targeted placement of defibrillators in public places such as airports and shopping centres, frequented by a large number of susceptible people, could, at best, increase overall survival from 5.0% to 6.3%. These sites are the most suitable for locating public access defibrillators on grounds of effectiveness, cost effectiveness, maintenance, and training. People working in these areas could be trained in their use, rather than relying on the actions of untrained bystanders. Further expansion of the coverage of public access defibrillators to enable lay use in all potentially suitable sites would require much greater resources and would produce little additional improvement in survival. Public access defibrillators should not be provided in preference to the expansion of defibrillation given by first responders or increased cardiopulmonary resuscitation by bystanders.