This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn.

The placing of automated external defibrillators (AEDs) in selected public locations was studied.

Device.

Cost-utility analysis.

The study population comprised a hypothetical cohort of the general public.

The setting was selected public locations, such as international airports, public sports venues, golf courses, county jails, health clubs and shopping centres. The economic study was conducted in the USA.

The effectiveness and resource us data were estimated from studies published between 1985 and 2002. The price year was 2002.

The effectiveness data were derived from a review of completed studies.

A Markov model was constructed to estimate the cost-effectiveness of the two strategies in a general cohort of Americans in public locations. The model was based on a decision tree and the two main branches, corresponding to the two strategies, were similar except for the deployment of the AED device in strategy 2.

The outcomes assessed in the review were the probabilities of the following:

cardiac arrest,

AED use on arrest victim,

initial resuscitation with EMS,

initial resuscitation with AED,

surviving to hospital discharge with EMS or AED,

dying in the first year after cardiac arrest,

dying in subsequent years after cardiac arrest, and

surviving unimpaired, moderately impaired, or severely impaired.

Also assessed were:

the mean life expectancy for arrest survivors;

the quality of life in unimpaired, moderately impaired, or severely impaired persons; and

the annual probability of AED use in international airports, public sports venues, golf courses, county jails, health clubs, large shopping malls, community centres, primary care practices, hotels, restaurants, and retail stores.

Although a formal review of the literature was undertaken, specific inclusion criteria relating to the design of the primary studies were not reported.

MEDLINE was searched from 1966 to 2002 using the keywords "cardiac arrest", "defibrillation" or "defibrillator", and "emergency medical services". In addition, the bibliographic references of the identified studies were also reviewed.

Not stated.

Not stated.

Nineteen primary studies were included in the review.

When multiple estimates had to be combined, the estimate selected for the base-case analysis was derived from the study that showed the most sound and robust methodology. The other estimates found in the literature were used to create a range of values. When none of the studies was methodologically superior, the authors reached a consensus on the base-case estimate and value ranges.

Not stated.

The probability values were:

0.2 (range: 0.01 - 1) for cardiac arrest,

1 (range: 0.6 - 1) for AED use on arrest victim,

0.25 (range: 0.15 - 0.35) for initial resuscitation with EMS,

0.50 (range: 0.25 - 0.70) for initial resuscitation with AED,

0.10 (range: 0.02 - 0.20) for surviving after hospital discharge with EMS,

0.25 (range: 0.20 - 0.50) for surviving after hospital discharge with AED,

0.20 (range: 0.10 - 0.40) for dying in the first year after cardiac arrest, and

0.15 (range: 0.10 - 0.20) for dying in subsequent years after cardiac arrest.

The mean life expectancy for arrest survivors was 5 years.

The effectiveness parameters were as follows:

the probability of surviving unimpaired was 0.87 (range: 0.80 - 0.95),

the probability of surviving moderately impaired was 0.09 (range: 0.05 - 0.15),

the probability of surviving severely unimpaired was 0.04 (range: 0 - 0.10);

the quality of life was 0.85 (range: 0.7 - 1) in the unimpaired, 0.20 (range: 0 - 0.40) in the moderately impaired, and 0.1 (range: 0 - 0.20) in the severely impaired; and

the annual probability of AED use was 0.47 in international airports, 0.1 in public sports venues and golf courses, 0.09 in county jails, 0.08 in health clubs, 0.07 in large shopping malls, 0.03 in community centres, 0.01 in primary care practices and hotels, 0.001 in restaurants, and 0.0005 in retail stores.

The summary benefit measure was the quality-adjusted life-years (QALYs). These were calculated using the decision model. A discount rate of 3% was applied to those benefits incurred in the future. Utility values, as well as survival data, were estimated from the literature review. The life-years gained were also reported, but were not used as the denominator of the cost-effectiveness ratio.

A 3% annual rate was used to discount the costs occurring in the future. The unit costs were only reported separately from the quantities of resources used for AED devices. The categories of costs in the analysis were AED, hospitalisations (for those who survive to discharge or die), and future medical costs in the first and subsequent years. The AED costs were for purchase, maintenance, training, and supplies per use. The cost/resource boundary adopted in the study was unclear. The unit costs were generally estimated from published studies and ranges observed in the literature were used in the sensitivity analyses. To assess the costs of AEDs, the manufacturers were also contacted. Resource use data relied on the probability values estimated in the literature review and used as inputs in the decision model. All of the costs were inflated to 2002 values using the medical component of the consumer price index (3.9% annual inflation rate).

Statistical tests of the costs were not conducted in the base-case.

The indirect costs were not included in the analysis.

US dollars ($).

One- and two-way sensitivity analyses were conducted to deal with the uncertainty in the parameters used in the decision model. The ranges of variations for both the costs and probability inputs were estimated from the review of the literature. A Monte Carlo simulation was performed in which the probability of cardiac arrest was kept at 20% and the probability of AED use at 100%, while varying all other model inputs. The cost variables were assigned log-normal distributions while all other variables were assigned normal distributions.

Strategy 2 resulted in a gain of 0.114 QALYs (0.15 life-years) per site in comparison with strategy 1.

Strategy 2 led to an incremental annual cost of $3,400 per site relative to strategy 1.

An incremental analysis was conducted to combine the costs and benefits of the two strategies considered.

The annual cost per QALY gained with strategy 2 over strategy 1 was $30,000 ($22,700 per life-year gained).

In most of the variations carried out in the sensitivity analyses, the cost per QALY remained below the threshold of $50,000. However, some variables resulted in the cost-effectiveness ratio exceeding the fixed threshold. These were the annual probability of cardiac arrest per site, the probability of surviving to hospital discharge with EMS, the cost of hospitalisation for arrest survivors, and the life expectancy for arrest survivors.

Monte Carlo simulations showed that 87% of all the simulations cost less than $50,000 per QALY. Also, the probability that the cost-effectiveness ratio was above $100,000 was less than 1%.

A policy of deploying automated external defibrillators (AEDs) in public locations represented a cost-effective strategy in comparison with care provided by emergency medical services equipped with AEDs (EMS-D). This conclusion depends strongly on the observed probability of cardiac arrest, which unfortunately represents the most uncertain variable.

The rationale for the choice of the comparator was clear. Strategy 1 (EMS-D) represented the routine intervention for the care of individuals suffering cardiac arrest in public locations. You should decide whether it represents a valid comparator in your own setting.

The effectiveness data were retrieved from published studies and a formal review of the literature was performed. The authors reported the methodology used to identify the primary studies and to combine the estimates found in the literature. However, the design of the primary studies was not described and it was not possible to assess the internal validity of the studies. Likewise, it was unclear whether the authors considered differences across the primary studies when estimating the effectiveness. However, sensitivity analyses were conducted using the ranges of variations found in the literature.

The QALYs were used as the summary benefit measure. These appear to have been appropriate for measuring the impact of the intervention on patient survival and quality of life. Data on survival were also reported. The use of QALYs simplifies comparisons with the benefits of other health care interventions. Quality of life values were estimated from two studies that used either the EuroQol questionnaire or the Health Utilities Index. The authors noted that, to carry out a conservative analysis, the potential utility gains arising from the increased sense of security afforded by AED availability were not considered. Discounting of the benefits was relevant and was appropriately performed.

The perspective of the study was misreported in that the authors stated that costs relevant to society were included. However, the indirect costs, which would have been relevant, were not considered. Resource use data were estimated from probability values and the sources of the cost data were reported. The price year was given, as was the inflation rate used to inflate the costs. Discounting was applied since the costs were incurred over more than 2 years. The costs were not treated stochastically in the base-case, but probabilistic distributions were assigned to costs in the Monte Carlo simulations.

The authors stated that their findings confirmed the results of earlier studies. The issue of the generalisability of the study results to other settings with different rates of cardiac arrest was implicitly addressed in the sensitivity analyses. It was also addressed more explicitly in the use of probability values coming from a large metropolitan area with a diverse population, which may be representative of the overall US population. The authors stressed that, when wide ranges of values were obtained from the literature, their choices were always conservative, thus favouring strategy 1. It was noted that the cost-effectiveness of AED deployment strongly depended on the benefits of EMS-D. Consequently, in many public locations with low cardiac arrest rates, the option of encouraging the optimisation of existing EMS may represent a more efficient solution.

Due to the uncertainty in location-specific cardiac arrest rates, the study results provided interesting insights for decision-makers in terms of the cost-effectiveness of AED deployment. The analysis also suggested that the AHA guidelines may be over-conservative, and that PAD may be cost-effective even at lower rates of cardiac arrest, as shown in the sensitivity analysis. The authors noted that the results of an ongoing trial (the Public Access Defibrillation Trial) would contribute to an appropriate evaluation of the deployment of AED in public locations. Further research should assess which individuals and locations may benefit from PAD.

This study had no external funding.