Health System Costs of Out-of-Hospital Cardiac Arrest in Relation to Time to Shock
Background— Early defibrillation results in higher admission rates and healthcare costs. This study determined the healthcare resources used and related medical costs after out-of-hospital cardiac arrest (OHCA) in relation to time to shock. We assessed the incremental healthcare costs per life gained from reduction in time to shock of 2, 4, and 6 minutes.
Methods and Results— Clinical and costs data of patients in witnessed OHCA with ventricular fibrillation as initial rhythm were collected. Each patient’s time to shock was estimated and assigned to 1 of 3 categories: ≤7 minutes (early), 7 to 12 minutes (intermediate), and >12 minutes (late). Incremental cost-effectiveness analysis and Monte Carlo simulation compared scenarios of reduction in time to shock of 2, 4, and 6 minutes. Six-month survival was 22%. Mean prehospital, in-hospital, and posthospital costs in the first half-year after OHCA were €559, €6869 and €666. Mean costs were €28 636 per survivor and €2384 per nonsurvivor. Among patients shocked early (n=24), 46% survived, with costs averaging €20 253. Of the intermediate group (n=149), 26% survived, with costs averaging €31 467. Among patients shocked late (n=135), 13% survived, with costs averaging €27 781. The point estimates of the incremental cost-effectiveness ratios of reduction of time to shock of 2, 4, and 6 minutes compared with baseline were €17 508, €14 303, and €12 708 per life saved, respectively.
Conclusions— Costs per survivor were lowest with the shortest time to shock because of shorter stay in the intensive care unit. Reducing the time to defibrillation increases the healthcare costs by an acceptable amount according to current standards and is economically attractive.
Received March 16, 2004; revision received June 17, 2004; accepted June 18, 2004.
The use of automated external defibrillators (AEDs) by lay rescuers such as police, firefighters, flight attendants, and security personnel has been shown to improve survival of out-of-hospital cardiac arrest (OHCA).1–3 The cost-effectiveness of these early defibrillation programs critically depends on the incidence of cardiac arrest in relation to the required amount of defibrillators and the expected gain in time to defibrillation. Various studies have shown that AED programs appear to fall within the conventional standards of cost-effectiveness.4–9 With improvement in out-of-hospital care, more patients will survive to hospital admission, and therefore healthcare costs will increase. In 3 studies, 2 on public access defibrillation and 1 on the cost-effectiveness of AEDs onboard aircraft, healthcare costs were taken into account.4,5,9 These studies used 1 fixed price for a survivor and 1 fixed price for a nonsurvivor of OHCA, but shortening the time to defibrillation may benefit the clinical condition of the patient on arrival in the hospital and the subsequent in-hospital course.
We studied the costs of prehospital, in-hospital, and posthospital health care in a cohort of victims of OHCA in whom the time to defibrillation of each individual could be related to their healthcare costs in the first 6 months after resuscitation. Furthermore, we could determine the incremental healthcare costs per life saved, derived from scenarios of reduction in time to shock of 2, 4, and 6 minutes.
Study Area, Emergency Medical System, and Patient Selection
The study area included the city of Amsterdam and a surrounding suburban and rural area with a total population of 1.6 million contained in 885 square kilometers. Between January 2000 and January 2002, data of all patients in OHCA, identified by the emergency medical system (EMS) dispatch center, were collected prospectively in the setting of a controlled trial of the effect of the use of AEDs by first responders.10 The standard EMS consisted of ambulances manned by a team qualified to perform advanced cardiopulmonary life support. Police and firefighter first responders were equipped with an AED. When first responders with an AED arrived at the scene, they used the AED. Paramedics could take over the resuscitation by connecting their own manual defibrillator. We included patients with a witnessed cardiac arrest and an initial shockable rhythm and excluded patients below the age of 18 years and patients with a cardiac arrest of traumatic origin.
Trained data collectors were notified shortly after the ambulance was dispatched and arrived on scene by car. Data were obtained on the circumstances of the arrest, the estimated moment of collapse, witnesses, bystander cardiopulmonary resuscitation, and relevant time points by directly interviewing all persons involved. The continuous rhythm data from the AEDs and manual defibrillators were downloaded onto a laptop at the scene. Time of call was obtained from the computer of the dispatch center. Deviations of defibrillator clocks, clocks in the dispatch center, and all personal clocks were corrected by comparison with a standard radio-controlled clock. An experienced research nurse assessed the survivors at 6 months after the resuscitation and classified the patients according to the 5-point scale of Glasgow-Pittsburgh Overall and Cerebral Performance Category (OPC and CPC).
Rhythms were categorized as shockable (ventricular fibrillation/ventricular tachycardia), asystole (disorganized rhythm with amplitude <100 μV), or organized rhythm. Time to shock was defined as the time from the moment of the collapse, as estimated by the data collector, to the delivery of the first shock. Return of spontaneous circulation was defined as return of a palpable pulse for at least 15 seconds.
Medical ethics committees from the participating hospitals and EMS approved the study and considered the study exempt from having to obtain informed consent before treatment. Informed consent was obtained from patients or family members after resuscitation.
Primary End Point
Survival 6 months after the cardiac arrest was the end point of effectiveness.
Costs were calculated as the product of volumes of healthcare resources used and their 2001 unit costs. Only medical costs were taken into account. An account was kept of all volumes used during hospitalization from medical records per patient, as follows: days of admission in intensive or coronary care units, cardiology, neurology, or other wards; diagnostic investigations, including laboratory testing; and therapeutic procedures. Volumes of postdischarge health care were recorded by assessing medical records and interviewing the general practitioner and the patient or family members at 6 months after the OHCA. These volumes included days of care in nursing homes and rehabilitation centers, readmissions to hospital, visits to outpatient hospital clinics and to the general practitioner, treatments by a physiotherapist, and hours of home care.
Because of the observed high volumes and the potentially costly nature of inpatient hospital days, standard cost-accounting procedures based on the 2001 hospital ledger were followed to estimate true costs of personnel involved and materials used. Additional costs of overhead per inpatient day were derived from existing Dutch healthcare costing guidelines.11 The total costs per inpatient hospital day were differentiated for academic and nonacademic hospitals
Costs of the outpatient hospital consultations and of home support were based on estimates of the true costs of these services available from 2 national standard cost-accounting studies.12 Costs of ambulance care, emergency room treatment, out-of-hospital consultations by the general practitioner and the physiotherapist, and days in a nursing home were based on national census data (eg, total expenditures for general practitioner care divided by the number of insured patients and the average number of consultations per insured patient).13
For the less frequently observed major diagnostic procedures (eg, coronary angiography, electroencephalography) and interventions (eg, angioplasty, coronary bypass surgery, implantable cardioverter defibrillator implantation), either standard charges were used as a surrogate for costs or cost data from prior research14 were used, adjusted for the year 2001 using price indices for the healthcare sector. Costs are reported in 2001 Euro (€).
Patients were classified in 3 groups according to time to first shock. Because the American Heart Association/International Liason Committee On Resuscitation guidelines consider shock delivery within 5 minutes after call a high-priority goal, and we observed a median time from collapse to call of 2 minutes, we defined group 1 as shock delivery within 7 minutes after the collapse.15 We divided the remaining patients into 2 approximately equally numbered groups: shock delivery between 7 and 12 minutes (group 2) and >12 minutes (group 3) after the collapse. For these groups, 6-month survival and mean total healthcare costs during the first half-year after the resuscitation were calculated.
Scenarios were made of reduction in time to shock of 2, 4, and 6 minutes. To apply a scenario, we subtracted 2, 4, or 6 minutes, respectively, from the measured time from call to shock and reclassified each patient to 1 of the 3 “time to shock” groups, with the associated survival and costs. The overall survival and costs after resuscitation were calculated for each scenario.
Statistical Analysis and Sensitivity Analysis
Unit costs were applied to the volume data of the individual patients to arrive at cost per patient. Descriptive statistics were used to express the use of healthcare resources and related costs for the 3 groups in time to shock. To test differences in costs between the 3 groups and subgroups of survivors and nonsurvivors, the Jonkheere-Terpstra test was used (a nonparametric test for ordered differences among groups16). Point estimates of average healthcare costs and of the probabilities of survival of the 3 groups were used to calculate the incremental cost-effectiveness ratios of the scenarios of reduction in time to shock of 2, 4, and 6 minutes versus baseline. To account for data uncertainty, a Monte Carlo simulation was performed that consisted of 500 first-order runs, each representing 500 second-order runs.17 The simulation used the observed6 distributions of healthcare cost data for the patients alive or deceased in the 3 groups. The probabilities of patients being alive in each group were assumed to follow β-distributions.18 Incremental cost-effectiveness scatterplots and acceptability curves for willingness to pay values up to €100 000 are presented for each comparison of time-reduction scenarios with baseline.
To investigate the robustness of the present results regarding changes in the underlying unit costs, we performed an additional Monte Carlo analysis using different unit costs for the important units derived from published data from another Dutch cost-effectiveness analysis, after price indexing, for stenting versus balloon angioplasty.19
A total of 583 consecutive patients were enrolled during the 24-month study period. Of these, 331 patients (57%) had an initial shockable rhythm, 23 of which were nonwitnessed. Therefore, 308 patients were included in the analysis. Baseline and process characteristics of these patients are shown in Table 1.
Of the 308 patients, 144 (47%) were admitted to the hospital. Seventy-two patients (23%) survived to hospital discharge. At 6 months, 67 (22%) of 308 patients were still alive. In 44 (66%) of 67 of the survivors, the CPC and OPC scores were 1. A CPC and OPC score of 2 (some limitation) was scored in 15 patients. Eight patients (10%) had a CPC and/or OPC score of >2, which indicated poor cerebral and/or overall performance.
The Kaplan-Meier survival curves for the 3 groups for time to shock are shown in Figure 1. There was a significant difference in survival at 6 months (P<0.001).
In Table 2, the volumes and unit prices of healthcare utilization during prehospital, in-hospital, and postdischarge treatment are shown separately for the survivors and nonsurvivors. The lengths of stay in the different departments for all 144 patients are shown in Figure 2.
Costs and Cost-Effectiveness
Mean prehospital, in-hospital, and posthospital costs in the first half-year after OHCA were €559, €6870, and €666, respectively. For survivors and nonsurvivors, total healthcare costs were €28 636 and €2383, respectively. The distribution of the costs data was highly skewed, which is typical for costs data; few patients incurred particularly high costs. Table 3 shows the costs for survivors and nonsurvivors for the 3 groups in relation to time to shock. The difference between the in-hospital nursing day costs for the survivors in the 3 groups was significant (P=0.045). Survivors who were defibrillated more rapidly needed fewer days of intensive care; the mean duration of intensive care unit (ICU) stay for early, intermediate, and late time to shock was 1.4, 2.5, and 4.4 days, respectively (P=0.02).
The point estimate of the incremental cost-effectiveness ratios for the scenarios of reduction in time to shock by 2, 4, and 6 minutes related to baseline were €17 508 (median €13 921; IQR €1886 to €30 654), €14 303 (median €13 227; IQR €4916 to €27 982), and €12 708 (median €12 326; IQR €4996 to €30 429), respectively. These costs represent the additional costs for health care during the first half-year for each life saved as a consequence of the reduction in time to shock.
Figure 3 shows cost-effectiveness planes based on the Monte Carlo simulation for each scenario with a 95% CI and the corresponding acceptability curves for different willingness-to-pay values up to €100 000. The Monte Carlo analysis based on price-indexed unit costs from Serruys et al19 did not deviate notably from the reported results.
This study measured, in patients in OHCA found in ventricular fibrillation, the healthcare costs in the first 6 months and related these costs to the interval between collapse and first shock and found lower in-hospital nursing day costs in the earliest-shocked patients. Although overall medical costs for survivors were higher than for nonsurvivors, we demonstrated that reduction in time to shock, which results in a higher proportion of survivors, was still cost-effective and perhaps even cost-saving.
Prehospital costs for survivors and nonsurvivors were of similar magnitude; in-hospital costs for nonsurvivors were much lower than for survivors, both for nursing-day costs and diagnostics and intervention. The majority of patients dying in-hospital do so within the first week of admission to the ICU, whereas costs of complex diagnostic and therapeutic interventions are spent on survivors later in their hospital stay. An important observation was that survivors who received an early shock had significantly lower in-hospital nursing-day costs because of significantly fewer admission days in the ICU. The lower cost from in-hospital nursing care for the early-defibrillated survivors is not reflected in the total cost in the early-defibrillated group. With late defibrillation, the high proportion of “cheap” nonsurvivors has a large favorable impact on the total costs. Conversely, the high total cost of medical care in early-defibrillated patients is still relatively low because of the high proportion of survivors.
Monte Carlo simulation showed that system improvements resulting in a time gain of 2, 4, or 6 minutes have a bias-corrected 0.95 probability of being cost-effective at willingness-to-pay values per life saved of, respectively, €55 329, €54 453, and €60 580, based on healthcare costs in the first half-year after the resuscitation. There is even a 20% to 30% chance in each scenario that saving a live is cost neutral or cost saving.
In general, wealthier countries are willing to pay more than poorer countries for 1 gained life-year. In the United States, a cost-effectiveness ratio of <$50 000 per life-year gained is regarded as economically attractive; in the Netherlands, that number is <€20 000 per life-year gained. A cost-effectiveness ratio of >$100 000 per life-year gained is generally regarded as economically unattractive.22 The willingness to pay for a resuscitation presented in this study applies to 1 whole life saved, but the cost measurement was limited to 6 months, at which point virtually all costs of the resuscitation and related medical care had been incurred. We assumed that medical costs after 6 months will be similar to those of cardiac patients who have never been resuscitated. The expected long-term survival among OHCA survivors was recently described with an observed 5-year survival rate of 79%.23 The quality of life, measured by the health utility index of survivors of OHCA as described by Stiell et al,24 was measured as 0.80 (IQR 0.50 to 0.97), which compares well with age-adjusted values for the general population (0.83). When we conservatively estimate that an OHCA survivor lives on average another 4 quality-adjusted life-years and consider that the average yearly costs will probably decrease again after the “expensive” resuscitation period, all reductions in time to shock appear attractive according to standards for cost-effectiveness.
Improvements in the EMS have proved feasible.1,2,25 The Casino project reported an extremely large improvement in response time of 6.4 minutes compared with conventional ambulance service.1 In the controlled police study in Miami-Dade County, Florida,2 police were equipped with AEDs, and the time from collapse to arrival decreased by 2.7 minutes. The Piacenza Progetto Vita study showed a decrease in time to arrival of 1.4 minutes.25 Our controlled study showed a decrease in time to shock of 1.7 minutes.10 It is likely that, depending on dispatch of ambulances and first responders, programs will result in modest improvements in median time to shock. Even then, the proportion of patients “shifting” to the shortest time to shock may increase substantially.
For obvious reasons, patients in the present study could not be randomized as to the time to shock, and the groups were created retrospectively. We found no differences with regard to baseline characteristics between the 3 groups; therefore, it is unlikely that the differences in costs between groups were the consequence of the baseline states of the patients.
In this analysis, only the costs and benefits of patients with an initial shockable rhythm were included, because the survival benefit of early response is to be expected mainly in patients with an initial shockable rhythm. A shorter response time is associated with higher probability of an initial shockable rhythm,26 and therefore the actual benefit of early response could be even larger than reported.
The present study included only medical costs. Nonmedical costs such as costs of sick leave were not taken into account; the average age of the patients was 66 years, and most of them had already left the workforce.
The way earlier defibrillation can be achieved is not addressed in this study, will be highly dependent on the local situation, and cannot be easily included in a model. It could involve such things as changes in dispatching, organization of EMS, training of first responders, purchase and depreciation of AEDs, and material and personnel costs. From the perspective of society, these costs must be added to the healthcare costs to fully describe the costs involved.
Dutch costs may not be directly transferable to costs in other countries because of differences in costs of healthcare resources and differences in practice patterns and resources used. The hospital costs for 1 survivor as found in the present study (averaged over all shock delays) were €24 921, lower than the fixed hospital costs per survivor assumed by Nichol et al,5,9 whereas the cost of hospitalization for 1 nonsurvivor were in the same range. However, the composition of the various cost elements will be similar between countries, with the possible exception of ICD implantation policies. In the present study, 28% of survivors received an ICD, which reflects the practice of the years 2000 to 2002, during and after which the rates of ICD implantation tended to increase. Therefore, we performed a recalculation for a situation in which 70% of survivors received an ICD, based on information from the AVID registry that 29% of all patients had an OHCA in the setting of a Q-wave acute myocardial infarction, which in general is not accepted as an ICD indication.27 Cost-effectiveness ratios for this 70% ICD implantation rate in the 3 scenarios of reduction in time to shock of 2, 4, or 6 minutes are €32 470, €31 106, and €31 058, respectively. Although such an implantation policy results in an increase in the cost per survivor, reducing the time to defibrillation remains economically attractive.
In conclusion, costs per survivor were lowest with the shortest time to shock, mainly because of less use of ICU beds. The increase in healthcare costs resulting from increased survival by shortening the delay to defibrillation is low by the standards of health economics. Shortening of time to shock is economically attractive from the perspective of healthcare economics and may even be cost-saving.
This study was supported by the Netherlands Heart Foundation (grant 98.179), the Dutch Ministry of Health, Welfare and Sport (grant LV 359042), and the Dutch Ministry of the Interior and Kingdom Relations (grant 1074808).
Myerburg RJ, Fenster J, Velez M, et al. Impact of community-wide police car deployment of automated external defibrillators on survival from out-of-hospital cardiac arrest. Circulation. 2002; 106: 1058–1064.
Nichol G, Hallstrom AP, Ornato JP, et al. Potential cost-effectiveness of public access defibrillation in the United States. Circulation. 1998; 97: 1315–1320.
Nichol G, Valenzuela T, Roe D, et al. Cost effectiveness of defibrillation by targeted responders in public settings. Circulation. 2003; 108: 697–703.
Van Alem A, Vrenken R, De Vos R, et al. A controlled clinical trial in the use of the automated external defibrillator by first responders in out-of-hospital cardiac arrest. BMJ. 2003; 327: 1312–1316.
Oostenbrink JB, Koopmanschap MA, Rutten FFH. Handleiding voor kostenonderzoek, methoden en richtlijnprijzen voor economische evaluaties in de gezondheidszorg. Amstelveen: College Voor Zorgverzekeringen; 2000.
Chamuleau S. Intercoronary Derived Physiological Parameters for Clinical Decision-Making in Patients With Multi-Vessel Coronary Artery Disease [thesis]. Amsterdam, Netherlands: University of Amsterdam; 2001: 116–133.
Guidelines 2000 for cardiopulmonary resuscitation and emergency cardiovascular care, part 4: the automated external defibrillator: key link in the chain of survival. Circulation. 2000; 102: 60–76.
Siegel S, Castellan N. Non-parametric Statistics for the Behavioral Sciences. New York, NY: McGraw-Hill; 1988: 216–223.
TreeAge Software Inc. DATA 3.5 for Healthcare User Manual. Williamstown, Mass: TreeAge Software, Inc; 1999.
Fryback DG, Chinnis JO, Ulvila JW. Bayesian cost-effectiveness analysis: an example using the GUSTO trial. Int J Technol Assessment Health Care. 2001; 17: 83–97.
Serruys PW, de Bruyne B, Carlier S, et al. Randomized comparison of primary stenting and provisional balloon angioplasty guided by flow velocity measurement. Circulation. 2000; 102: 2930–2937.
Paris M. Organization of Economic Co-operation and Development (OECD): main economic indicators: purchasing power parities. Available at: http://www.oecd.org/std/ppp. Accessed September 14, 2004.
Campbell M, Torgerson D. Bootstrapping: estimating confidence intervals for cost-effectiveness ratios. Q J Med. 1999; 92: 177–182.
Mark DB, Hlatky MA. Medical economics and the assessment of value in cardiovascular medicine: part I. Circulation. 2002; 106: 516–520.
Stiell I, Nichol G, Wells G, et al. Health-related quality of life is better for cardiac arrest survivors who received citizen cardiopulmonary resuscitation. Circulation. 2003; 108: 1939–1944.
Capucci A, Aschieri D, Piepoli MF, et al. Tripling survival from sudden cardiac arrest via early defibrillation without traditional education in cardiopulmonary resuscitation. Circulation. 2002; 106: 1065–1070.